IK-Map: An enhanced workspace representation to support inverse kinematics solvers

We present an approach to improve the performance of general purpose inverse kinematics (IK) solvers which are based on iterative gradient descent algorithms. The proposed IK-Map is used to represent the whole workspace of the manipulator through a voxelized data structure, similar to existing approaches, e.g. reachability or capability maps. We extend the reachability map approach by additionally storing reference IK solutions, which can be used to seed iterative IK solvers during online processing. This information can be used to limit the effect of well-known issues with local optimization schemes based on gradient decent methods, such as local minima or constraint violation. We evaluate the approach with a simulated model of ARMAR-4, showing that classical generic Jacobian-based IK solvers can be improved in terms of success rate, performance, and quality of the resulting IK solutions.