Robotic Task and Motion Planning for Long-Horizon Problems

The project focuses on planning methods for robots that must solve complex, multi-step tasks in structured or semi-structured environments, potentially under visual and language-based task specifications. Long-horizon robotic problems require reasoning across multiple levels of abstraction, including task sequencing, symbolic consistency, geometric feasibility, and execution-aware replanning. Task and motion planning provides a principled framework for addressing these challenges by combining high-level symbolic reasoning with low-level motion generation and feasibility checking. This internship emphasizes AI planning methods, including search algorithms, PDDL-based modeling, and related approaches to sequential decision-making in robotics. Depending on the project scope, vision-language models (VLMs) may also be incorporated for semantic grounding, subgoal proposal, perception, or search guidance.