Quality Cost A* Path Planning for Multi-Sensor Fusion in Corridor Smoke Scenarios

Area of Science:

• Robotics
• Sensor Fusion
• Environmental Hazard Navigation

Background:

• Indoor fire smoke significantly degrades camera and Lidar sensors, compromising robotic navigation safety.
• Current path planning methods inadequately address sensor degradation, often using empirical penalties without physical grounding.

Purpose of the Study:

• To develop a physically-based path planning algorithm, Quality Cost A* (QC-A*), for robotic navigation in indoor fire smoke.
• To improve robotic safety and reliability by accounting for sensor degradation and enabling fault tolerance.

Main Methods:

• QC-A* maps Fire Dynamics Simulator (FDS) visibility data to sensor perception quality using Koschmieder and Beer-Lambert laws.
• A cost function integrated into the A* algorithm guides paths away from high-attenuation zones.
• Multi-sensor fusion is employed for fault tolerance during sensor-specific failures.

Main Results:

• Simulations across four smoke scenarios demonstrated QC-A* significantly reduces low-visibility hazard zones (40.7% to 19.6%) and enhances perception quality (0.067 to 0.177).
• QC-A* achieved a 96-100% planning success rate under sensor failure tests, outperforming Camera-Only (48%) and Lidar-Only (70%) approaches.
• The algorithm shows a favorable safety-efficiency balance, with a modest 15.3% path length increase in heavy smoke.

Conclusions:

• QC-A* provides a physically grounded, interpretable, and generalizable framework for robotic path planning in fire environments.
• The method effectively models sensor degradation, prioritizing perceptual safety while maintaining navigation efficiency.
• QC-A* enhances robotic resilience and safety in hazardous, smoke-filled indoor settings.