Toward a Co-Design Framework for Bipedal Robots

In this paper, a co-design framework for developing bipedal robots capable of steady energy-efficient dynamic locomotion is proposed. In contrast to conventional design methodologies, where the mechanical structure, the walking pattern generator, and the control parameters are sequentially developed, the proposed approach integrates all these components in a unified optimization process. The design problem is formulated as a multiobjective dynamic problem to simultaneously optimize the structural parameters of the robot, the walking pattern generator, and the gains of the low-level controllers. The general objective is to maximize the walked distance while minimizing the energy consumption during locomotion. A modified multiobjective version of the Differential Evolution algorithm was employed to solve the optimization problem, in combination with a mechanism for constraint handling based on Pareto Dominance. The framework generates a solution set with feasible designs, each including the optimal dimensions of the robotic links, the corresponding walking patterns and the controller gains. As a case study, a bipedal robot with anthropomorphic dimensions was synthesized and evaluated using multi-body simulation. The obtained solutions were analyzed in terms of Cost of Transport and compared to available bipedal commercial robots. These results showed that the proposed framework can generate feasible solutions, that were in average 10% more efficient than those of the commercial robots, highlighting the advantages of co-design strategies for the development of bipedal robotic systems.