Modular Topology Optimization of a Humanoid Arm

The development of humanoid robots is typically driven by control, kinematics and actuator design, often resulting in unnecessarily heavy structural components. Topology optimization can be used to reduce weight. Unfortunately, topology optimization is numerically expensive for a multi-component robot system that assumes different poses and considers several materials for each segment. This paper proposes a top-down approach where components of the robot are optimized separately. In the first step, requirements on the stiffness of the entire structure for different poses are decomposed into precise displacement constraints for each component. In the second step, these constraints are used for topology optimization of the individual components for each material. The approach is applied to a two-segment humanoid arm for ABS, aluminum and titanium. The resulting designs, obtained at moderate computational cost, satisfy all stiffness requirements for all poses considered.