Map2Route improves language-based route planning on semantic maps
Map2Route: Benchmarking Compositional Language-Grounded Route Planning over Semantic Maps
Robotics
Summary
Planning routes using everyday language is hard because instructions can be complex and involve conditions like places to avoid or preferences about the path. The authors created Map2Route, a new test set with lots of detailed route instructions on maps that label objects and areas. They also developed a method called Grounding2Route to better follow these complex instructions and plan routes. Their method does much better than previous attempts but still struggles compared to human performance, showing there's more work to do.
What this means in practice
- •For navigation system developers: Develop more accurate voice-guided navigation by integrating complex language commands about places to avoid or preferences along routes.$Commercial implications: Enables building smarter consumer GPS apps that better follow nuanced human navigation instructions, improving user satisfaction and safety.
- •For robotics engineers: Program robots to interpret detailed spoken or written directions involving relational and comparative descriptions for indoor navigation tasks.
Authors
Muyi Bao, Hang Xu, Jingfan Tang, Zihan Liu, Yuxin Cai, Chen Lv, Wenshan Wang, Ji Zhang
Abstract
We introduce Map2Route, a human-curated benchmark for compositional language-grounded route planning over pre-built semantic maps. Map2Route contains 1,000 episodes across 40 scenes, where instructions use relational, comparative, and nested descriptions to identify route-relevant objects and regions, while specifying ordered must-pass regions, must-avoid requirements, five categories of soft preferences, and spatial and route-stage scopes, which is partially tested by existing works. Alongside Map2Route, we propose Grounding2Route, which combines executable code-as-grounding with verification-guided repair and scope-aware planning.Across seven representative adapted baselines, Grounding2Route substantially outperforms existing methods in all metrics. Despite these gains, a substantial gap to human demonstrations remains, highlighting the difficulty of Map2Route and the considerable headroom for future progress. Additional qualitative results and resources are available on https://anonymous.4open.science/w/Map2Route-F05F/.