Removing Infrastructure Barriers in Human-Robot Collaboration Through Wireless Reconfigurable Cells

Research Context and Motivation

Human-Robot Collaboration (HRC) is essential for modern industrial tasks like remanufacturing, which require high flexibility and frequent workspace reconfiguration. Traditional robotic setups rely on extensive power and data cabling, which creates physical hazards, limits mobility, and makes rapid layout changes expensive and slow. This paper addresses these infrastructure barriers by proposing a wireless, 5G-based, battery-powered multi-sensor platform that offloads computationally intensive perception tasks to the edge.

System Architecture and Approach

The researchers designed a modular, robot-mountable platform that integrates directly onto standard cobot flanges. The system utilizes a custom carrier board featuring a System-on-Module (SoM) and a 5G modem to transmit high-bandwidth camera streams without cable constraints. To ensure safety and adaptability, the team implemented a computer vision pipeline based on the YOLO architecture. This pipeline performs real-time object detection, oriented grasp-pose estimation, and robust hand recognition. By detecting human hands, the system can dynamically suppress robot motion to prevent accidents. The model was trained using a combination of synthetic data and real-world images, utilizing domain randomization to maintain accuracy across varying lighting and background conditions.

Experimental Validation

To demonstrate the portability and robustness of the architecture, the system was deployed and tested at two international sites: HUN-REN SZTAKI in Hungary and the University of Oslo in Norway. The researchers evaluated the system across a diverse range of 5G environments, including private Standalone (SA), public SA, and private Non-Standalone (NSA) networks. The experiments confirmed that the architecture can achieve round-trip response times as low as 12 ms, which is sufficient for safe, real-time HRC. While the results validate the feasibility of wireless, cable-free workcells, the authors emphasize that current 5G infrastructure still faces significant interoperability challenges that must be resolved for widespread industrial adoption.