Nonlinear Oscillations For Planar Bipedal Locomotion And Its Implementation On FPGA

The study of dynamics and control for bipedal locomtion is essential for improving hu-manoid and rehabilitation robotics. Current achievements in bipedal locomtion, such as theASIMO humanoid from Honda, are largely based on a over-restricted control theory. On theother hand, nature evolution has advanced much more dexterous solutions. This thesis focusedbio-inpsired controlling method for bipedal locomotion and the hardware implementation forit. Firstly, simple nonlinear oscillators are applied to a three-link planar biped by driving itsjoint angles. Following previous paper[1,2], the oscillators are obtained by solving linearizeddifferencial equations of hybrid dynamics of the robot. Periodical stability is analysed usingthe Jacobian matrix that are obtained by linearized approximation of the robot’s Poincaré re-turn map[3]and stability area is given on the oscillators’ parameter plan. Simulation resultsare given subsequently. Secondly, a scheme for efficient hardware implementation of centralpattern generators (CPGs) on Field Programmable Gate Arrays (FPGAs) is proposed. A re-vised distributed-arithmetic (DA) algorithm is applied to the implementation to maximize theutilization of look up tables (LUTs) in FPGAs. The proposed scheme performances satisfactoryexperiment results which have a correlation coefficients of0.99with simulation ones. In themean time, it demonstrates74%reduction in LUTs consumption,75%in registers and100%inembedded multipliers.