Development of hardware and algorithms for a multi-functional capacitive proximity sensing system

This thesis focuses on the development of a multi-functional capacitive proximity sensor to improve the worker safety during the industrial human-robot interactions. The sensor is to be mounted on the worker and used to maintain a safe distance between the worker and robot or the parts moved by the robot. The response of a capacitive proximity sensor is a function of the actual distance as well as the geometry of the approaching object. This uncertainty can lead to a wrong estimation of distance or possibly a missed detection. The proposed sensing system in this work aims to solve this issue. Three sensing capabilities, namely distance measurement, surface profile recognition, and parallel motion tracking are implemented in a single platform. These capabilities are achieved by a capacitive sensing element coupled to reprogrammable interface electronics. The sensing element features a 4×4 matrix of electrodes that can be reconfigured to different arrangements at run-time to obtain information on the desired parameters of interest (i.e., distance, shape, and trajectory). The control modules are mapped on a field programmable gate array while the capacitance generated by each configuration of electrodes is measured and quantized by a capacitance-to-digital chipset. Digital filters are used to pre-process the raw capacitive data in order to compensate for random walk and environmental interferences such as temperature and humidity variations. Statistical learning methodologies are applied to classify objects and calculate distance values. Quantitative regression models are built to seek out distance values while classification tools including K nearest neighbors, neural network, and support vector machine are employed to recognize the surface profiles. The performance of the sensing modalities is experimentally assessed with lab equipment as well as on an industrial robot. The system can detect objects and classify their geometries at distances up to about 20 cm with high accuracy. Three different surface profiles can be recognized by all the classifiers. Recognizing the shape of the object improved the regression models and reduced the close-distance measurement error by a factor of five compared to methods that did not take the geometry into account. The capability of tracking the parallel motion is demonstrated by combining the capacitive responses from different electrode connection configurations. The breakthroughs made through this work will make capacitive sensing a viable low-cost alternative to existing technologies for proximity detection in robotics and other fields.