A Tracking-Guidance People Method using a Quadruped Robot and Artificial Vision

摘要

This paper proposes a tracking-guidance people system, developing and deploying a vision-based approach for identifying individuals by utilizing distinct light spectrum ranges, specifically Thermal (for tracking purposes) and RGB (for guidance). In both cases, a lidar system complements the robot's distance estimation to the person, environment mapping, and reactive navigation during the tracking and guidance phases. The system has been tested in structured scenarios (for guiding visitors or first responders through different instances) and scenarios with poor light conditions in potential surveillance cases. Comparative evaluations against state-of-the-art developments have validated the person detection method's robustness and the sensory system's practical functionality for decision-making during the tracking and guidance phases. A video of the test phase is available in the Link.