Motion generation of multi-legged robot by using knowledge transfer in rough terrain

Walking motion generator of multi-legged robot is very complicated operation because there are many degrees of freedom required to be considered. Especially in computational intelligence approach, many iterative calculations are required for convergence the result. Furthermore, in case of the change in environment has to recalculate the walking motion again and then increase of the calculation cost is a big problem. In this study, we propose the extraction and reuse method of walking knowledge for multi-legged robot by using computer simulation. Sequences and patterns of motion are formed by using minimal generation gap model in genetic algorithm and open dynamics engine was applied to experiment. Finally, we discussed the efficiency of knowledge transfer for reducing the calculation cost.