RTLOCs: RGBT-LiDAR multimodal fusion for object detection of robot in outdoor

Multi-sensor modality fusion-based object detection systems represent a critical technology for robotic operations in outdoor environments, Contemporary outdoor robotic systems predominantly rely on the fusion of RGB and LiDAR data or RGB and thermal imaging, yet continue to face substantial challenges in environmental adaptability. To address this issue, we propose a hybrid fusion architecture that incorporates thermal imaging into the RGB-LiDAR fusion framework, there by enabling simultaneous perception through three distinct modalities: thermal, RGB, and LiDAR. Furthermore, we constructed a multimodal dataset named UJN-RGBTL tailored for outdoor autonomous robotic navigation based on our robotic platform. Experimental validation on an additional benchmark demonstrates that our approach achieves significant performance improvements compared to unimodal methods, verifying the efficacy of the proposed algorithm and exhibiting robust performance in real-world robotic deployment scenarios.