the eye robot project

Biological Inspiration

Mobile agents, whether natural or artificial, are all faced with the problem of disturbances in respect to their vision system while moving. This phenomenon is easily observed when looking at, typically shaky, video footage shot with a hand-held camera whilst walking. By decoupling the vision system, the eye, from the main body or head, nature has come up with an efficient way of tackling the crucial task of gaze stabilisation.

In many mammals the eyes are rotated in the opposite direction of the head movement in order to stabilise the image on the retina. This behaviour is referred to as the vestibulo-ocular reflex (VOR). As its name implies, it is not driven by the vision but the vestibular system, the balance organ in the inner ear. The head rotation is sensed by the semicircular canals, then relayed to the vestibular nucleus in the brainstem. The brainstem neurons innervate the motor-neurons of the extraocular muscles which rotate the eyeball. Due to the large latency in the vision signal (approximately 100ms) a closed-loop control based on the visual slip signal is not possible. However, the vision signal is used in nature to act as a training signal to fine tune this reflex.

It is well understood in the neuro-science community that it is the cerebellum (latin, little brain) that is required not only for the tuning of the VOR, but all motor-functions. The cerebellum is a very regular structure and each microzone's functionality is largely defined by it's inputs and outputs. We believe that by investigating the role of the cerebellum in VOR motor learning in a robotic environment, we are able adopt some algorithmic solutions from nature in order to build more flexible, robust and adaptive robots. On the other hand, we can provide neuroscientists with test-bed to assess their models in a 'real world' setting.

Engineering Implementation

To sense rotations in the robot setting we employ MEMS gyroscopes from Analog Devices&trade, the ADXRS300. We arranged three of them in a mutually orthogonal setup in order to sense rotations around the yaw, pitch and roll axis.

During the course of the project we experiment with different actuation mechanism for the artificial oculomotor system. Initial tests of the algorithms are carried out on a electrically actuated system, which we force (using digital filter models) to behave 'muscle like'. In a second set of experiments we are migrating to a oculomotor system which is driven by pneumatically actuated 'artificial muscles'. To test the effectiveness and stability of our algorithms we developed an electrically actuated 3D platform to reproduce typical rotational velocity profiles the eye in a human head experiences. This platform is not electronically connected with the eye mechanism and the sensory coupling is only achieved via the gyroscopic system mounted on the platform. This approach mimics the sensing of head rotations via the semicircular canals of the inner ear. We use a simple vision processing algorithm to track the position of a laser projection in order to generate an teaching signal for the cerebellar inspired adaptive filter structure. If our model of brainstem and cerebellum do not generate an appropriate driving signal for the mechatronic eye, visual slip (i.e. an apparent motion of the visual target) will occur. To be clear, this vision signal is not used as a feedback signal to drive the eye, but purely as a teaching signal to fine tune the feed forward structure of brainstem and cerebellar circuitry. As long as the visual slip is correlated with the vestibular sensation this fine tuning contributes to a more and more stable vision signal, which is crucial for 'higher order' image processing task like edge detection, object recognition and trajectory detection.

Video Gallery