Evolving Routing Algorithms with Genetic Programming
E. Luckschandl · 2001
Chapter twenty-two reports on using genetic programming techniques to construct dynamic routing algorithms. A tool called the Java Method Evolver (JME) has been developed which automatically generates Java methods under input constraints corresponding to the variables in the network. A general communication network simulator is used to allow different network topologies to be defined. By repeatedly feeding the same call sequence into the simulator, the effectiveness of different routing algorithms under different conditions is studied. An important achievement is the architecture of the system, enabling the JME to generate Java bytecode and dynamically feed it into the running simulator.