Potential for biomimetic artefacts with vibrissal active touch
Vision and touch complement one another. By closing our eyes, we can confirm that the capacity to reach out and explore surrounding surfaces with our fingertips is no real substitute for being able to look and see. However, as users of ‘virtual reality’ systems can confirm, a visual experience—however realistic and compelling—can lack the element of engagement and grounding if you cannot touch the objects you are seeing. By building robots, or other intelligent artefacts that can observe the world, but that cannot physically sense its surfaces, we deny our artefacts a capacity possessed by even the simplest animals.
Fingertip touch and whisker touch
Research on the artificial equivalent of human touch, like that on synthetic vibrissal sensing, is a relatively new and emerging field. Insights drawn from one field of active touch will inform and advance the other. However, for many practical applications, there are good reasons why a vibrissal sensing system might be preferable to a finger-like probe. Specifically, skin-style sensors suffer from the problem that the transducer is at the periphery of the device and therefore can be easily damaged by an unexpectedly sudden, abrasive, or noxious contact. The “business end†of a vibrissal sensor, in contrast, is a passive, easily replaceable, shaft; the sensitive transduction apparatus is stored safely in-board at the base. Indeed the natural whisker can sustain considerable damage and still remain useful. In this regard, rats show a remarkable capacity to adapt relevant neural substrates to changes in the configuration or length of their whiskers, and we would expect to provide similar robustness and adaptivity in the artificial vibrissal systems we build here.
The long and short of whisker sensing
Rats and shrews have two main types of whisker sensor: the longer, actuated macrovibrissae and the shorter, more densely arranged microvibrissae. Though both types are capable of detecting tactile features such as texture and shape, there appears to some specialisation and complementarity of function. Specifically, the longer macrovibrissae serve as a highly effective means of pin-pointing interesting objects in space, whereas the microvibrissal array may be the most optimal device for close examination of surface properties. In this project we will devise analogs of each and of their combination.
Feeling our way
Actuation of the macrovibrissae must provide significant advantages since it requires expenditure of energy. Two likely benefits are that it provides more degrees of freedom for sensor positioning, and that it allows the animal to sample a larger volume of space with a given density of whiskers. However, beyond this, we believe that the ability to employ alternative whisking strategies in different contexts may constitute the principal gain. In other words, independent control of the whiskers allows the animal select both the position of the shaft or tip of the whiskers, and whisker motion parameters, so as to provide the maximum amount of task-relevant information. An artificial vibrissal sensor with this capacity will be very different from the blundering binary collision-detectors that provide the tactile sensing competence of most contemporary robots. The BIOTACT sensor will, literally, “feel its wayâ€, dancing round an object with purpose and accuracy to extract a rich tactile percept of its contours.
Potential applications
We envisage a variety of possible industrial and societal applications for vibrissal sensing some of which are outlined below.
i.  Object detection and sorting tasks. The task of sorting items by texture, for classification or quality control, could be effectively addressed using tactile sensing. The remarkable speed and accuracy with which the star-nosed mole, with its unusual tactile sensing organ (see panel 5), detects, identifies, and then consumes tiny prey animals is an inspiration in the this regard.
ii. An immediate impact of vibrissal object recognition could occur in consumer robotics. For example iRobotTM's automated vacuum cleaner currently operates using a navigation algorithm and collision avoidance ability but with no object recognition—all objects seem the same to these robots, and consequently they tend to get lost. Tactile texture and shape recognition could provide useful landmarks for self-localisation and mapping, and could help distinguish different fabric types on floors/furniture to allow the selection of optimal investigation and cleaning strategies.
iii. Texture detection in guidance systems for autonomous off-road vehicles, climbing robots, and planetary explorers. Vibrissal sensing of ground and surface texture could aid in the control of mobile automata exploring rough terrain. For these applications accurate assessment of frictional forces is important for optimising control.
iv. Search and/or rescue robots are often required to operate in cluttered, heavily particulated environments such as dust- or smoke-filled collapsed buildings, where even bright illumination will not allow good visual acuity. An artificial whisking system could provide effective local navigation for a robot, helping it to pick its way through rubble and detritus.
v.  Inspection in enclosed environments, such as ducting systems, could use vibrissal sensors as an aid to search and to investigate and quantify blockages or damage.
vi. Observations of whiskered sea mammals demonstrate that vibrissal systems can also operate effectively underwater. We see applications such as searching for objects in turbid water, for instance, at the bottom of a muddy river. Another possible application, which has been discussed with British Nuclear Fuels, is the investigation of liquid nuclear waste decommissioning tanks.
vii. Drawing on the prey-capture abilities of shrews, vibrissal sensor systems might also be incorporating into automated devices for pest identification and control.
viii. Microvibrissal arrays could be adapted to form an enhanced artificial ‘fur’. Such material could provide a robot, or other artefact, with an exterior touch-sensitive cladding. Alternatively it could be used to enhance human perception, for instance, as cladding for a prosthetic limb. As the technology is reduced is a micro- or nanoscale such systems will become similar to skin like sensors.
ix. In medicine, efforts to help deaf children speak clearly have recently included insertion of thin binary-pressure pads in the mouth so that the deaf person can visualize the differences in tongue contact during speech. This could be greatly enhanced using a variable, direction-sensitive pressure pad (similar to our proposed micro-vibrissal technology) to measure tongue movements.
x.  Interest in artificial tactile sensing has recently emerged in the area of minimally-invasive surgery. Here surgeons are concerned about the loss of tactile information during operations since they are no longer able to use their fingertips to assess tissue properties. A number of sensor types have been trialled but there is much scope for improving on their design and sensitivity.
