Abstract
In this article, a new 7-DoF humanoid arm that can operate a steering wheel anthropomorphically, and a task-based design method for such redundant manipulators that work with loads in constrained spaces are devised. The latter consists of a controller-integrated optimization method (COM) for structural parameter optimization, and a combinatorial filtration method (CFM) for actuator selection. The COM is based on kinetostatic analysis and implemented through the covariance matrix adaptation evolutionary strategy (CMA-CS). In the optimization, a collision-free control (CFC) method is integrated for the accurate workspace evaluation and the manipulability index for characterizing the kinematic performance. The CFM is based on the results of the COM and a dynamic analysis. In actuator selection, a library of available candidate motors and reducers is constructed for global selection based on the design objectives and the characteristics for more accurate joint parameters. The effectiveness of CFC in structural parameter optimization is verified in the CFM. Thus, a 7-DoF humanoid arm for driving is achieved; its performance is evaluated in real-world experiments to assess the validity of the design.