Evolutionary algorithm for global design of locomotion systems

Proposes a method for the design of complex mechatronic systems consisting more precisely of locomotion systems. This method is based on a modular approach to the system design. The objective in the design process is to find an optimal mechanical architecture and an associated control to achieve displacements over complex surfaces. For this, we use an evolutionary algorithm integrating a dynamic simulation of the robotic system in its environment. We use a hybrid encoding which allows for a simultaneous evolution of the mechanical structure and its control system. Specialized genetic operators have been designed to manipulate this encoding and to adapt their action to the evolving population. The robot performances are evaluated through a simulation in which all criteria can be computed and exploited in real time. Hierarchical evaluation is then suggested for the improvement of computing time and the process is illustrated through a set of design examples