Advances in Autonomous Robotics and Path Planning

The field of autonomous robotics is moving towards leveraging large language models (LLMs) and their variants to enhance path planning and decision-making capabilities. Researchers are exploring innovative approaches to integrate LLMs with traditional planning algorithms, enabling robots to interpret high-level semantic instructions and navigate complex environments. The use of LLMs as teacher models to train lightweight small language models is gaining traction, allowing for efficient deployment on edge devices. Additionally, the development of hierarchical planning frameworks and probabilistic program induction methods is improving the ability of robots to reason under uncertainty and adapt to dynamic scenarios. Noteworthy papers include: SmallPlan, which presents a novel framework for leveraging LLMs to train small language models for path planning tasks. Semantic Intelligence, which integrates GPT-4 with A* planning to enable low-cost robots to exhibit intelligent, context-aware behaviors. MORE, which enhances the capabilities of language models for zero-shot mobile manipulation planning. NAMO-LLM, which proposes a sampling-based planner guided by LLMs to compute feasible plans for navigating among movable obstacles.

Sources

SmallPlan: Leverage Small Language Models for Sequential Path Planning with Simulation-Powered, LLM-Guided Distillation

Semantic Intelligence: Integrating GPT-4 with A Planning in Low-Cost Robotics

LLM-Guided Probabilistic Program Induction for POMDP Model Estimation

MORE: Mobile Manipulation Rearrangement Through Grounded Language Reasoning

HCOA*: Hierarchical Class-ordered A* for Navigation in Semantic Environments

RobotxR1: Enabling Embodied Robotic Intelligence on Large Language Models through Closed-Loop Reinforcement Learning

Learning Symbolic Persistent Macro-Actions for POMDP Solving Over Time

NAMO-LLM: Efficient Navigation Among Movable Obstacles with Large Language Model Guidance

Built with on top of