FBCCD: A Forward and Backward Cyclic Iterative Solver for the Inverse Kinematics of Continuum Robot

Abstract

This paper proposes a new numerical approach called Forward and Backward Cyclic Coordinate Descent (FBCCD), which is based on the Cyclic Coordinate Descent (CCD) algorithm. A specific set of solutions can be found from infinite solutions of multi-segment continuum robots using the iterative numerical algorithm. Inspired by the Forward and Backward Reaching Inverse Kinematics (FABRIK) algorithm, the inverse kinematics (IK) of a multi-segment continuum robot is divided into two phases: a forward iteration of end coordinates and a backward iteration of end direction. Forward and backward iterations correct and compensate each other, making the end pose close to the target. By altering the goal function of a single iteration, the FBCCD algorithm can also be applied to the continuum robot with a movable base. The numerical experiment results illustrate that this algorithm is with higher convergence rate and effectiveness compared with some of the most popular IK approaches. The average operating time for a five-segment continuum robot is 361 ms and the average number of iterations is 22.89.