To achieve a high degree of autonomy, an agent usually needs somekind of memory mechanism. We present a new approach to the evolution ofagents with memory, based on the use of genetically programmed networks.These are connectionist structures where each node has an associatedprogram, evolved using genetic programming. Genetically programmednetworks can easily be evolved into agents with very differentarchitectures. We present experimental results from evolving geneticallyprogrammed networks as neural networks, distributed programs and rulebased systems capable of solving problems where the use of memory by theagent is essential. Comparisons are made between the performance ofthese solutions and the performance of solutions obtained by otherevolutionary strategies used to evolve agents with memory
展开▼