Introducing Weighted Intermediate Recombination in On-line Collective Robotics, the (µ/µ W , 1)-On-line EEA

Amine Boumaza (1)

1 LARSEN - Lifelong Autonomy and interaction skills for Robots in a Sensing ENvironment

Weighted intermediate recombination has been proven very useful in evolution strategies. We propose here to use it in the case of on-line embodied evolutionary algorithms. With this recombination scheme, solutions at the local populations are recombined using a weighted average that favors fitter solutions to produce a new solution. We describe the newly proposed algorithm which we dubbed (µ/µW, 1)-On-line EEA, and assess it performance on two swarm robotics benchmarks while comparing the results to other existing algorithms. The experiments show that the recombination scheme is very beneficial on these problems.

Domaines: Intelligence artificielle [cs.AI]

Amine Boumaza. Introducing Weighted Intermediate Recombination in On-line Collective Robotics, the (µ/µ W , 1)-On-line EEA. Applications of Evolutionary Computation, Apr 2019, Leipzig, Germany. ⟨hal-02185694⟩