Multi-sensor fusion and semantic-geometric navigation for robotic monitoring of improper micromobility parking on pedestrian pathways

초록

Shared micromobility vehicles (SMVs), such as e-scooters and e-bicycles, have become an integral component of urban transportation systems. However, improper parking of SMVs on pedestrian pathways disrupts mobility and poses growing safety concerns. This article presents an autonomous robotic monitoring system designed to detect, geolocate, and report improperly parked SMVs in urban environments. To navigate safely, the robot employs a multi-sensor fusion localization framework for reliable positioning in global navigation satellite system (GNSS)degraded areas, alongside a semantic segmentation network for pathway-constrained autonomous navigation. Simultaneously, a deep learning-based perception module detects improperly parked SMVs and transmits cropped vehicle images and geospatial locations to a cloud-based monitoring system. This cloud-connected monitoring architecture issues real-time notifications to shared micromobility service providers (SMSPs), allowing immediate visual verification and intervention. Field experiments over a 1.32 km urban-like campus route demonstrated reliable end-to-end performance, including a low intervention ratio of 2.28%, a detection precision of 86.2%, a recall of 92.6%, an F1-score of 89.3%, and an average cloud-reporting latency of 2.27 s. These results validate the feasibility of the proposed system as a practical expert monitoring tool for improving micromobility management and ensuring the safety and accessibility of public spaces.

키워드