Minimal representation multisensor fusion using differential evolution

Abstract

Fusion of information from multiple sensors is increasingly used in planning and control of robotic systems. The minimal representation approach provides a framework for integrating information from a variety of sources, and uses an information measure as a universal yardstick for fusion. In this paper, we evaluate a differential evolution approach to the search for minimal representation solutions. Experiments in robot manipulation using both tactile and visual sensing demonstrate that this algorithm is effective in solving this difficult search problem, and comparison with a more traditional genetic algorithm shows distinct advantages in both accuracy and efficiency for the differential evolution approach.