Graphs of Convex Sets for Lifelong Manipulation Motion Planning

Authors

Yakovlev K.

Annotation

Robotic manipulation includes many complex tasks that cannot be solved using predefined trajectories alone and require more flexible and sophisticated planning methods. One promising direction is planning with Graphs of Convex Sets (GCS) in the configuration space. This approach has been shown to be effective on many robotic manipulation tasks. However, many practical manipulation scenarios are online: goal states may arrive sequentially, and the planner must reuse previously constructed information while adapting to new requests. In such settings, constructing a complete decomposition of the entire scene in advance may be unnecessary and inefficient, especially in high-dimensional configuration spaces where building convex regions is computationally expensive. Therefore, in this work we focus on an online planning setup in which goal states are sent to the system iteratively. Rather than relying on a fixed precomputed decomposition of the whole scene, we incrementally refine and expand the graph of convex regions only when needed, as new goal states arrive. We evaluate our approach on different planning scenes and compare it with sampling-based and optimization-based baselines in terms of planning time and trajectory quality. The results show that our method is faster while remaining competitive in trajectory quality.

External links

DOI: 10.1007/978-3-032-34387-1_11

Download the collection of works  (PDF) from Springer Nature: https://link.springer.com/content/pdf/10.1007/978-3-032-34387-1.pdf

ResearchGate: https://www.researchgate.net/publication/411904512_Graphs_of_Convex_Sets_for_Lifelong_Manipulation_Motion_Planning