I looking to build virtual world simulator to get students and non-academics interested in space exploration and AI and wanted run down the basic concept with the people here to get there thoughts. The idea is based around the exploration of an alien planet through the use of Van Neumann machines, which are basically autonomous robots. The “player” designs robots using a component system and programs the robots AI. Effectively each component has a function, a power consumption, and size. In the initial development stage components consists of different types of sensors, construction components, and power generation. The only thing size does for now is determine the number of resources required to build a machine. The programming will be done in c# and I was planning on using an event based system for handling the “players” AI code. Effectively the robots sensors will trigger user written events as well as there being a timer based event. So if the robot had a forward facing thermal sensor with a range of 10 meters. An event would be fired when a temperature change was detected within that range. How the robot responds would depend entirely on how the player programmed it. This should be robust and simple enough to allow people to utilize whatever AI technique they choose. The player can send the following commands to the robot from the AI: Turn X degrees left Turn X degrees right Move X meters forward Move X meters backwards Activate component Deactivate component Sleep A robot is destroyed if: It falls more then 5 times it height Enter a temperature zone >50 or <-50 Runs out of power For the first development iteration I was planning on simply creating a virtual world with the robots having the responsibility of mapping it. Once that is done I was planning on adding more “game” aspects to it such interacting with objects, the environment, and other players robots. What are peoples thought on his approach to a virtual world and AI programming?

The general concept sounds good.

I see a potential flaw in your design however and that is you provide no real way to get or handle real time input. You can move the robots forward and all of that, but you can't tell where an obstacle is.

If this is for students you might want to consider making a graphical programming interface... It seems like the students might have trouble remembering your syntax. Baring that you may want to consider using a premade scripting system such as Lua to reduce development time.

Quote: Original post by shadowisadog
The general concept sounds good.

I see a potential flaw in your design however and that is you provide no real way to get or handle real time input. You can move the robots forward and all of that, but you can't tell where an obstacle is.

Your right there is no direct or real time control of a robot. Determining where an obstcale is comes down to the robots sensor components and the logic the programmer has programmed into their AI. The echo location senor allows the robot to recieve contour information of the surounding area in the form of a height map. But that data only means as much as the AI logic the programmer has written makes of it. So if the echo location sensor detects a 10+ meter drop, 10+ meters wide, two meters ahead then unless the programer logic treats that as an obstacle then robot will just walk off that cliff.

Quote:
If this is for students you might want to consider making a graphical programming interface... It seems like the students might have trouble remembering your syntax. Baring that you may want to consider using a premade scripting system such as Lua to reduce development time.

I was planning on target it at university students and above. But I could look at the various existing scripting langagues once I have the framework in place to give it a wider appeal.