A Hybrid Control Framework for Dynamic Locomotion of Bipedal Mobile Robot

Dynamic legged locomotion for Biped robot is challenging because of the instability of robot structure itself and short stance time while moving. In this paper, we propose a hybrid controller combining force-and-moment based Model Predictive Control (MPC) and use a whole body controller to optimize the joint control for a 12 degree-of-freedom (DoF) bipedal robot. In the framework, optimal reaction forces and moments are found by MPC using single rigid body dynamic model, including 3-D forces and 3-D moments at each foot. Then the reaction forces and moments will be further optimized by WBC controller, considering the surface contact constraints. Finally, we use the recursive Newton-Euler algorithm to map the optimized forces and moment to joint torque of 6 actuators on each leg. The proposed controller is validated by a 12 degree-of-freedom bipedal robot on Gazebo. The robot can achieve walking speed of 1.5m/s with accurate velocity tracking and the controller can also complete lateral moment, turning and static double leg stance using the same set of control parameters.