Tactile Sensor Arrays for Incipient Slip Detection: Simulation, Instrumentation, Optimisation and Algorithm Development

Tactile sensing allows humans to directly engage with their environment, significantly shaping the way we interact with the world around us, playing a crucial role in our daily interactions. In the realm of robotics, tactile sensing holds immense potential, particularly in enabling the autonomous performance of tasks involving grasping. Integrating tactile sensing into robots is crucial for equipping them with human-like touch capabilities. A vital facet of human touch is slip detection, which involves identifying relative movement between an object and the skin, and aids in determining optimal grasp strength. This capability enables robots to detect grasp insecurity and trigger a response from preventing object drops in manipulation tasks. This thesis delves into the realm of tactile sensing for robotic applications, aiming to enhance slip detection capabilities and promote incipient slip phenomena. The research encompasses a multifaceted exploration of simulation exploration, sensor design and optimisation, algorithm development, and real-world application. In this thesis, a simulation model was established to validate the operational principles of the PapillArray sensor (a novel tactile sensing methodology), providing a platform for refining its design and accounting for potential manufacturing variations. An improved slip detection algorithm, capable of detecting rotational slips, was developed for array-based sensors like the PapillArray. Their efficacy was verified with ground truth reference slip instances through external camera tracking, enabling algorithm parameter optimization, comparison, and performance evaluation. The proposed algorithm was successful in recognizing slip occurrences using PapillArray data (precision of 85% and recall of 90%), and in detecting incipient slip before gross slip occurs across a range of velocities of translational and rotational movements. An innovative optical instrumentation method was then introduced, significantly reducing the sensor thickness to just 7 mm while maintaining performance. Tactile sensor arrays were built and calibrated using this transduction method, featuring an increased number of sensing elements and an optimization of sensor geometry to encourage early incipient slip occurrences and extended incipient slip duration across a wider range of gripping forces to allow sufficient time for a robotic gripper to react to the detect slip events. The developed slip detection algorithm proved effective on the new tactile sensor array, with the results showcasing improved incipient slip behaviour achieved through the extension of sensing elements from 9 (3x3 array) to 25 (5x5 array) and the design optimization of the pillar height differences. The 5x5 optimized sensor gave the best F1 score (0.70) and the best percentage of tests (96%) where incipient slip was detected before gross slip happened. Finally, the practical application of the developed algorithms and tactile sensors is demonstrated in a real-time robotic manipulation task, validating the algorithms' efficacy, and assessing the tactile sensors' capabilities in object handling. This thesis aims to highlight the significance of incipient slip detection for robots, mirroring its importance in human interaction. Advocating for the integration of slip detection into future tactile robot systems that operate within our complex environment. The thesis makes several valuable contributions to the realm of tactile robotics by advancing understanding in the domains of tactile sensor development, incipient slip, and tactile robot grasping.