AI-Powered Collision Avoidance Safety System for Industrial Woodworking Machinery

Applying Artificial Intelligence technology to safety-critical industrial equipment requires preliminarily studies on the efficacy and limitations of such technology, to enable the definition of normative certification frameworks. In this chapter, we present the prototype of an ultrasound-based collision avoidance system for industrial woodworking machinery. Using a single ultrasound sensor, the prototype can identify the presence of an operator in less than 13 milliseconds, with high sensitivity (97.3%) and specificity (98.6%) also in the presence of noise. The solution presented is able to leverage increasing amount of data over time to increase accuracy, improving the model while always keeping the inference adequate for the memory, power and latency constraints of real-time execution on an embedded microcontroller unit.