Whiskered robots This article summarises the motivation and history of whisker inspired tactile sensors for robotics and the contribution that this field of research has made in our understanding of the biological analogue.
Touch is currently an under utilised sensory mode in robotics, with vision remaining the preferred method of spatial exploration. There are many examples in the animal kingdom of creatures that live in confined and visually occluded environments where a developed sense of touch rather vision is advantageous for survival. Such environments are challenging operational domains for mobile robots as conventional proximity sensors and artificial vision systems do not perform well. Perhaps more persuasively, there are many situations where a developed sense of whisker-like touch on a robot would serve as a beneficial complement, rather than replacement, for vision. Of particular interest is that the flexible shaft of a whisker provides an intrinsically compliant interface between the robot and a surface under inspection thereby minimising the risk of damage to both. It follows that this reduces the required precision for motion planning whilst not compromising the efficacy of the tactile inspection. In this article the use of whiskers as a method of endowing robots with a sense of whisker touch is summarised through a brief history of whiskered robots. This is followed by a discussion on the utility in adopting a biomimetic approach to whiskered robot development and how this has resulted in a 2-way exchange of ideas between engineers and biologists.
Whisker based sensors have been deployed on mobile robots since the mid 1980's. Initially they were used as simple binary proximity sensors to assist in navigation and obstacle avoidance (Russel, 1984), (Brooks, 1984), (Hirose et al., 1990).
Examples of whiskers used on robots for more detailed spatial exploration can be divided into either active or passive touch approaches. Active touch whiskers measure the bending torque of the whisker as it makes contact with objects during a controlled movement, analogous to the whisking behaviour of rats (Ueno & Kaneko, 1994), (Kaneko et al., 1998)(Top left panel of Figure 2). Passive touch whiskers measure the torque of whisker bend in response to contacts made with objects as the platform on which the whiskers are mounted moves passed them, i.e., the whiskers are not `directly' actuated. For example spring loaded potentiometers have been extensively used to measure this torque (Russel, 1992), (Jung & Zelinsky, 1996), (Wijaya & Russel, 2002). More recent examples of active whisker sensors for surface inspection use load cells to measure the whisker deflection torque (Clements & Rahn, 2006), whilst others use resonating piezoelectric stimulators (Muraoka, 2005) or strain gauges (Solomon & Hartmann, 2006)(Lower left panel of Figure 2). There are also a number of examples of more biologically plausible whisker sensory arrays installed on mobile robots to conduct biomimetic study of the whisker sensory system. The aMOUSE project integrated a bilateral array of real rat whiskers glued to the diaphragm of electrohet microphones and mounted onto a small mobile robot platform to investigate texture classification of surfaces (Lungarella et al., 2002)(Top right panel of Figure 2). The ability to extract surface shape was also demonstrated using an actuated array of flexible artificial whiskers instrumented using Hall effect sensors at their base (Kim and Moller, 2004)(Lower right panel of Figure 2).
The Whiskerbot platform (built as part of EPSRC grant no. GR/S19639/01) had a small array of glass-fibre moulded whiskers instrumented at their base using micro-strain gauges to measure 2-dimensional deflections of the shaft. This analogue information was converted into a series of empirically based spike train models of the rat vibrissal primary afferent (Mitch et al., 2004) and passed to an embedded real-time spike based model of the rodent trigeminal complex and superior colliculus (Pearson et al., 2007).
The SCRATCHbot (Spatial Cognition and Representation through Active TouCH bot) developed as part of the ICEA project (Integrated Cognition, Emotion and energy Autonomy, EU FP6) had a much larger number of whiskers and motile degrees of freedom to position the array within the environment. Hall effect sensors were used for measuring the deflections in the whisker shafts and standard DC motors used to actuate the "whisking" behaviour observed in rats. This platform was used to reproduce different implementations of whisking pattern generation and coordination to allow a quantitative comparison of each hypothesis with regards to the quality of sensory information derived. It was also used to demonstrate a simple model of tactile attention that could be used to direct the exploration of the robot through its environment. Both of these research questions were developed further using the Shrewbot platform (Shown in Figure 1) which was built as part of the BIOTACT project (Biomimetic technology for vibrissal active touch, EU FP7) along with the G2 Sensor which was used to develop algorithms to extract texture and object form from an active tactile whisker array (Sullivan et al., 2012)(Shown in Figure 5). Both of these platforms used miniaturised modular whisker sensorimotor assemblies to form their array.
To design an artificial whisker sensory system that can derive useful tactile information from the environment gives rise to a number of lines of inquiry that are of equal importance to both engineers and biologists. From an engineering perspective, one issue is to identify any advantages of using whiskers for tactile exploration over other approaches, for example, based more directly on human touch (cutaneous touch). Through the observation of whisker development (Sullivan et al., 2006), array morphology (Brecht et al., 2006), and how animals position and move their whiskers during natural exploratory behaviour (Grant et al., 2009), we can intuit advantages such as robustness, speed of response and size of sensory field. Other behavioural studies have found that whiskered animals, such as rat, can discriminate intricate surface features with the same acuity as human touch (Carvell, 1990). Specific questions that are generated through the development of an engineering specification have also inspired further biological investigation. For example, to determine how best to move an artificial array of whiskers to gather information from the environment, video footage of rat whisking has been analysed to develop posits for possible control strategies (Mitchinson et al., 2007). Intriguingly, these model control strategies were then physically tested using the robotic artifact (Pearson et al., 2007) which inspired further investigation of the original biological analogue, such as, how to deal with "noise" generated by whisker self-motion (Anderson et al., 2010).
A biomimetic design approach that encapsulates a two-way exchange of ideas and skills, can be applied to many areas of the design of a useful whisker based sensory system. In addition to modeling the physical mechanics of whiskers to develop a sensory tool for robotics, whiskers also act as extremely powerful "probes" with which to observe the function of the brain. The neural components of the whisker sensory system of the rat maintain an exquisite degree of topological preservation from whisker follicle through to cortex (Kleinfeld et al., 2006). This allows for controlled observations of the neural response to stimuli applied to the whisker shaft at multiple levels of this neuraxis. Model systems can be constructed from such observations which are then reinforced, or at least validated, by behavioural observations of the animal, an approach known as neuroethology (Camhi, 1941). Using robotic artifacts to test such models (computational neuroethology) introduces a degree of experimental flexibility to allow experiments that may be either impossible or unethical to achieve through animal testing. It also provides roboticists the opportunity to work with neuroscientists to evaluate novel brain-based control approaches for future autonomous robotic systems.
- Russel, R.A. (1984). Closing the sensor-computer-robot control loop Robotics Age : 15-20.
- Brooks, R.A. (1989). A robot that walks: emergent behaviours from a carefully evolved network Neural Computation 1: 253-262.
- Hirose, S.; Inoue, S. and Yoneda, K. (1990). The whisker sensor and the transmission of multiple sensor signals Advanced Robotics 4(2): 105-117.
- Ueno, N. and Kaneko, M. (1994). Dynamic active antenna: a principle of dynamic sensing IEEE International Conference on Robotics and Automation : 1784-1790.
- Kaneko, M.; Kanayama, N. and Tsuji, T (1998). Active antenna for contact sensing IEEE Transactions on Robotics and Automation 14(2): 278-291.
- Russel, R.A. (1992). Using tactile whiskers to measure surface contours International Conference on Robotics and Automation : 1295-1299.
- Jung, D. and Zelinsky, A. (1996). Whisker based mobile robot navigation International Conference on Intelligent Robots and Systems 2: 497-504.
- Wijaya, and J.A., Russel (2002). Object exploration using whisker sensors Australlasian Conference on Robotics and Automation : 180-185.
- Clements, T.N and Rahn, C.D (2006). Three-dimensional contact imaging with an actuated whisker IEEE Transactions on Robotics 22(4): 844-848.
- Muraoka, S. (2005). Environmental recognition using artificial active antenna system with quartz resonator force sensor Measurement 37: 157-165.
- Solomon, J.H. and Hartmann, M.J. (2006). Robotic whiskers used to sense features Nature 443: 525.
- Lungarella, M.; Hafner, V.V.; Pfeifer, R. and Yokoi, H (2002). Whisking: an unexplored sensory modality International Conference on Simulation of Adaptive Behaviour : 58-69.
- Kim, D and Moller, RM (2004). A Biomimetic Whisker for Texture Discrimination and Distance Estimation International Conference on the Simulation of Adaptive Behaviour : .
- Pearson, M.J.; Pipe, A.G.; Melhuish, C.; Mitchinson, B. and Prescott, T.J. (2007). Whiskerbot: A Robotic Active Touch System Modelled on the Rat Whisker Sensory system Adaptive Behavior 15(3): 223-240.
- Mitchinson, B.; Gurney, K.N.; Redgrave, P.; Pipe, A.G.; Pearson , M.J.and Prescott; T.J (2004). Empirically Inspired Simulated Electro-mechanical Model of the Rat Mystacial Follicle-sinus Complex Proceedings of Royal Society London(B) 271: 2509-2516.
- Sullivan, J.C.; Mitchinson, B.; Pearson, M.J.; Evans, M.; Lepora , N.Fand Prescott; T.J. (2012). Tactile Discrimination Using Active Whisker Sensors IEEE Sensors 12(2): 350-362.
- Sullivan, R.M.; Landers, M.S.; Flemming, J.; Young, T.A. and Polan, H.J. (2003). Characterizing the functional significance of the neonatal rat vibrissae prior to the onset of whisking Somatosensory and Motor Research 20(2): 157-162.
- Brecht, M.; Preilowski, B. and Merzenich, M.M. (1997). Functional Architecture of the Mystacial Vibrissae Behavioural Brain Research 84: 81-97.
- Grant, R.A.; Mitchinson, B.; Fox, C.W. and Prescott, T.J. (2009). Active Touch Sensing in the Rat: Anticipatory and Regulatory Control of Whisker Movements During Surface Exploration Journal of Neurophysiology 101(2): 862-874.
- Carvell, G.A. and Simons, D.J. (1990). Biometric analyses of vibrissal tactile discrimination in the rat The Journal of Neuroscience 10(8): 2638-2648.
- Mitchinson, B.; Martin, C.J.; Grant, R.A and Prescott, T.J (2007). Feedback control in active sensing: rat exploratory whisking is modulated by environmental contact Proceedings of the Royal Society, B 274: 1035-1041.
- Anderson, S.R.; Pearson, M.J.; Pipe, A.G.; Prescott, T.J; Prescott , T.J.and Dean; P. (2010). Adaptive Cancelation of Self-Generated Sensory Signals in a Whisking Robot IEEE Transactions on Robotics 99: 1-12.
- Kleinfeld, D.; Ahissar, E. and Diamond, M.E. (2006). Active sensation: insights from the rodent vibrissa sensorimotor system Current opinion in Neurobiology 16: 435-444.
- Camhi, J.M (1941). Neuroethology: nerve cells and the natural behaviour of animals Sinauer Associates Inc., .