KGLAMP: Knowledge Graph-Guided LLM for Multi-Robot Planning
A new framework called KGLAMP has been developed by researchers to assist large language models in planning and replanning for diverse multi-robot teams using a knowledge graph. This graph captures relationships between objects, spatial accessibility, and the abilities of robots, allowing the LLM to produce precise PDDL problem specifications. It acts as a continuously updated memory that integrates fresh observations and initiates replanning when inconsistencies arise. This method overcomes the shortcomings of traditional PDDL planners, which depend on manually created symbolic models, as well as LLM-based planners that frequently overlook the diversity of agents and uncertainties in the environment. The findings are available on arXiv (2602.04129v2) and focus on long-term missions that necessitate coordinated planning among various capabilities.
Key facts
- KGLAMP is a knowledge-graph-guided LLM planning framework for heterogeneous multi-robot teams.
- The knowledge graph encodes object relations, spatial reachability, and robot capabilities.
- The framework guides the LLM in generating accurate PDDL problem specifications.
- The knowledge graph serves as a persistent, dynamically updated memory.
- It incorporates new observations and triggers replanning upon detecting inconsistencies.
- Classical PDDL planners require manually crafted symbolic models.
- LLM-based planners often ignore agent heterogeneity and environmental uncertainty.
- The paper is published on arXiv with identifier 2602.04129v2.
Entities
Institutions
- arXiv