Development and Experimental Validation of Control Algorithm for Person-Following Autonomous Robots

Abstract

Automatic guided vehicles, in particular, and industrial autonomous mobile robots, in general, are commonly used to automate intralogistics processes. However, there are certain logistic tasks, such as picking objects of variable sizes, shapes, and physical characteristics, that are very difficult to handle fully automatically. In these cases, the collaboration between humans and autonomous robots has been proven key for the efficiency of industrial processes and other applications. To this aim, it is necessary to develop person-following robot solutions. In this work, we propose a fully autonomously controlling autonomous robotic interaction for environments with unknown objects based on real experiments. To do so, we have developed an active tracking system and a control algorithm to implement the person-following strategy on a real industrial automatic-guided vehicle. The algorithm analyzes the cloud of points measured by light detection and ranging (LIDAR) sensor to detect and track the target. From this scan, it estimates the speed of the target to obtain the speed reference value and calculates the direction of the reference by a pure-pursuit algorithm. In addition, to enhance the robustness of the solution, spatial and temporal filters have been implemented to discard obstacles and detect crossings between humans and the automatic industrial vehicle. Static and dynamic test campaigns have been carried out to experimentally validate this approach with the real industrial autonomous-guided vehicle and a safety LIDAR.

Domain Adaptive Subterranean 3D Pedestrian Detection via Instance Transfer and Confidence Guidance

“…Introducing 3D pedestrian detection for subterranean scenes is capable of acquiring the locations of pedestrians in an emergency situation, thus providing instruction for evacuation and rescue. Compared with images captured by the RGB camera, the LiDAR point cloud has the ability to assist in locating the 3D spatial position of pedestrians and is hardly affected by the poor illumination in the subterranean scenes [4]. However, existing LiDAR-based 3D pedestrian detectors rely on large-scale annotated point cloud datasets for supervised training.…”

Comparative Analysis of Metaheuristic Optimization Methods for Trajectory Generation of Automated Guided Vehicles

“…These vehicles take care of many logistical tasks, tedious for the human operator. In many cases, they share space with workers [1]. In the realm of autonomous guided vehicles (AGVs), trajectory planning is crucial for safe and efficient navigation.…”