Learning Singularity Avoidance

07/11/2018

∙

by   Jeevan Manavalan, et al.

∙

0

∙

share

With the increase in complexity of robotic systems and the rise in non-expert users, it can be assumed that a constraint in a task is not explicitly known. In tasks where avoiding singularity is critical to its success, this paper provides an approach, especially for non-expert users, for the system to learn the constraints contained in a set of demonstrations, such that they can be used to optimise an autonomous controller to avoid singularity, and thereby unpredictable behaviour when carrying out the task, without having to explicitly know the tasks constraint. The proposed approach avoids singularity by maximising task manipulability throughout the motion of the constrained system, and is not limited to kinematic systems. Its benefits are demonstrated through comparisons with other control policies.