Evolved controller for simulated car racing using Genetic Programming

For computational intelligence to be useful in creating game agent
AI, we need to focus on creating interesting and believable agents
rather than just learn to play the games well. To this end, we
propose a way to use multiobjective evolutionary algorithms to
automatically create populations of Non-Player Characters (NPCs),
such as opponents and collaborators, that are interestingly diverse
in behaviour space. Experiments were conducted where a number of
partially conflicting objectives are defined for racing game
competitors, and multiobjective evolution of Genetic
Programming-based controllers yield pareto fronts of interesting driving bahaviours. This work has been performed ny Alexandros Agapitos and Julian Togelius in the University of Essex, UK.

Category:

Gaming

Tags:

• Genetic Programming
• Artificial Intelligence

[virtualspecies]

Nice! More details about that particular problem would be great - what did the function/terminal set look like?

The 'impressiveness' of problems solved by GP sort of depends on the complexity of functions it had at its hands, i.e. combining very high-level instructions (avoid-wall, turn-90-degrees) is much easier than evolving a controller with low-level instructions (read sensor input, change thrust, turn wheels) :-)