Language-Model Guidance in Bio-Inspired Robot Path Planning: Reward Design and Reliability

Abstract

Llm-Guided Robot Planning now turns on the ability to balance performance with evidence quality, resource limits, and transfer across settings. This comparative analysis considers using language-model priors without surrendering feasibility, reproducibility, or collision safety. It synthesizes 1 focal paper with 13 independently retrieved publications verified through persistent DOI or publisher records. The analysis is organized around reward specification, search dynamics, constraint handling, sim-to-real transfer, and auditability. The analysis declines to treat reported outcomes as directly interchangeable, the review compares problem boundaries, design logic, and conditions of validation. Across the literature, the literature consistently implies that advances in LLM-guided robot planning become credible when representation, objective, and evaluation protocol are evaluated together and when uncertainty about distribution shift is reported explicitly. This organization relates method selection to decision risk and exposes recurring transfer threats, and proposes a research agenda centered on explicit comparators, boundary tests, and reproducible artifacts.