Experimental Evaluation of Sensor Redundancy and Filtering for Human Tracking in Industrial Collaborative Applications

Collaborative robots are assuming an increasingly relevant role in manufacturing scenarios. However, they are often used in a conservative way. To overcome this limitation and to implement an effective human-robot collaboration it is crucial to improve the knowledge about the sensory system which supervises and ensures safety during the collaboration. In this work, an experimental scenario has been developed to investigate the effect of sensory redundancy and filtering on the safety and fluency of the collaboration. Various strategies, such as camera placements, levels of occlusions, and obstacles (static and dynamic) are considered to evaluate the raw and filtered data from one and multiple sensors. A detailed comparison has been drawn for the various human tracking methods through the introduced metrics, to study the effects and shortcoming of the employed filters and fusion techniques when used for collaborative applications, specifically through a Speed Separation Monitoring safety controller.