D-LGP: Dynamic Logic-Geometric Program for Reactive Task and Motion Planning

Many real-world sequential manipulation tasks involve a combination ofdiscrete symbolic search and continuous motion planning, collectively known ascombined task and motion planning (TAMP). However, prevailing methods oftenstruggle with the computational burden and intricate combinatorial challenges,limiting their applications for online replanning in the real world. To addressthis, we propose Dynamic Logic-Geometric Program (D-LGP), a novel approachintegrating Dynamic Tree Search and global optimization for efficient hybridplanning. Through empirical evaluation on three benchmarks, we demonstrate theefficacy of our approach, showcasing superior performance in comparison tostate-of-the-art techniques. We validate our approach through simulation anddemonstrate its reactive capability to cope with online uncertainty andexternal disturbances in the real world. Project webpage:https://sites.google.com/view/dyn-lgp.