CORNAV: Construction-Aware Reasoning for Robot Navigation on Active Worksites

2026-10-02

AI Summary

This study addresses the localization inaccuracies and operational risks faced by construction robots due to the absence of architectural drawings, schedule awareness, and safety context. We propose a BIM-free, construction-aware active inference navigation framework that achieves semantic localization by aligning open-vocabulary 3D scene graphs with CAD blueprints. Furthermore, the method translates project schedules into time-varying dynamic constraints and leverages large language models for safety verification and path planning. Experimental results demonstrate that the proposed system increases task success rates from 13% to 72.2%, completely eliminates hard no-go zone violations, and reliably rejects hazardous commands triggered by erroneous labels, thereby enabling safe autonomous navigation in complex construction environments.

Abstract

The construction industry faces persistent labor shortages, low productivity that costs the global economy over $1.6 trillion annually, and one of the highest injury rates among major industries. These factors motivate the use of autonomous robots to improve efficiency and worker safety. Existing language-grounded navigation systems, however, rely on semantic scene understanding alone and lack access to construction-specific context such as architectural plans, evolving work schedules, and safety constraints. As a result, they localize permanent building features unreliably and cannot safely navigate active jobsites. We present CORNAV, a blueprint-grounded, schedule-aware navigation framework that operates from 2D CAD drawings and project schedules without requiring a Building Information Model. CORNAV aligns architectural blueprints against hierarchical open-vocabulary 3D scene graphs to ground object queries, converts project schedules into time-varying navigation constraints, and validates requests through an LLM-based safety module that escalates hazardous zones before planning. An A* planner then enforces mandatory exclusion zones while preferentially avoiding higher-risk areas. Across an indoor office and a real construction site, blueprint grounding raises task success from 13.0% to 72.2% over semantic retrieval alone, schedule awareness eliminates all hard-zone violations, and the safety module correctly rejects hazardous requests arising from mislabeled project schedules.

Problem

• Robot Navigation
• Construction Sites
• Language-grounded Navigation
• Safety Constraints
• Blueprint Grounding

Innovation

• Blueprint-grounded Navigation
• Schedule-aware Constraints
• 3D Scene Graphs
• LLM-based Safety Module
• Construction Robot Navigation