Active Sensing

Many animals actively control the position of their sensory apparatus so as to ‘home-in’ on interesting features of the environment and so maximise their capacity to obtain useful, behaviourally-relevant information. As strongly visual creatures, we are most familiar with this in the domain of vision. Salient stimuli falling in the periphery of the visual field cause rapid movements of the eyes and head that bring the highest-definition area of the retina, the fovea, to bear upon points of interest. However, our eye-movements are also inextricably linked to cognitive processes concerned with current goals and motivations. As such they are strongly dependent on the task in hand—gaze is directed not towards the visual features with the greatest brightness or colour contrast, but at those places in the world that can provide the best information for “getting the job done”. Eye movements are also proactively and predictively controlled. Thus, for instance, a batter in the game of cricket will fixate the expected bounce-point of the ball, just ahead of its impact, so as to best observe its subsequent movement from the ground to the bat. Research in active vision has forced a rethink of some fundamental assumptions about the computations underlying visual object recognition, scene-analysis, and the visual guidance of movement; it is also transforming the way we design artificial vision systems. Insights from this field are even raising basic questions about the nature of visual consciousness, blurring traditional boundaries between perceiving and acting, and placing new emphasis on the sensorimotor-dependent nature of experience.

Active Touch

Visual sensing has been extensively researched both from analytical perspectives in psychophysics and neuroscience, and through synthetic approaches in computing and robotics. Whilst there is still much to be learned, the BIOTACT project proposes to investigate active sensing in an equally important but relatively under-explored modality—that of tactile experience or touch. For humans, the sensations we obtain through touch depend on how we position and move sensitive surfaces such as the tips of our fingers. Active control of the hand and digits is fundamental to this tactile sense—we stroke a surface to detect texture, palpitate gently to judge shape, press to determine hardness, and so on. This is even recognizable from our language—the verbs “to touch” and “to feel” evoke the notion of action or manipulation as much as they do that of sensation. Many animals, particular ones that are nocturnal or inhabit poorly-lit spaces, have evolved more specialist tactile sensory systems. BIOTACT will focus on two such animals—rats and shrews—that have active touch systems, pictured below, based on arrays of tactile hairs known as whiskers or vibrissae.

In advancing the understanding of tactile sensing in animals, and in developing artificial systems with similar, useful sensory capacities, we anticipate that insights will be obtained concerning tactile perception and cognition that are as important as those revealed by research on vision. In particular, we expect to demonstrate that active touch sensing is likewise regulated by cognitive processes linked to goals, motivations and memories; and that it is also predictive and proactive—anticipating environmental structure and positioning sensors where the best information can be obtained. At the same time, touch is a very different modality to vision. For instance, the three-dimensional nature of the world is more directly accessible through touch than through vision, however, the sampling of the world is typically sparser, leading to a greater need to integrate information over time. By exploring the remarkable sensory capacities of animals that are tactile ‘specialists’ we will seek to understand such differences and identify the strengths and weaknesses of touch as a useful sense modality for intelligent artefacts. By developing sophisticated biomimetic tactile sensors we expect to transform the way that engineers and roboticists think about touch sensing—not as just an add-on to systems that are largely reliant on visual processing, but as a radically different means through which to explore and experience the world.

Project Goals

The primary goal of the BIOTACT project is to develop novel artefacts and computational methods for active touch based on the investigation and understanding of biological vibrissal sensory systems.

The principal artefact we will create will be a novel active tactile sensing array, called the ‘BIOTACT sensor’, with many hundreds of vibrissa-like sensing elements, that can fulfil a wide variety of sensing functions well beyond the capacity of existing tactile sensors. This goal will be achieved by developing a succession of prototype sensor designs. Through closely co-ordinated research in neurobiology, computational modelling, and engineering we will emulate, using these active sensing devices, the capacity of natural vibrissal systems to

1. generate fast rhythmic whisking behaviour, exquisitely modulated by sensory feedback, that allows precise positioning of the vibrissae in a manner that maximises information uptake;
2. accurately determine object properties such as position, shape and texture; encode tactile memories that can support recognition of familiar items and places; and make rapid and accurate decisions about object identity and spatial location.

A second artefact to be developed will be an autonomous shrew-like, whiskered robot with the ability to determine, using actuated vibrissae, the shape, position, and movement, of small, mobile target objects, sufficient to allow their detection and tracking. The functionality of this robot will be modelled on the remarkable capacities of insectivores to seek-out, identify, track, and capture fast-moving prey animals.

In addition to these scientific and technological objectives the project will also include a range of networking and training programmes designed to foster and accelerate research within the wider community of active touch researchers, and to train young scientists in the multi-disciplinary skills required to take the field forward.