No abstract

No abstract

No abstract

Active touch sensing can benefit from the representation of uncertainty in order to guide sensing movements and to drive sensing strategies that operate to reduce uncertainty with respect to the task at hand. Here we explore learning approaches that can acquire task knowledge quickly and with relatively small datasets and with the potential to be exploited for active sensing in robots and as models of biological sensory systems. Specifically, we explore the utility of deep (hierarchical) Gaussian Process models (Deep GPs) that have shown promise as models of episodic memory processes due to their low-dimensionality (compactness), generative capability, and ability to explicitly represent uncertainty. Using data obtained in a robotic active touch task (contour following), we show that both single-layer and Deep GP models are capable of providing robust function approximations from tactile data to angle and sensor position, with Deep GPs showing some advantages in terms of accuracy and uncertainty quantification in angle discrimination.

Perception of physical features through touch requires the execution of exploratory movements. Modifying the state parameters of the sensory apparatus to obtain relevant information to achieve a task contributes to an efficient manner for exploration of object properties. These principles have served as inspiration in the development of robotics and autonomous systems. Following the contour of an object poses multiple challenges in the perception of the geometry of an object such as identifying the angle of the sensor relative to the edge to perform tangential exploratory motions, and localising the sensor to place it where the angle tends to be perceived with more accuracy. The variability in the acquisition of tactile data may induce inaccuracies in the predictions from the sensor model. This work examines the influence of integrating proprioceptive information for the assessment and update of the parameters of a Bayesian probabilistic model. This inclusion leads to an increment in the number of task completion relative to performing the task with a model trained solely with data collected offline. Studies in biological touch suggest that tactile and proprioceptive information converge synergistically to drawing conclusions about the feature that is in contact with the sensory apparatus, this work provides a method for improving the modelling of the sensor responses to actively perform object exploration under variability of tactile data in the acquisition process.