Radar and Infra-Red array Sensor Fusion in a Robotized Environment: An Experimental Study

The paper describes a sensor fusion architecture and develops a detection/localization algorithm that fuses the sensor data obtained from several environmental sources, namely a network of 122 GHz frequency modulated continuous wave (FMCW) radar, a 100 GHz imaging camera, and a network of Infra-Red (IR) array sensors. The capabilities of the proposed system are addressed based on several measurements and use cases tailored for a robotized environment and human-robot fenceless cooperation scenarios in manufacturing. The paper proposes methods for verification of safety procedures, analyzes localization, human-robot distance estimation accuracy, and end-to-end latency introduced by fusion and machine learning algorithms. It also quantifies the interference/noise sources introduced by moving robots/machinery, for different collaborative setups. The proposed platform is shown to effectively exploit heterogeneous sensors and data fusion techniques to improve performance. Lessons learned from the experiments are also discussed as well as the relative strengths/limitations of the proposed system.