Adding Vision to Khepera: An Autonomous Robot Footballer

This project investigates the evolution in simulation of robot controllers capable of performing a hard task, playing football, in the real world. It is argued that for any reasonably interesting task, robot controllers are too di cult to design, and that an evolutionary methodology must be employed. It is further argued that evolution in the real world is too slow for such a task, and that evolution in simulation must be used to produce good robot control structures. The techniques of minimal simulation, where the robot controller is forced to ignore certain features through making those features unreliable, are used to construct a simulated environment for a robot with vision system. A xed architecture neural network provides a sensorimotor control system for the simulated robot, and evolution is used on an encoded version of the network to produce good solutions. Two experiments are presented; nding a white stripe in a dark arena, and nding a tennis ball and pushing it into a goal. In both scenarios, good controllers capable of performing the same behaviours in simulation and in the real world, are evolved only once su cient unreliability is incorporated into the simulation. The success in evolving in simulation a robot controller incorporating distal visual environment input data and displaying the same behaviours in both simulation and the real world, goes some way to addressing the arguments that evolution in simulation is only suitable for toy problems.