LLMs End2End Planning:
In this project, we explore the application of Large Language Models (LLMs) for autonomous path planning and decision-making. By leveraging the advanced reasoning capabilities of LLMs, we aim to create a robust end-to-end planning system that can understand complex spatial and contextual instructions. Our approach integrates LLMs to interpret natural language prompts, translate them into actionable plans, and dynamically adapt to obstacles and changes in the environment. This project investigates LLMs as standalone planners and in hybrid setups, where traditional algorithms like A* or RRT enhance reliability and precision.
A RAG system to retrieve many-shot demonstrations for LLM-based planning with structured reasoning.
A VLM-conditioned 2D diffusion planner for vision- and language-guided trajectory generation.