Velocity Control of Serial Elastic Actuator Based on Ekf Estimator and Neural Network

Yichao Mao,Rong Xiong,Qiuguo Zhu,Jian Chu
DOI: https://doi.org/10.1142/9789814525534_0091
2013-01-01
Abstract:Serial Elastic Actuator (SEA) is a kind of advanced actuator which can efficiently adapt environment disturbance because it has better compliance against external impact than traditional actuator. Previous research work of SEA control are mainly focused on torque control. However, in applications such as manipulator motion planning, the position following accuracy is considered a more critical performance metric, which leads to the requirement of delicate velocity control of SEA. Different with previous PID control method, this paper proposes a velocity controller based on neural network and extended kalman filter (EKF). Neual network is utilized to learn a feed-forward dynamic model of the controller under the condition with a certain payload and pose of SEA, which can compensate nonlinear disturbances due to gravity and the elasticity of spring. And the proposed neural network model is adjusted adaptively using observation data of the actuator payload and configuration acquired by a EKF estimator. Experimental results demonstrate the validity and robustness of the proposed method.
What problem does this paper attempt to address?