Design and Implementation of Running Motion Generator for Humanoid Robots using SLIPM and CPG

The purpose of this research is devoted to establishing a running generation technique for bipedal robots. The motion generator is based on the Bipedal Spring-Loaded Inverted Pendulum Model (Bi-SLIPM) and is designed to stabilize running motions. This work incorporates human motion features during the generator process. First, this paper introduces the bipedal robot used for the tasks. The 2D Bi-SLIPM is employed to generate the running pattern. To ensure the applicability of the running motions generated by the 2D Bi-SLIPM in real-life scenarios, this paper introduces the concept of Discrepancy of Leg Compression (DLC) based on the 2D Bi-SLIPM. Additionally, a Central Pattern Generator (CPG) is integrated as an auxiliary synthesizer to generate DLC-CPG motions. These motions serve as the motion generators for the abduction direction of the hip and the ankle. Moreover, to address the challenges posed by tremors and inherent instability in running movements, this paper proposes the integration of feedback data from an inertial measurement unit. The insights gathered from the inertial measurement unit are incorporated into a heuristic algorithm, which systematically explores and refines parameters for the gait model. Through this iterative training process, the resilience and stability of the running gait are significantly improved. Ultimately, the training results demonstrate the successful realization of the proposed motion generation and control techniques on the physical robot.