Evolutionary development of path planning algorithms
- LLNL
This paper describes the use of evolutionary software techniques for developing both genetic algorithms and genetic programs. Genetic algorithms are evolved to solve a specific problem within a fixed and known environment. While genetic algorithms can evolve to become very optimized for their task, they often are very specialized and perform poorly if the environment changes. Genetic programs are evolved through simultaneous training in a variety of environments to develop a more general controller behavior that operates in unknown environments. Performance of genetic programs is less optimal than a specially bred algorithm for an individual environment, but the controller performs acceptably under a wider variety of circumstances. The example problem addressed in this paper is evolutionary development of algorithms and programs for path planning in nuclear environments, such as Chernobyl.
- Research Organization:
- Lawrence Livermore National Laboratory, Livermore, CA
- Sponsoring Organization:
- USDOE Office of Defense Programs (DP)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 3712
- Report Number(s):
- UCRL-JC-131904; DP0213000; ON: DE00003712
- Country of Publication:
- United States
- Language:
- English
Similar Records
A theoretical comparison of evolutionary algorithms and simulated annealing
Synthesis of logic circuits with evolutionary algorithms