Inverse Kinematics Solver Based on Evolutionary Algorithm and Gradient Descent for Free-Floating Space Robot

This paper investigates the inverse kinematics (IK) problem of free-floating space robot (FFSR) and proposes an IK solver called IK Solver Based on Evolutionary Algorithm and Gradient Descent for FFSR (EA &GD-Based IK Solver for FFSR). The IK problem for FFSR aims to find a configuration satisfying: the end-effector (EE) reaches a specific pose; the base attitude angle in this configuration is as close as possible to its nominal value. This IK solver includes the initial value module and the gradient descent module. The first module uses evolutionary algorithm to get an initial configuration, where the base attitude angle is set to its nominal value, and EE pose corresponding to this configuration is very close to the goal EE pose. The second module uses the initial configuration as initial value and obtains the gradient by projecting the EE pose error using the pseudo-inverse of the Generalized Jacobian Matrix (GJM). It can further reduce the EE pose error in the first module and obtain a configuration as a solution to the IK problem. As the EE pose corresponding to the initial configuration is very close to goal EE pose, the second module will not bring too much change in the base attitude. Therefore, the base attitude angle of the configuration obtained by the second module is very close to the nominal base attitude angle.