Cognitively inspired artificial bipedal humanoid gait generation

摘要

The biomechanical gait is a complex and robust task that demands little conscious effort of an individual to perform. This article proposes a methodology for artificially generating a bipedal humanoid gait pattern for the knee and hip leg joints in the sagittal plane. The lower limbs movement is divided into three complementary patterns of movement, each with a specific function, which can be modulated individually and superposed to compose a gait pattern. This strategy is based on the potential field navigation algorithm concept [1] and the architecture for its application is based on the AuRA cognitive architecture [2]. The methodology was validated using a MATLAB implementation of a simplified kinematic model of the legs movements. The results show movement patterns close to the biomechanical ones at the joints. When applied to the kinematic model, those movement patterns generate feet trajectories that are consistent to the biomechanical gait both in shape and amplitude. The strategy was also embedded in a microcontrolled prototype, revealing low computational demands as well as another view for the generated movement.