Cyclic gait planning and control of underactuated five-link biped robot during single support and impact phases for normal walking

Gait planning is an important key in motion control of biped robots. This paper focuses on developing a cyclic gait based on capturing real human data using Kinect-V2 camera and a control for normal walking of underactuated five-link biped robot during the SSP and impact phase. Since human motion is done optimally following the principle of optimality, the ankle and hip trajectories of a human subject during single support phase (SSP) are captured, filtered and averaged over steps to remove random variations in the data. The ankle and hip trajectories are fitted using suitable polynomial functions depending on geometrical, velocity and acceleration constraints. Then, the desired joint angle profiles are obtained as a function of those ankle and hip trajectories using inverse kinematics. Additionally, feedback linearization algorithm for controlling and stabilizing the underactuated biped during normal walking is proposed. The mathematical model of underactuated biped robot is divided into two subsystems: actuated states and underactuated one. In order to control and stabilize all these system states, the control signals of actuated states affect the underactuated state indirectly. The controller parameters are optimized using Genetic Algorithm. Finally, the stability analysis of the biped robot during normal walking is discussed using the linearized Poincare map. The results show that the proposed gait trajectory generation and control are suitable for the underactuated biped robot to walk normally in a natural manner. Also, the biped accurately follows the desired trajectories.