HOSA: An End-to-End Safety System for Human-Robot Interaction

The advancement of collaborative robotics increases process efficiency. However, humans are still part of the loop in many deployment scenarios. They are unpredictable factors that may potentially become at risk. This work proposes a safety system for Human-Robot Interaction (HRI), called HOSA, and discusses the decisions made from its modular architectural design phase to a real scenario implementation. HOSA is an end-to-end system that considers the information provided by sensors available in the environment, the communication network to transport the information, the reasoning in the information, and the interface to present the risks. It applies deep learning algorithms to detect HRI collision risk and the use of Personal Protective Equipment (PPE) based on surveillance camera images. Also, it considers knowledge representation based on Ontology, Software-Defined Wireless Networking (SDWN), and a user interface based on augmented reality. The benefits of the proposed design are evaluated through a use case of a HRI scenario for radio base station maintenance. The architecture scales with the number of devices due to semantic descriptions and an adequately provisioned communication network. It demonstrates the system’s efficiency in detecting risk during HRI tasks and alerting people in the scenario. The conducted experiment shows that the system takes 1.052 seconds to react to a risky situation.