A Natural Language Interface for Multi-Constraint Spatiotemporal Planning Via LLM-Parameterized Mixed-Integer Scheduling and A*
Sean Ye, Matthew Luebbers, Matthew Gombolay
AI summary
Problem
Traditional spatiotemporal planning requires rigid, expert-level manual specification of constraints and cost functions, making it time-consuming and inaccessible for non-experts.
Approach
The authors developed a neurosymbolic architecture that uses an LLM to translate user prompts into Python code, which dynamically parameterizes a Mixed-Integer Linear Program (MILP) scheduler and an A* motion planner.
Key results
- Designed a language-to-spatiotemporal planning interface mapping natural language to formal constraints
- Formulated a neurosymbolic architecture bridging LLM-generated code with MILP scheduling and A* motion planning
- Demonstrated significant reductions in cognitive workload and improved usability via a 30-participant user study
- Achieved comparable path optimality to templated interfaces while reducing planning time
Why it matters
Bridges the gap between human intent and complex automated planning, making advanced spatiotemporal tools practical for non-experts in robotics, logistics, and naval operations.
Abstract
Spatiotemporal planning is critically important in fields like robotics, logistics, and naval operations, especially for problem specifications involving multiple constraints. Tra- ditional approaches place the burden on end users to manually specify cost functions, constraints, or model parameters, a time- consuming and laborious process often resulting in less-than- ideal plans. We present a novel architecture integrating an LLM-based natural language interface with MILP scheduling and A* motion planning for multi-constraint spatiotemporal planning. We validate our LLM-planning approach through a within-subjects user study using a simulated maritime route- planning domain against manual control, and against au- tonomous planning with classical template-based constraint specification. Results showed our LLM-planning approach not only improved usability and reduced workload over alternative input modalities but also maintained the path optimality of traditional constraint specification interfaces while decreasing planning time. These findings demonstrate that bridging LLM- powered interfaces with robust schedulers and motion planners can enhance human-autonomy interaction in complex planning tasks, potentially making advanced spatiotemporal planning tools more practical for a broader range of users.