Robot Shaping: An Experiment in Behavior Engineering
Robot Shaping: An Experiment in Behavior Engineering
Solving Non-Markovian Control Tasks with Neuro-Evolution
IJCAI '99 Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence
Evolution of stable ecosystems in populations of digital organisms
ICAL 2003 Proceedings of the eighth international conference on Artificial life
COOPERATIVE COEVOLUTION OF MULTI-AGENT SYSTEMS
COOPERATIVE COEVOLUTION OF MULTI-AGENT SYSTEMS
Cooperative Coevolution: An Architecture for Evolving Coadapted Subcomponents
Evolutionary Computation
Managing team-based problem solving with symbiotic bid-based genetic programming
Proceedings of the 10th annual conference on Genetic and evolutionary computation
Design of evolvable computer languages
IEEE Transactions on Evolutionary Computation
Hi-index | 0.00 |
Evolutionary problem decomposition techniques divide a complex problem into simpler subproblems, evolve individuals to produce subcomponents that solve the subproblems, and then assemble the subcomponents to produce an overall solution. Ideally, these techniques would automatically decompose the problem and dynamically assemble the subcomponents to form the solution. However, although significant progress in automated problem decomposition has been made, most techniques explicitly assemble the complete solution as part of the fitness function. In this paper, we propose a digital-evolution technique that lays the groundwork for enabling individuals within the population to dynamically decompose a problem and assemble a solution. Specifically, our approach evolves specialists that produce some subcomponents of a problem, cooperate with others to receive different subcomponents, and then assemble the subcomponents to produce an overall solution. We first establish that this technique is able to evolve specialists that cooperate. We then demonstrate that it is more effective to use a generalist strategy, wherein organisms solve the entire problem themselves, on simple problems, but that a specialist strategy is better on complex problems. Finally, we show that our technique automatically selects a generalist or specialist strategy based on the complexity of the problem.