Evaluation of 3Di cameras in UGV applications

Abstract

Unmanned ground vehicles (UGVs) enable autonomous operation in challenging and dynamic environments, making them valuable for applications such as search-and-rescue, industrial automation, and exploration. The effectiveness of UGVs depends on robust sensing and mapping capabilities, typically provided by sensor systems such as Light Detection and Ranging (LiDAR) sensors and vision-based cameras. However, these technologies have limitations. LiDAR, while accurate, is expensive and struggles with reflective surfaces, while stereo vision has depth estimation problems in low texture environments. Alternative or complementary technologies are needed to improve UGVs performance.This study investigates the integration of 3Di cameras as an alternative sensing modality for UGV applications. The 3Di sensor captures high-resolution depth in- formation, which is processed and incorporated into a LiDAR-based Simultaneous Localization and Mapping (SLAM) framework. The primary objective of this study is to assess the capability, accuracy, and repeatability of 3Di-based SLAM applications compared to a state-of-the-art LiDAR system. The data show that LiDAR achieves an average mapping error of 0 5 mm, 80% less than the 3Di camera. However, the camera accuracy is not affected by lighting. Both sensors encounter difficulties with transparent surfaces, but are capable of detecting light absorbing materials. The presence of uneven structures causes inaccuracies, especially with the 3Di camera. Though, the camera outperforms LiDAR in cliff detection, identifying cliffs up to 0 does not detect them at all..6m away, while LiDAR.With the given SLAM settings, LiDAR provides high accuracy mapping but is not capable of detecting hazards such as cliffs. The 3Di overcomes these drawbacks and thus improves navigation safety, highlighting its potential as a complementary sensor for UGV applications.