Behavior evaluating indices for robot behavior learning and evolution

This work is concerned with the behavior evaluating indices introduced to establish a method for quantitative evaluation of robot behaviors. The indices are defined using terminology in the Intelligent Composite Motion Control, ICMC. The ICMC aims for intelligent robots that can step by step realize highly complex behaviors from fundamental motions. By applying the indices to a robot soccer problem by legged robots we present how the index values change as the behavior evolve from fundamental motions to complex behaviors. and study the cross relationship between them. The validity and usefulness of the indices are then discussed.