A local manipulation path replanning algorithm on deformable linear objects for collisions resulted from model deviation

Robotic manipulation on deformable linear objects asks for action path planning firstly. However, the simplified cable model used for promoting planning efficiency tends to deviate from the real shape of manipulated deformable linear objects, which will lead to collisions and the original planned manipulation path failing, which brings a block between path planning and its robotic implementation. Aimed at the planning failure problem, this paper raises a novel replanning algorithm. Firstly, the deviation and resulted collision situations are classified and the severity of deviation is assessed with a partition policy to determine the moment for performing path replanning. Then the model of the cable in collision is formulated, and a refined artificial potential field is constructed to generate a constrained collision-free path for the deformable linear object trapped in deviation and collisions. Moreover, a time-decaying policy is implemented to fuse the original planned path and the replanned path to avoid further collisions. Finally, the effects of the algorithm are verified by applying it to multiple experiment scenes constructed with the physics engine. It can be concluded that with the replanning scheme, the effector keeps the manipulated cables from severe deviation and collisions adaptively and the safety of the off-line planned path is improved.