Space Station Sim Dev Dynasty

[SportsDino]

So this all started in the thread 'Economic Simulation Games' (not to be confused with Erotic Stimulation Games which undoubtably would sell better).

Economic Simulation Games - Front Office Football Central

-----

Well this dynasty is about taking it from a rough idea and into a hopefully amusing game. I'm starting it a little past scratch, and I'll skimp on technical details to avoid boring people to death, but it might be interesting to watch what goes through my head. Now I'm not a pro-game developer, this is a garage operation, and its also a hobby, so do not take any of this at too high a standard and er... I'm sure the people out there who actually have made games will laugh their ass off at parts! Suggestions are welcome, but no guarantees I will listen (I am a stubborn bum for the most part).

-----

So the goal:
A complete game, not necessarilly game studio standard, but it shouldn't look incomplete. I'm no graphics wizard, but they should be clean, functional, and at least run smoothly for gameplay purposes. My core concern is gameplay, I'm not looking to develop the deepest plot a game has ever seen (though I love to write), nor the prettiest, but it should do something unique and do it well. I am hoping it is fun, but it will be a matter of taste and of course execution.

How this might pan out:
Unsure at the outset, but my current scheming is to build the entire engine with a sort of wireframe 2d graphics approach and no frills interface (basically a pumped up text game) and release a playable engine demo as a major public milestone.

From there as much gameplay beta I can convince people to do on essentially an ugly but hopefully increasingly complete game to get the mechanics right, before moving on to translating that engine into the 3d space universe and interface I want for the finished game.

At all points the game will be free, I will of course prefer people to only use the packages I distribute or permit (for stat tracking and quality concerns). It will not be an open source project, modding of things I can easily support will be available, but I don't owe people anything in that regard. I am deliberately nerfing/obfuscating the AI's abilities to the level I need to support just this game, because if I do enjoy this I might want my next project to be for-profit (if you can do anything well you should get paid for it).

So that is the overview of what I intend to deliver, next post, some abstracts about the game!

[SportsDino]

Game Abstract:

The game is a space station simulation game, a blend of multiple gameplay concepts but basically in the genre with SimCity, Civilization, or Railroad Tycoon.

The player character is a young, somewhat wealthy, character sent out from Earth with a fistful of credits to a station to make their name. The initial conditions are somewhat similar to Tropico, you will have something that vaguely corresponds to 'backgrounds' and environment the station is within.

The game has two eras of gameplay, solar system and interstellar, with the latter starting the game at a more technologically advanced state with an already developed game world. The basic theme is Pioneer Era in Space, the basic game starts in a newly minted station with a sort of small town atmosphere, although options to start with a more developed station can again be set in the initial conditions.

The style of the game is a political-economic simulator. You are essentially trying to optimize your personal holdings/influence within that particular station. The game is set entirely within the confines of that singular station, although there is an entire simulated galaxy out there that will also lend an inter-station component to the gameplay. The original design does not include multi-station play... you interact with Earth or other stations only through communications and trade (and well potentially conflict, although this is not a military sim). The other entities are basically simple-simmed, they do not contain the same robust internal simulation of the player's station.

Like many other strategy games, building, research, and production are important gameplay mechanics... but, I'm hoping these will be pursued in a somewhat unique way I will describe in more detail later. However, that is not the focus of the game... often those are the main tools of your basic strategy game. Instead, this game attempts to focus on the tools of politics and economic planning as its primary interaction. Hopefully I get it right and its still fun, but its something that needs to be explained probably since I think it will be a bit out of the norm.

Finally, the game is built around a sandbox engine, but the final graphical version of the game will include a story campaign that takes a space station from its first founding in the Sol System (you get to pick where though, be it orbit of Earth or Pluto!) through to the end of the tech tree. This is one continuous game, with set story events that will shake things up in a variety of ways and contain a few overarching storylines with branching.

Once the campaign mode is complete all of the events and characters will be unlocked in an advanced sandbox mode, along with an additional set of crazy events. These will all be parameterized and randomized, so eventually you'll recognize all of them... but they won't be immediately recognizable in each playthrough instance (i.e. characters names will be randomized, some of the actions/varaibles will be tweaked so even if it was in the story campaign in the sandbox mode it could occur under very different circumstances so you can't easily predict it). It won't be perfect, think Left4Dead style rearrangement of the map each time.

Altogether, I'm hoping the sim engine is deep enough to make a good sandbox game, while the story mode is more of a tweaked playthrough built to deliberately challenge (as in average difficulty progression is: tutorial mission, sandbox, advanced sandbox, story). I hope to give a good thrill in the campaign, but the meat and potatoes as it were rest in the engine itself being amusing time sucker like SimCity or Civ.

Up next, the central play mechanic! Hopefully friday!

[SportsDino]

Ah hell, I'm bored and got a few minutes, so I'll give it a start.

-----

In this game, your character is not a god. You don't have perfect observability, you can't place buildings or objects or cities upon command. You don't manage producing troops, or positioning them.

Your avatar is a dude, a wealthy dude with some starting connections, but to get anything done in this game you need to convince people to do what you want. The game centers around conversations, negotiations, and policy design... even more so than logistical design.

Some examples (this assumes the basic sandbox premise, pioneer station):

At the start of the game your character basically has two options.

1. They have their inheritance, quite a tidy sum of cash. The station is a living economy, with a population of individually modeled agents all trying to survive, and all working in businesses and station government roles to maintain and expand that station economy. Naturally, you can take that chunk of cash and buy a business (or share of a business).

Depending on your stake, you can either be top dog, or at the least a major executive. You could of course just sit on the owner's board and collect your dividends too. However, the station is a busy place, and often you will be called upon to handle some management functions (and of course that lets you have more of an impact than passively collecting your profits).

2. You have money, and though new to the station there are always politicians in need of campaign funds. By buddying up with a local council member you can start making the connections that may one day see you in an election of your own. Besides your cushy government compensation you can always seek less scrupulous ways to pad your bank account (laws for cash isn't unheard of). Or maybe you just want the adoration of the masses and continued growth of your power on the station.

Either way, membership in the council, the faction parties, or local special interests (businesses or otherwise) are all ways you can climb the ranks of the 'political' ladder. Whether you prefer to be loved or feared, from these positions you can have a direct impact on the stations development (and perhaps skim some of that pork for yourself).



So you can run a business or you can schmooze people, or some combination, what are the details?

Well the bulk of the gameplay as mentioned before is conversations, your character walks about the station and you are involved in meetings, you meet up with movers and shakers one on one, so on and so forth... but basically your day is to convince people or make the decisions that will have some effect on the characters, economy, or policies of the station.

For instance, running a business you can hire/fire employees, conducting interviews, set production strategies and goals, negotiate contracts with suppliers/customers, beg for bailouts, promote synergy... essentially your workers spit out all the tinker toys or food cubes, your job is to come up with the plan that those workers are following.

Delegation is a big deal, and understanding others is key. These are not worker drones, everyone from your CEO to your janitor has a little mind (it may be simple, particularly in the case of CEOs, but it is their's). You don't need to tell your 10 employees to each build 20 goods... you set a goal for 200 units of product, they will self-organize to hit that goal. If you set an unreasonable goal, they will fail, they may even warn you of that fact in advance... but the game will let you make a bad plan, it is just out of your control whether it gets executed.

The best way to think about the gameplay is not 'I'm building 4 factories and shipping yada per day, making 12000 credits an hour. and power gaming. Instead think of it as pruning trees. Create a plan and your little simlings will follow that plan, and if it is general enough, keep intelligently expanding it as makes sense. It can get to the point where you just say 'keep making more money' and the little manager will go off and set up the construction of the 4 factories, planning of the production schedules and sales all for you... while you are off working on another scheme.

This is a game where the more you micromanage, the slower you probably end up succeeding. The real challenge is hopefully going to be in your ability to react to circumstances, plan, and convince others. This of course is even more obvious in the political side of things where you can't even really issue an order, only pass laws and proposals which you can design, but do not actually implement yourself.

[SportsDino]

Okay, i wanted to put out more updates but I got sidetracked on a few things and it took me a while to get back to game programming.

Anyway here is the state of the project:

1. Engine spawns a set of agents which are nothing but variable containers at the moment with little pea-brains. The expectation is to scale up to 10,000 simultaneous AI citizens at a time. Each one has their own little life which is pretty detailed, but the processing is scaling friendly (as level of detail works for graphics, I run a similar concept with my little agents so they seem really smart when you are interacting but for the most part are simplistic with bursts of genius when they are in the background).

2. Created a 'conversation database' which contains hundreds of mini-dialogue templates. Most of these are functional at the moment... although I plan to add at least a hundred drama/comedy/tragedy templates to the mix before first beta. Yes, the conversations inevitably will seem cookie cutter after enough repetitions (its basically form letters on steroids).

3. The conversation AI (which will be plugged into all those pea-brains eventually) is testing out well. Agents have a collection of personality traits and the variability has been pretty good. It is not exactly scripted so players might be a little surprised with the NPC conversation of this game (that is the point, maybe I can sell the engine or at least inspire game studios to try harder)... the agents have their own agendas and they are coming across very well. This game does not depend on 'skill checks' to succeed at convincing, you need to be able to read the agent and play on their personality or facts to get them to do what you want.

4. Developed a set of goods and an initial tech tree for the game. This is not Civ, Master of Orion, or Rise of Nations... the techs are mostly linked to the ability to create new goods (or create existing goods in new ways), and they are not +10% to asskicking type techs either... the tech does a particular thing. For instance, you might invent the personal hovercraft, and before you know it you have a hovercraft industry employing the citizens, and you see little hovercrafts buzzing around the space station. You might invent a cure to the space plague, which does nothing until an outbreak breaks out. Etc.

5. Currently working on the representation of the space station itself. This is somewhat impeded by trying to settle on the graphic design and interface (full 3d or top down layout or top down with layers). Ideally I want a full 3d station, but I don't know if I have the resources to pull it off, yet alone the art skill to make it presentable. I think I can get more forgiveness at bird's eye view graphics!

6. Conversation GUI is in place, think Civ diplomacy screen completely turned on its head and remodeled. There is a conversation bin where you can pull out templates (or pull up saved or recently used lines). Then there is a point and click interface for configuring it into a statement. While you are talking the game is paused since it is a little slow to put together a convo (but I use all the extra cycles in the background to process more AI, but thats getting technical). I think these controls are reasonably fast to use, and I've put in search and customizable hotkeys (all of which are set only while in convo mode so you can have a different set while on the map interface). Takes some getting used to I admit, but if you ever wanted to control what your player character said to an NPC (such as say 'Fuck You' to the evil overlord in the middle of his evil speech).... well this system can do it! Only thing that remains is I'm considering creating a shorthand system so you can type (with intelligent auto complete) and compose statements quickly only with the keyboard.

7. The player can access the full conversation database (with the exception of plot elements the character is not aware of). In story mode as your character uncovers more details of the game world the conversations you can hold with characters expands. This is mostly not the focus of this release, however I am hoping I can pull this convo-system into future games like conventional RPGs where mysteries and character specific dialog will be more common and critical. In this game it is mostly functional with a few key points in story mode.

8. This may be changed by the graphical layout, but I have a tentative skeleton of laying out the station map. This is actually pretty important, the station is its own little bustling location that is constantly changing as the game progresses. The player has a physical avatar that needs to navigate around the station to perform some tasks (although its the future so of course they have great telecomm and many game options can be performed that way)... but the navigation system is mostly camera control and point/click. You find the medbay for instance and click on it to say 'go here' and your little dude goes there in the best manner possible. Sort of like the Sims 3 really, except hopefully smarter (because I dumb down the nav-paths). By the first release i need to get this thing humming along smoothly, because I'm trying to put as much visual feedback into the map as possible to avoid having to bring up details panels on every little thing. One thing I am toying with is popping up more information overlays as the camera zooms in or out on locations in the station. Some success, some failure in this regard (mostly because I won't settle on 3D versus forced perspective and I'm hacking too much).

9. People Graphics. Well they suck, I'm terrible at making people, and in a game mostly about talking to people that is pretty bad. At the moment I'm settling on stationary photos of my mostly stick figure like people. I'd love to procedurally generate face pictures so I can make it seem like each person is semi-unique without actually doing art (sort of like the 'create a character' you see in sports games these days).

10. Building Graphics. Somewhat better. Whether it is 3D or top down, all of the location models are built on variances of the futuristic sleek down to grungy rust metal sci-fi theme. Not studio quality by far, but I don't think the station will look too-embarassing. I'm seriously considering the forced perspective simply so I can make all my model components only need to worry about being assembled in 2d space. This does make a good simplification of my art design in this respect (not to mention dead simple nav-paths!).

11. Object Graphics. Again, back to suck. I love architecture, but creatures and things I am terrible at modeling in 3d, and I can't draw 2d. I've considered making the first beta graphics-lite and seeing if I can drum up support through hopefully amusing gameplay to get either an art community interested in it, or enough interest I can justify hiring artists to redo all the people and objects in the game (and probably some of the building components too). for now, sticking to my strengths, and letting people rely on my extensive text labels to tell the apple blobthing from the laser pistol blobthing!

12. Map editing. The functional pieces of this are coming together for 2D space stations. I'm thinking of killing the full 3D construction, so this might be ready to enter the next phase soon. Constructing the station is a semi-complicated process... you need to configure some superstructure lego-blocks first (for instance a new wing of the space station, or a giant corridor, or where the solar panels hook on). Then for most of the economic units (say pistol factories, or farms, schools, etc...) you need to assign them with variable size tetris blocks that fit within the superstructure. I'm not liking how little variability there is... but for a first game I think I'll just live with it. I think some particular parts of the game will work beautifully (for instance a sort of Deep Space 9 like promenade section creates pretty neato looking areas that look sort of like a cool shopping mall, whereas the corridor section always looks very blah). I think the fun of the map editing process will be one of those things limited by time, the closest parallel will be Sim-City-ish. You can place a building, like a school, but you have no real control over customizing it. According to procedures the school will have a certain graphical style (that can change based on tech or some parameters, but only to the extent of what I created art assets for)... but you can't reshape its size or positioning in any way. The one cool thing I am doing, you can create 'zones' and the AI itself will populate that zone with businesses/services based on the economic model... and change it as needed over time. In fact, that is how most of the construction occurs, the user doesn't really layout buildings for the most part.

13. Economics engine. This is basically unchanged from a generic sim engine I have sitting around. I've yet to adapt it for this game in detail although I did make sure it interfaced with the new agent design. The plan is for each of those 10,000 some agents to be living a full little life, they have a house, a job, and various shenanigans they get into or out of as the game plays out. I've been building these simulators for a while now, so it will be exciting to see one in game form for the first time out in the public. But basically more details to come as I nail down parts.


------

so from here, what next.

Well, I'd like to start an article on the plot design for the story. I've got an abstract, but no real outline yet. I also have to gauge spoilers versus getting feedback (not that I'm expecting all that much feedback, so I might just go full spoiler on the general theme of the plot, but not go into any of the details that will make for juicy reading in game).

Story mode is not the focus of teh game, this is more a pure sandbox sim, but I think it will help guide some of the other design aspects to make for a better sim than just thinking about it as an economic modeler would do. A good challenge helps foster creativity, my opinion anyway.

I also think a run-through of an average 'day' of gameplay would be cool at one point to get a perspective on whether it would be fun or dull.

------

[JetsIn06]

I would buy this game. Keep it coming! Sounds extremely cool and interesting, and it's great to hear about the process.

[SportsDino]

Settled on a graphic design for the game:

-----

Map View:

Top-Down, essentially uses color coded squares with labels (think shopping mall style store directory).

Motion is pan in 2D with zoom in/out (but always with the 'camera' pointed at the same top down orientation, with no rotation). Level of detail is used to increase the amount of information popups based on which zoom tier you are on. To simplify, zoom in/out is done at steps, basically you have way zoomed out (basically fit space station in view at all costs), high zoom (able to see a lot of the station but very little labels/data), and a couple more intermediate levels down to lowest zoom (information rich, but really looking at an area not too much larger than the other view I will describe soon).

The game is built with a scroll wheel mouse in mind, clicking middle button will advance to the next level of zoom where you pointed to. Double clicking will go to lowest zoom (or high zoom if at lowest). Scroll wheel will go up or down a level. So navigating the map view should be fairly rapid and intuitive, just aim and zoom.

Right click allows you to access location sensitive menus, which can do many of the games tasks. Left click I'm not quite sure yet... it might do multiple things based on various context.

Map view has no rendered graphics, its all purely functional and schematic like. A large amount of gameplay will take place in this view (simply because its the most efficient).

-----

Man on the Street View:

This is where you see your player avatar, as well as other characters in the game. The graphic inspiration is isometric based on the original Fallout 1/2, however I'm adding rotation.

Say you are in a corridor, it has shops on both sides. You will see only one side of the corridor at a time, which will be rendered 3d building fronts. The other side is not in view, but will be represented as a foreground shadow element with informational labels. When you 'rotate' the view, the side you were looking at before is moved to the foreground, and the shadow side becomes visible.

The layout of the view is such that your character and the space he can walk in is centered, so that the shadowed side does not obscure your view of the full navigable space at any point. The buildings that are in view are set up against the wall at all times, so they define one edge of the space you can run in. The shadows define the other side. All area is in view, so basically you see through buildings to get the lowest edge of the space in view at all times, the shadows are basically labels of what is there... not really a graphical element. This is to avoid the issue in Fallout where you would run up to the wall towards the player and need transparent bubble to see what is down there (so rats for instance could sneak up on you completely out of view if you were unlucky).

When you reach an intersection you get a camera specific to that intersection (T-junction or 4-way). This lets you see all buildings accessible from that intersection.

This graphic scheme requires renders of the following permutations:
- Against the wall (i.e. one side of the corridor)
- Facing towards player straight on (T-junction intersection)
- North-east, South-east, North-west, South-west corner views (for 4-way intersection, and the south-east and south-west corner of a T-junction)


So why this view?

1. I can do this without using a 3D engine. It is possible to assemble this as pre-rendered sprites with a tile system like Fallout 1, and all of the characters as simple animated sprites moving between tiles. Granted 3D power is pretty cheap these days, but there is some technical investment I would need to do there compared to a more restrainted 2D approach.

2. I'm nostalgic for that period of graphics in general.

3. I can make the camera control very simple and other than the corridor issue, the user will rarely have to manually rotate. Even in the corridor case, the user can functionally use the shadow versions of the buildings so they don't have to rotate in order to play the game.

4. I can make the buildings as high detail as I want if they are prerendered sprites without worrying about polycounts and such. So the initial graphical quality will probably be slightly higher.

5. I don't need to do complicated 3d algorithms to chop up the model based on field of view, I can force the perspective to only show a little bit of the station as needed with very simple logic.

6. Tile based navigation and animations will cut down the amount of work I need to do.

7. I can make the people somewhat vague-ish looking similar to Fallout... hopefully enough so to cover up my sucktitude.

8. I can dress up the set, for instance with furniture, based on tiles, and maybe some set animations for using them. Don't need to worry about collision testing as much.

9. I've decided to avoid exterior shots of the station, basically to cut down the amount of work needed to handle that case. Wasn't adding much to the gameplay, although it would be nice to have if time opens up.

What is this view used for?

1. Only view where people are visible.
2. The view you get when inside a building.
3. Most of the graphical splendor takes place here.
4. Gives that personal avatar connection feeling
5. Conversations take place here, the convo-interface overlays the view in a manner where you can access all the controls but still see everyone you are talking to. This handles the one-to-one and group conversation case. The output from the agents will be visible over their respective characters (with speech bubbles) and also in a summary pane (for convenience and history tracking).
6. I don't do any Sims-style maintenance (i.e. you don't make your sim go to the bathroom), but I plan to have a lot of objects you can interact with around the station for giggles or alternative sources of doing things. For instance you can walk up to the hovertram and zip around the station, or go up to a news kiosk and get the paper, etc.
7. All travel can be performed from this view, although the map view will basically give Oblivion style fast travel (time will pass, roughly a bit longer than what it would take you to walk it).


Expectations:

This view will probably be used when the player is first starting out and exploring the game, with declining use to particular situations when the user has learned they can do a number of things faster from map view. However, this will make the game feel more like an RPG in a few key ways, which will hopefully keep it from becoming a boring numbers fest, and also make it easier to give it a heart when it comes to story mode.

Also as a player myself I like to SEE the massive empire I've constructed. I can see schematics all day of how many businesses I own and all that jazz, but seeing the little buildings themselves and the people bustling around them, although complete fluff for the most part, to me has always been a pleasurable part of any sim game.

Another side benefit of this design is it is yet another piece that can be salvaged for future for-profit games. People might get sick of the forced perspective and sort of cookie cutter designs after a while... but it does make location design and production pretty efficient and fast (decreases the variability of the layouts I can build, but streamlines a lot of the steps so I can assembly line more raw art into the games, a tradeoff, but one that makes sense for a single dev or even a small part-time team). I'm thinking if the game succeeds in getting an audience a more RPG-focused follow-up in the same universe would make an excellent if non-traditional sequel (completely switch gameplay genres, but keep a lot of the art and engine components, just switch them from economic-political sim towards some more traditional RPG gameplay design).

I'm always about the reuse of stuff.

------

Anyhoo, next steps is building a bunch of widgets for Map View, and then getting the simulator to start building and running stations. I'm switching to a full 2d art design so I'll need to hack together some supporting code for a bunch of stuff, but that gets into boring details that will take forever to do, but aren't really worth the time to discuss here.

[path12]

The way you describe the individual people remind me of citizens in Tropico. Am I understanding that correctly?

I remain quite interested in seeing how this turns out.

[SportsDino]

Tropico is probably the closest parallel yes.

I guess I could break down what an 'agent' is since that is something I've got approaching a near complete state in terms of basic technology:

1. There is one agent for each simulated citizen of the station.

2. Each has a home, most likely a job, similar to Tropico (though the particular mechanisms in my opinion are more refined).

3. Each has needs such as hunger, shelter, luxury, but its not like the Sims where the agent is constantly trying to essentially manage entropy of a bundle of need bars. All needs are averaged out long term and treated in bulk with probability of the related actions to serve those needs being carried out.

4. What the above means is, with some probability watching a single agent it can live unrealistically, for instance you may see it never eats for a whole month in a row, then goes out to eat every night of the next week... my system doesn't correct for such imbalances (other than making them improbable). Unless I command it to... but that imposes a constraint that makes certain aspects of the AI take up more computational complexity for a non-necessity, and no one really cares that each agent does the eat poop sleep cycle perfectly anyway. I'm not simulating that!!!

5. Agents have an inventory, and their house has inventory, and every building in the game has inventory (sometimes broken down by furniture having inventory). However, other than tools most inventory is handled in bulk terms. An agent doesn't buy an apple, it buys a months supply of apples, and so on. The way agents process inventory is actually one of the useful innovations of my design... I can force it to think on a more detailed basis, but it terms out doing planning in bulk if often more efficient. Indeed, the best way to make it do detailed inventory management is merely run the bulk model and then set it to work in discrete plans in the short term (as in it has a plan for the month, but within that larger plan it rapidly develops short term loops to make a realistic simulation of each day).

6. So like Tropico, each agent is running around the map doing various activities related to acquiring services for themself, doing their job, or seeking entertainment. The wealth model is more detailed, instead of a generic 10 per hour with some fraction representing your maximum housing rent you can pay, the game engine can handle such complex things as an agent having a mortgage and use of debt/savings to manage short term and long term supplies.

7. Agents in this game are not broken down into factions. They have a set of personality traits, a set of 'knowledge' (facts they have picked up usually from work/news), a relationship network, and statements they have decided reflect themself over time (and that can change). Since politics is a major aspect of the gameplay, most agents affiliate with some form of party, or multiple parties, or an independent with certain leanings. These party affiliations are handled as decision by the agent, and can change as conditions in the sims life changes.

8. All agents are operating on a long term plan of their own devising, this reacts to changing conditions as needed. An agent doesn't just exist to cycle money in the engine, they are trying to work toward some goal. Some of those goals may be simple, mere survival or couch potatoe. Others might be trying to strike it rich, or become famous, etc...

9. In the short term, time is blocked out in the agent's schedule according to probability to service needs/luxuries and obligations. If imbalances exist the probability for servicing that issue is increased, but there is no penalty to the sim for ignoring any particular thing for a long time. For instance, maybe the long term plan includes going to the doctor every six months, but for whatever reason that slot never lands into the schedule for two whole years! The probability will increase steadily that time will be devoted to that slot, but the sim will not start having poor health or any negative effects.

10. There is an exception to the above, performing certain tasks does certain checks on various statuses that may result. For instance, catching the dormant space plague. If you had went to the hospital, it might have been detected before an outbreak, and in that case, yes the sim would have better health if they stuck to the long term plan.

11. So why do 9 + 10, aren't they contradictory? Well no. The physical event of going to the doctor needs to occur, as this causes the doctor's revenue to get incremented, the space plague cures to be administered, and all sorts of necessary transactions for the accuracy of the economic model. However, using a probabilistic system over a hardcoded script allows for very adaptive agent planning and a more realistic looking simulation, en masse. It also can be coded to enforce detailed schedules within some tolerance (say six months + three months leeway before it forces a doctor visit into the agent's schedule at all costs, probability be damned)... at the expense of slightly increasing the resources to process schedules.

12. On average, I've found the probabilistic system to rarely have these anomalies anyway assuming the amount of schedule space is relatively large. The normal result is pretty similar to real life, constant things with obvious short term gain occur most of the time, with the random special thing happening every once in a while co-opting the basic routine. The strength is that I can add a lot of routines to such an AI without really carrying about the AI design all that much... the equations tend to sort everything out.

13. Each agent can talk to the player. And when you talk with them will give a detailed run down of their life, in general... although if you ask them specifically about what they did on January the 4th 2136 they won't know. They basically have no short term memory beyond a day or two, and that bit of memory is just due to some caching I do for efficiency/safety's sake.

14. An agent has very detailed processing when they are creating a plan or talking to the player. So I simulate very detailed world, by really only doing the work when it is needed. So each NPC can go deep (although some NPCs are dull and others more lively based on traits)... so unlike Tropico, you can find Joe Schmoe (say you are mayor) and ask him specifically what gets his goat and what it would take to get his vote. He will tell you, it often will be self-serving, ideological, or in direct conflict to the platform that may be best for the station or to get you elected... but its the player's job to consider what an agent says versus what will really be the deciding factor. Generally they ask for the sun if you would let them (who wouldn't) but when push comes to shove they have to vote for someone (or not at all) and really you only need to be better than the opposition to win. But I try to provide plenty of ammo to shoot yourself in the foot, I thought it would be a good idea for a game to not have it be so straightforward and easy.

15. The entire economy is based on real transactions. Most of the planning is done based on abstract plans (for flexibility/speed) but if the player wants they can get a detailed report on how many twinkies were sold in the last year, or how many people died of space plague.

16. Agents have a lifespan and family connections (as well as friends/enemies). They also have, as I said, a collection of statements that define them. This is basically how I store historical information, such as "I hate the Joneses because they stole my newspaper business!". The introduction of this kind of memory (personal through self-statements, general through news/facts) will be a key distinction from Tropico or the Sims where relations are really just a like/dislike slider.


Anyway, I guess I've gone way into the category of too much information. Obviously agent design is something that excites the hell out of me, particularly in tricks to create deep intelligence (or seeming deep intelligence) that scales to large numbers efficiently. Really, once you see it I would think the general statement you would make it out as is:

"Tropico-like but more detailed."

Like Tropico however, the player isn't expected to obsess about a particular agent, but merely the summary opinion of the population as a whole (and I've been workin on lots of graphical widgets to aid that process).

[path12]

Thanks for the detailed response. I think that generally individual needs is a pretty sticky problem in programming a game system, and think you've taken some interesting choices.

One question that strikes me -- I don't find points 9 and 10 contradictory as much as points 4 and 10. But even then it looks as if an agent could be satisfying needs as it 'slots' but possibly an unforseen event either prompts a certain action in a certain slot?

Also, while I think it necessary for an interesting game for the player to have imperfect information, I didn't notice if there are specific ways that a player will be able to at least improve his chances of making the right call in determining what a particular agent (or group of agents) is really concerned about. Without that it seems you could end up just making random guesses -- which is not fun in the long run.

[SportsDino]

Good questions.

Individual agents in games is actually a very messy problem, even the simplest robot vacuum cleaner game can have incredibly huge computational complexity with a simple world and set of rules. I think players underestimate the difficulty of AI programming in general (although I'm right in there with them bashing every game's AI on some aspect, including my own!!!!).

Most games simplify the processing per individual to keep things feasible. Also the scripts they run for the agent are fairly predictable after playing long enough, although they may be pretty efficient (for instance Civ 4 AI is pretty good at expanding quickly and if you don't adapt your play to take that into account you can have a rough time as your tiny nation is surrounded by massive AIs).

My method is not naturally better, I don't have that much ego until its proven in action, but I've made some very unconventional decisions in order to try and break the glass ceiling of AI capability as it were. I'm hoping to succeed, but I honestly can't claim victory. So far I'd say I've managed to build one hell of an underlying engine for a game like Sim City, or Roller Coaster Tycoon. But those games derive their fun from the sandbox environment and construction, the AI isn't the star of the show. If I can apply these theories to a game like Civ (or this game I'm building), that in my mind is proof that these ideas are awesome and have great potential in the future.

As a consequence though, I've made interesting choices as it were. Some of these are reductions based on the structure of my design, and are fundamental aspects of the technology. Others are simplifying assumptions that can be improved with additional modules on my engine (I've described a few points in my post on agents where this is possible, its a matter of investment in additional complexity, several places I have chosen to do so already and others I have not). Whether it works or not, I truly do think that players will notice it is distinct from anything they've played, hard to say though whether it will be for the better!

------

4 and 10 are somewhat contradictory, how best to describe it...

An agent's standard of living is tracked abstractly, for instance hunger.

When an agent eats, for instance at a fast food joint, it does not increment any status indicator. There isn't a bar in the agent called 'calories' for instance, such that you eat a hamburger it goes from 0 to 500 (Sims style).

Rather it can track the last time the agent has ate something... and nothing prior to that.

So then it models the next transaction for eating fast food as:
- Desire for food
- Desire for X type of food
- Desire for X type of food served fast
- Proximity to fast food joint
- Time since agent last ate

All of this is rolled into a probability factor, and all of those probabilities compete against each other for the potential that a particular block on the agent's schedule will be to eat at the fast food joint.


The consequences of such a design:
1. It is possible that the agent could eat at a fast food joint three times in a row, and then never do so again for a month. Anything is possible, merely improbable dependant on the model.

2. The agent's 'routine' is essentially an equilibrium around the various probability factors in play. While it can vary, without disruption to this equilibrium it will converge to some pattern behavior (such as almost always eating at the fast food place next to where the agent works, usually visiting the same hangouts more than others, etc...)

3. The only way to impose programmer defined patterns on behavior is to add more conditions to the model. For instance, if the agent must eat everyday, we can guarantee it does so with 100% probability. However, I hate to do this, because such constraints almost always end up dominating behavior. At best it looks like I just ran a bunch of equations to create a purely scripted AI (i.e. boring), at worst it goes maladaptive with the agent doing highly suboptimal behaviors that look funky, decrease economic efficiency, or give easily exploitable loopholes to the player.

4. In practice, this tends to lead towards behavior that is quasi-realistic. I could apply this algorithm to my daily life and it would fit, I tend to eat fast food at locations near my work, I have certain things I enjoy and tend to waste my time on them more often than not (like game programming and message boards, lol), and I break out of my routine rarely but usually for some specific reason, and usually I try to fit in a few other unique errands along the way. For instance, if I go to the mall I stop at multiple places, even if I'm only seeking one thing I feel I really need. In general the system is pretty natural in its results, even though everything about how its built seems almost unnatural!


So to relate this back to standard of living, the model merely assumes the character is eating on time according to whatever plan it is running. If the agent avoids eating for a month, its still perfectly healthy! However, if I choose to how I would combat this is by adding what I call a status (or statement about the agent).

For instance, I could put a trigger that if the agent ever goes three days without eating they get a 'starving' status. If a status is in effect it skews probabilities and abilities of the agent as long as it persists. A reasonable design would be to simply set the probability of all luxury interests and non-survival motives to 0, and bump food. The only thing in the agent's queue then would be actions to keep it alive and eating, increasing the probability of food in the near future. In general this is how I prefer to handle forcing, rather than make something certain to happen, disinhibiting the agent from doing most things gets a pretty immediate response without making it act stupid in the short term (like drop everything and go to the nearest food vendor, such as skipping work).

Such status are rare, not because they are computationally impossible, but simply because for this particular game I don't think they are necessary. Generally the probability model does the trick, even fairly subtle models (I don't need to make the hunger/time curve exponential for instance, a nice clean linear relationship gets good bell curve like behavior without jumpiness in the agent's priorities).


The excellent thing about probabilities over scripts is that they stack tremendously well. Once you free up the constraint of hand-managing the thought process of the bot, adding new and even complex behaviors is just so incredibly easy, as long as you wrap your mind around the design. The most nearby parallel I can think of is subsumption architecture.

My method has particular quirks of course, understanding why would be a fun exercise in computer science but I'm already wasting a lot of bytes as it is!


So to answer the second part of that question, yes the agents can change slots on the fly. It actually is pretty dang fun to boot! Pull out a laser pistol and shoot someone in the middle of the space mall, and watch as everyone disrupts their perfectly precaculated lives and scrambles for cover. Reactivity is actually a strength of this system. Predictability is a weakness, to handle reactions I sometimes need to thrash slots that better fit the routine in order to make realistic behavior in the short term. I rely on probability to converge back to routine. However even in the short term there is unpredictability, each person will react to the murderer in their own way, on average probably the same (scream and run/hide), but others may try to call the cops, others might pull their own gun, etc.

I've got something I need to attend to right now, but I can discuss imperfect information some other time. I agree, it is very dangerous for the fun factor if it doesn't work. It actually is one of my concerns, mostly I've built hardcore econ simulations, but whether they make fun games will be a bit of a mystery!

[SportsDino]

When it comes to information it is a political/economics game so some of the common tools of those trades will be available:
- polls
- historical data (charts)
- models to estimate results of an option (not always accurate of course)
- expected political support (based on standard platform of a faction)

And so on. Basically as many as I can find a good use for.

Also the player won't have to seek out all information themself. There is a news system in the game which is actually a gameplay element, interstation news and local news is very useful for gathering summary style information, although I do allow it to be similarly biased by the reporters views.

For certain objectives, playing only with common knowledge will be limiting your character. For instance, you can pass a good deal of laws and such just reading the paper and following the polls/models, but if you have something controversial you will probably never be able to build a coalition just from summary information. To get individual votes you may need to talk to that particular character and probe what will make them fall into line, and that is where it can get dicey.

I'm determined that the depth doesn't become mere fluff either, there is more than just color commentary coming from the NPCs. If you never go into detailed information you'll basically be avoiding a lot of opportunities to use persuasion to increase your edge on the game, although I think it will be possible to play without doing it. I want the game to deliver a tangible benefit for strategically making the effort, but it will be a matter of balance and making things fun.

One on one conversations may be a necessary part of the story mode to get particular end states, the story mode is the hardest in the game though. Sandbox the detailed information situation will be a lot less obvious pressure, but include incentives for pursuing it.

The common way to think about it is, if you don't know anything special and you want to buy a trainload of toilet paper rolls... well you would probably make the contract for whatever the market price for toilet paper happens to be at the time. If you are digging for info, you might find out a contact somewhere along the distribution chain that can get you the shipment near wholesale, saving lots of money. But there is no guarantee such information exists, and if you are too gullible there is reason for an AI to try an fabricate information to get themselves a benefit. So digging for details can help or hurt you, based on your skill at reading an agent (and common/business sense).

[Neon_Chaos]

Sounds very interesting.

Have you ever played the game StarTopia?

[SportsDino]

Have not played StarTopia, I'll give it a wiki though.

[SportsDino]

Hmm, it looks interesting, I'd liken it a bit closer to the Tycoon style games then my efforts, but it actually looks interesting enough I'll probably track it down and give it a playthrough.

[SportsDino]

I will say, I strongly considered making a more fixed station design so I could make better graphics, but I'm very confident that I wouldn't compare to StarTopia on an art basis after seeing more screens of it! My station does grow in size as the game progresses (or shrinks, mwaahahhahaa!), but I found right away that making a consistent external view that meshed with the internal gets ugly in the art department for my limited resources.

I do hope though that the graphics I do get into my game have a feel of 'bustling small town' to them, and aren't a total joke at least. Switching to a sprite based engine I should be able to populate the screen pretty intensely to give the feeling of being part of a crowd.

[SportsDino]

The station is essentially a big lego puzzle, each component has attachment points. Its really quite simple, there are only a few real pieces (the rest are just variants on this):

- Corridor: 2 walls, 2 open ends.
- End-cap: basically a dead end with one open end, one main wall with two stub walls (no buildings)
- Intersection: 4 open ends, each corner is a single 'building'
- T-junction: 3 open ends, two corner buildings, and an opposite wall.

All of there are further varied by changing widths/heights. There are adaptor pieces for converting between these differences. The 1.0 version will not have elevation as a variable, so all stations are flat and 2d, although within the components themselves I may make graphical elements with a height component (usually this means adding ghosting, which may be necessary since I might just eliminate the rotation feature and put intelligent ghosting in its place).

Within the lego pieces I hope to go pretty hog wild, some will look like a shopping mall, others very utilitarian industrial, and still others will look like tight corridors you would see in a hotel for instance. The variety in the external portions of the map should be pretty nifty.

Within buildings they are all pre-rendered and are NOT LEGO BLOCK. So basically one hotel will always look the same, excluding furniture and any coloring style effects I may apply.

Also many station components are actually multiple of these lego blocks already assembled together. So instead of making a whole bunch of corridors with tram tracks in a ring around your station, the game will provide a 'traffic ring' component with the tracks already assembled, and it can be dropped into the station and connected in one motion. The AI usually expands the station with these super components in mind (with smaller components to basically fill in the cracks and make the pieces fit).

-----

Currently working on the top down map view. I plan to post screen shots/vids once I have it working and connected to the econ engine. This will probably be on the order of a couple months depending on how much free time I find. It is not too exciting perhaps, but the info overlays and labeling is CRITICAL for gameplay and I want to do it right and move on (I don't have luxury to leave things hanging, gotta wrap em up and move on to the next component if this project is to get done this century!).

When that project is done I can show a time lapsed simulation of a station from 'birth' to maturity, which should make the whole lego concept make perfect sense.

-----

Expecting to move on to the graphical person view immediately after. That should involve no new engine work at that point (for buildings anyway), but will be highly graphics programming intensive. Not to mention I'll need to work on my sprite animation and 3D art skills.

[SportsDino]

Switching from a 'fixed building' design to an internal layout editor. So if you go inside a 'hotel' you can build as you want BUT you are limited to the space specs of the building, and can only assemble from a collection of room components.

Some buildings will be pretty large, and I'll use this flexibility to assemble some of the more interesting looking locations.

The AI will tend to use basic super-components of the interiors, with a bit of other components to fine tune. The story version will probably ship with pre-designed locations (which are added to the station procedurally though) to allow me to create some consistent hand designed maps for the story mode.

Map design is not really a key part of the gameplay... however I thought I could use the interior builder to give a little more artistic flair to the game that the extenal corridors design somewhat lacks. Since I only have to worry about one render angle and no ghosting (interiors are all forced perspective) I can create a lot more components with simpler rules for how they are assembled.

Building interior will also be multilevel, so you can have grand staircases with balconies and what not. The only ghosting is based on elevation or certain obscuring 'furniture' elements like columns.

Inside a building there will be a mini-map similar to the top down map view for station nav, but you can not view both maps at the same time. I'm putting together the mini-map code with the map view code, so I started thinking up how I want these buildings to fit together.

I'm doubtful whether I will allow the user to directly layout buildings or not, I'm not particularly interested in making an editor in the detail enough to make it usable. Will probably allow the user to specify rooms and corridors, but like station externals they are all drag and drop pieces that lock together. Actually furniture and artistic elements will probably be under AI control with perhaps some color/style params the user can specify.

[SportsDino]

Thread is nearly two years old which is bad by even my standards of laziness.

The game has been scrapped for parts and overhauled to the point it is pretty much just best to start off from scratch.

So the new abstract:

- Still a political sim, from as low as a 'community leader' (a person of influence within a station, but only neighborhood sway and no real power) to president of an interstellar empire. De-emphasis on the communication gameplay style (will make a hell of an RPG out of it someday, but it doesn't really drive a sim game as well as I thought).

- Perspective shifts closer to the traditional omnipotent tycoon being, there are no face to face or avatar interactions anymore.

- Along with that the graphics step out quite a bit to the SimCity 'buildings and vehicles' level. Fits my abilities better in terms of art and design. Also the alpha version will be top-down, in anticipation of...

- Potential Xbox release if I can figure out their technology without too many portability problems and feel all the licensing bullshit is agreeable enough. I would probably min-price and give a generous evaluation version (no play limits but reduced feature scope and download size). If I go ahead with such a release it will be in isometric or 3D engine, all of the models from the start will be full 3D so their will be no rework in terms of modeling.

- There will be single and massively multiplayer modes. I prefer both being mostly a single player game enthusiast myself.

- I want to release this year, it may be as late as next Christmas, but I'm eager to release a game out in the wild.

- The advanced negotiation interface will still be used, but is not the core of the game anymore (was feeling too RPG and getting feature creeped to hell). In multiplayer it will also be accompanied by chat.

- Multiplayer will allow multiple players within a single station and the elections are considerably more vicious. It will not be a popularity contest, you cannot vote for other players, you will be elected solely on the basis of pleasing the simmies.

[SportsDino]

Premise of the Game:
In the late 21st century the result of climate change, reckless pollution, and unbalanced resource utilization leads to a movement towards more efficient industrial and commercial processes. This simultaneously leads to technology that makes the exploration of space less expensive, and an interest in products that can be created under space conditions that are more effective than their Earth-based counterparts.

Primary industries of the space economy included advanced computer chips, fusion power generation, 'honeycomb steel', and a number of other products that are best developed in low gravity conditions. The development of the alchemic reactor, a fusion device that can generate either power or any quantity of a particular atom imaginable, was discovered after years of experimentation in orbit. While incredibly expensive in terms of material inputs, the ability to draw in any basic substance and exchange it for power and rarer atoms fundamentally changed the nature of economics.

The moon became the largest exporter of exotic materials, its lower gravity, zero population, and plentiful quantities of worthless matter made it a natural site for alchemic reactors to slowly consume the planetoid for whatever became increasingly hard to extract from Earth. Earth of course remained the key economy of mankind, producing most of what was needed. And in between were the stations which handled zero-g manufacturing, processing of alchemic materials, and the transit either way.

This game takes place at the dawn of the Lightspeed Era. The first FTL drives have finally been stabilized, unfortunately the smallest vessel that can effectively use such a drive is the size of a small city. These stations are the new pioneers, spreading men to the furthest extents of the solar system and beyond. Millions of people join the attempt to settle the stars, eager to escape the cramped modern utopia of Earth with its strictly managed billions of residents. The new opportunities for wealth and glory exist out among the stars.


Scenario Design:
The player avatar is a character that can transit between any station or planet in the game, but can only get career positions on a station. They can gain power on a single station, transfer to another station, or build a political empire across many stations. Planets in the game are macro-economies that are always managed by an AI which is much simpler than the station AI. You can influence decisions that take place on the planets, but not manage specific development (as the planets themselves are big variable bags that just produce and consume massive quantities of goods based on population).

On a station a player gets increasing levels of control over policy, and limited control over construction (you use a familiar world-building tool common in sims to design expansions to the station in a ghost-mode, with the actual construction commencing after the proposal is approved and many of the details of how the zones fill out dependent on the simulator). The station growth dynamic is similar to SimCity, you place the 'roads' and zones, but sims build the businesses and homes. One twist of the game is that these buildings have a more detailed simulation than SimCity (they are more permanant and run through lifecycles) and you can negotiate or influence the particular buildings in the game through policy or conversation system.

While the goal of the game is to do whatever you want, the obvious progression path is increasing levels of political control on a station or system of stations. The single player 'campaign' is made up of missions at various stages of the history of the Lightspeed era, from the first pioneers to an interstellar clashing of rival empires in the far future. Each mission is unrelated to the others (you are a different person in each) and can be tackled in any order, but their is an option for a linear progression mode where you can advance through the missions in order and the post-conditions of each mission will feed into a simulation that generates the pre-condition of the next mission.

This means that you may see the station you first guide out from the solar system as the center of the distant empire that attempts to crush Earth in the future, or as a decrepit wreck just orbiting the sun. Each mission is designed with specific linear progression tie-ins to spice up the later scenarios with some branching, enough that by the end scenario it could play dramatically different than skipping to it directly.

All missions cover a time period, the player can continue playing on after the time elapses (as a normal sandbox game), but the timing loosely corresponds to how the central crisis of the mission is expected to wrap up.


Simulator Details:
The game simulates an entire interstellar economy in real-time, that may consist of hundreds of stations, and millions of sims. The station the avatar is currently located on receives the most detailed simulator resources, farther stations get progressively simpler simulations that generate realistic statistics with less processing bandwidth. Sims are generally simulated on a macro level through the use of personality profiles which react to station conditions. A mess of formulae eventually generate the population that commences with the voting during elections. The same thing does not work for every population, you need to 'take the pulse' through various statistic tools and craft a policy that most motivates a particular population to vote for you.

Generally a wealthy station with high standard of living is the best way to ensure political victory. Unfortunately you need to prove that it is you that led to such a situation and not some other guy. Also populations have a variety of issues they care about to different degrees, and unlike Americans they actually pay more attention to your actions than your speeches. The future version of news networks actually know how to breakdown policy in a way the masses can follow, so don't be surprised if the proposal to euthanize everyone at 60 leads to a landslide defeat at the polls. Or maybe in some communities that may be just the thing that gets you the job (it is possible in the engine although unlikely).

The economic sim is highly detailed, every unit of supply and demand is tracked to the smallest degree. If you fail to build farms or import food people will starve... the flow of goods is obsessively tracked across the entire simulation (assisted by the fact that stations themselves are the main cargo transfer device, although smaller orbital ships exist for connecting to nearby planets and stations). The economy of Earth in the game is like a massive blackhole, incredibly high supply and demand, the majority of which is domestically consumed on the planet itself. The game deliberately makes this macro-economy of a realistic multi-billion person scale set against the thousands or millions of your relatively tiny station. I really like my trade simulation, although it may be overly realistic. However, it is easy to figure out business in this game and there are plenty of helpful tools to guide you to solvency, only the power gamers might be overwhelmed trying to figure out how the greater economy works.

Although farther stations are not simulated in detail, particularly from an AI standpoint, they do have a completely accurate inventory of buildings and population and generate supply/demand at full precision. The inaccuracy of the simulation is that their decisions are made with much less complicated AI and they are only modeled at much rarer intervals rather than the constant up to the second simulation of the avatar's current station. What this means is that wherever you are at in the game world you see all the details of how buildings are located, little space trains scooting around the station, cargo haulers emptying bays to carry off to the space elevator to Earth, and so on. You can look in the news and hear about that station out around Pluto and even get a detailed manifest of what they want in their next shipment (communications are instantaneous) but it actually only makes upgrades to its buildings and population variables maybe once every ten minutes or so (although even here I cheat a bit to make it look more real, all of the statistics are equation based so all that changes at AI intervals are variables feeding into those equations, but the actual number of people is calculated so if you open up the Pluto station window and just watch the population number you will see it go up and down, other stations arriving at it and changing the cargo/population numbers, and so on... it all just happens because you are watching, if you do not watch then the simulator waits until an AI interval and then uses probabilistic methods to immediately calculate similar results as if it was being calculated every second.)

You cannot really see inside the station, you might see internal vehicles based on your station design, such as buses/trains/flying cars, but they are designed to 'look cool' but do not have simulation accuracy. In particular the time scale is highly accelerated most of the time but everything internal still moves at the same rate.

External graphics are simulated in real time (if you hit accelerated time you see the little ships and stations move faster). For instance, say you are approaching Earth orbit, you will see perhaps a dozen cargo ships launch out of your hangar and zip off towards the big blue planet (until they clip out), and eventually you will see them come back (or more likely others take their place, depends on your businesses). External traffic is designed to be deliberately chaotic, particularly around busy places like Earth orbit, the only limit is whatever the graphics can realistically handle.

There is combat in the game, but it is highly uncommon. Each piece of station superstructure has its own hit points, and station damage can lead to fatalities in your population and cargo loss (as well as productivity damage). I am not building a war-sim so the amount of weapons and battleships in the game is somewhat limited, although it has its own tech tree and industrial simulation. The game does allow piracy and warfare, some missions in single player even emphasize it. Be warned, the people still have the power, so your station of ultimate doom you work so hard to build may just boot your ass out of power so they can go ahead and sell the weapons rather than fire them. People don't like to die needlessly it turns out.

Your personal wealth is treated similar to Tropico. You do get a constant salary and can invest in businesses (no political ties regulations so go ahead and favor your own businesses in policy... but beware a corruption aware public). You can graft more cash for yourself. Personal wealth is useful if you ever want to change stations (money can buy influence after all), it also serves as a sort of score. It can also be used to help control zoning, if you put the money in you get a say in what gets built, so you can use personal wealth to finance businesses that help your political chances if they improve economy or standard of living. Evil uses of money include hiring assassins/brutes, obscene luxuries like gold plated yacht when leaving the station, bribes, and political contributions to campaigns (including your own).

Fundraising is not emphasized in the game, as the media will put your life history out there for free, but purchasing various advertisements in the game will let you put your own message out there for a price. It will sway some personalities more than others, also it is mostly useful for advancing into new territory, people tend to get vindictive over your past performance, and no amount of advertising will change their mind that you are an asshole after they see how you run your office. Only your most fervent supporters will contribute to your campaign, although you can aggressively fundraise if you wish. However, people tend to get pissed off if they are constantly asked for money, particularly if they don't have any, and some fraction of the populace will vote against the candidate that irritated them the most with requests for cash or annoying advertisements. Targeting is important, and the game is littered with tools for figuring out how to do that.

Technology and Goods:
There are numerous tech trees, mostly for commercial goods of varying qualities or production techniques. The game has a large number of goods, everything from base atoms (massive containers of everything from hydrogen to gold) to space ships. The best option is not always high priced goods, the margins on moon dirt, useful for nothing but being fed into an alchemic reactor, may be more valuable than the many inputs needed to spit out new stations (which itself is an industry in the game). So hauling millions of tons of dirt at good economies of scale might be better than selling billion dollar vehicles maybe once every year.

The sims are always trying to 'fill the void', that is make the most money with the least input and investment. They do not need to be directed to farm if food is needed, they will recognize the unfilled demand and invest in creating a supply so they get the cash from it. From an economic standpoint all the player can really do is try to position themselves in front of the void first so they get the highest return, as the sims can have a slower reaction time to recognizing an opportunity or looking ahead to the next best thing. For instance, if the tech tree is about to release wonder-widget, the toy every kid MUST HAVE, the information is available to sims but only ones closest to it may process what it really means and start gearing up production for wonder-widgets (although the AI can plan months or years in advance)... meanwhile you can know that wonder-widgets is in the pipeline and can choose to take a risk and build up the factory for them early (you can pre-build factories before tech completes, however research is probabilistic so delays could leave you paying the interest on unusable machinery that is not tooled because the research hasn't completed).

Research:
Research is an industry in the game which you can only direct through investment. There are two major types of research, exploration and refinement. Exploration is the first step of research, it starts from any vague topic that still has techs within it and progresses until it hits upon a tech within that topic. You do not know what the topic will be until research has progressed a certain point (similar to roller coaster tycoon actually).

Many exploration techs are simply topics which lead to more exploration avenues. It may also lead to concepts, something that may result in a tree of various refinement techs. A refinement tech results in a new set of goods, with a good defined as anything from raw material to entire space stations.

So your average research progression is you start on a vauge topic like 'station construction', it opens up a number of related topics, you research some of those and they begin to open up concepts such as 'Dyson Cylinders', you putz along on that concept a while and it opens up a bunch of refinement techs like 'cylindrical mega-structure', 'micro-stars', 'gravity-shaping' and so on. Finally you start on the refinement techs, each one opening up a set of goods which can be materials, components, or finished products.

All research proceeds on a probability distribution, with progress points accelerating as the idea becomes more developed. Some techs are co-researchable, they are related closely enough that progress points in one randomly distribute to the other research topics as well. This can only happens if the researchers are involved in a line of communication (with the quality of the communication effecting co-research probability). Most topics and concepts are co-researchable to a large number of other techs, while refinement techs are usually co-researchable to a small set of other refinement techs under the same concept parent and to a small set of topics/concepts that are children of the major concepts that are pre-reqs to the refinement tech. Essentially, all research can slip over into benefits for nearby research, but more general research like topics can slip into many more areas.

So the advantage of exploration research is you will find the big discoveries and open up new areas of technology and more rapidly progress overall, but you need refinement techs to actually get something you can sell for money. You can trade topics/concepts, but you cannot patent them. However, the techs stay with the research team that created them and must either be communicated to other researchers, published, or independently researched. So two groups can research 'Dyson cylinders' and both can come up with it on their own. They can publish if they feel like it (one way researchers get compensation) or they can secretly race for the refinement techs. Refinement techs can be patented, first to get there wins, and gets exclusive licensing rights. Patents are enforced by station treaties, you can break patent treaties which will prevent them automatically being communicated to you, but then you can infringe to your hearts content. Stealing patents will piss off people though, so its a risk.

All of this complexity is kept under the hood for the most part. As a player you can manage the research team, the communication policy, and select topics to focus on (more topics than you select may actually be researched). You can have zero involvement with research in the game, you have to go out of your way to get involved with research businesses or setting policy to influence research.

Technology tends to leak unless you are very restrictive with communication and keeping your researchers locked up. Also once something is discovered and published (even at low detail) others researching it will proceed at a faster rate. Once something is patented, even if it is not communicated, once it is sold someone can acquire it and reverse engineer it at a highly accelerated rate. You need to patent everything before selling it because it is assumed it would be reverse engineered shortly anyways. Advantages to public patents is it can be licensed to more producers, confidential patents can extend the length of time before some reverse engineers your product. Advantages to publishing at higher levels of detail will increase your researchers prestige, open up lines of communication to other research groups eager to work with you, and attract new researchers to your team. There also is some money in publication.