Multi-scale, Reactive Motion Planning with Deformable Linear Objects

Focus Research dealing with the manipulation of flexible objects has been on for a while. Most of the existing motion-planning algorithms that determine a trajectory from a start to the goal state revolve around traditional sampling-based and feedback control methods. What has not been done so far is the realisation of desired goal states from different initial states through reactive, global planning strategies. This study aspires in creating a robust and reactive planning strategy for deformable linear objects (DLOs). Since we concentrate on knot tying and untying as the basic tasks to be deployed on the DLO, our reactive strategy would be used to create a knot tying/untying planner. Results We are able to construct a reactive and global knot tying/untying planning strategy that can knot or unknot any particular knot type. Simple knot types experimented with are the overhand knot, slip knot and untight knot. We have also been successful with slightly more complex knots like the Hopf Link and the looped overhand knot. The planner shows resiliency to controlled perturbations applied on the system. This approach is compared with the traditional sampling-based Probabilistic Roadmap planner and it is seen that our approach yields results much faster and turns out to be more efficient. Another interesting result we came across is that the Minimum Distance energy that is used in this study to relax knots from their tangled state, is not unimodal as suggested in [20]. The energy unimodality was earlier empirically refuted in [22]. Conclusion Through this research we have been able to show that traditional methods for motion planning are slower as compared to the multi-scale, reactive methods that are faster and much efficient. The reactive modelling approach is thus ii the better alternative for building motion-planners especially those that involve deformable bodies.