NORM-Nav: Zero-Shot Mobile Robot Navigation with Natural Language Behavioral Constraints
Dongjie Huo, Junhui Wang,, Chao Gao, Yan Qiao, Dong Zhang, Guyue Zhou
AI summary
Problem
Conventional costmap-based robot navigation focuses solely on geometric feasibility, ignoring the semantic and social behavioral conventions expected in human-centered environments, while learning-based alternatives struggle with generalization and data scarcity.
Approach
The framework uses an LLM to parse natural language instructions into structured spatial, velocity, and traversability constraints, which are grounded via real-time vision-LiDAR perception and encoded as multi-layer costmaps compatible with standard grid-based planners.
Key results
- Proposes a zero-shot framework that translates natural language constraints into structured planning cues
- Introduces a modular plug-in design that augments standard costmap stacks without altering the underlying planner
- Demonstrates higher task success rates and better behavioral compliance in simulation and real-world tests
- Generates trajectories that more closely match human-preferred references than representative baselines
Why it matters
It allows mobile robots to navigate safely and socially appropriately in dynamic human environments without requiring extensive training or environment-specific tuning.
Abstract
Mobile robots operating in human-centered en- vironments must generate not only collision-free paths but also trajectories that follow local behavioral conventions. Con- ventional costmap-based navigation emphasizes geometric fea- sibility and often overlooks such requirements, which can result in socially inappropriate behaviors. This paper presents NORM-Nav, a zero-shot framework that integrates natural language behavioral constraints into costmap-based planning. An LLM parses each instruction into structured constraints and grounds them using real-time vision–LiDAR perception. These constraints are encoded as multi-layer costmaps that represent geometric, semantic, directional, and velocity cues and are directly compatible with standard grid-based planners. Simulation and real-world experiments indicate that NORM- Nav improves task success rates and produces trajectories closer to human references than representative baselines. The project website is available at https://ei-nav.github.io/NO RM-Nav.