화재 감시용 자율주행 로봇을 위한 복합 센서 기반 확률적 증강형 환경 인지 시스템

초록

This paper proposes a probabilistic environmental perception system based on multi-sensor fusion to enhance the detection reliability of autonomous fire surveillance robots. Since indoor fire signals are spatially heterogeneous and heterogeneous sensors exhibit distinct response and noise characteristics, reliable hazard perception is challenging with simple threshold-based approaches. The objective of this study is to visualize and quantify indoor fire risks in the form of continuous maps by integrating multi-channel sensor data with the robot's spatial information. To achieve this, the proposed framework normalizes multi-channel sensor data to estimate point-wise hazard levels and employs a probabilistic accumulation and spatial diffusion algorithm based on Gaussian kernels. Experimental results using indoor driving data demonstrate that the proposed method represents hazardous regions more consistently and confirms the effectiveness of fire-risk perception and map-based environmental representation.

키워드