The eye movement system of animals is an exemplary sensorimotor system, with a wealth of neurophysiological data. Many descriptive models have been developed to explain this data. To better understand the functional significance of the control architecture used in animals, we developed the Gazebot, a fast eye-head robot with a complete computer vision system and inertial sensing that approximates the vestibular system. The eye includes a stiff camera cable, similar to the optic nerve, which adds a complicated non-linear stiffness to the plant.
Despite these real-world complexities, we showed that the parameters of a non-linear controller inspired by Robinson’s model of the behavior of ocular motor neurons can be learned, and produces fast and accurate saccadic eye movements. We also showed that the retinal image can be effectively stabilized using the inertial sensor and vision, without a joint position sensor (as in animal eyes). The algorithm uses an adaptive controller that models the vertebrate Cerebellum for velocity stabilization, with additional drift correction. Our system can reduce camera image motion to about one pixel per frame on average even when the platform is rotated at 200 degrees per second. This work earned M.Sc. student Martin Lesmana the 2012 Master’s Thesis Award from CAIAC (the Canadian AI Association). Subsequently, we developed a general computational model of neural control of eye movements that proposes that the smooth pursuit system is an integrated part of the gaze stabilization system and utilizes the phylogenetically older mechanisms related to the vestibular ocular reflex (VOR).
References
- D. K. Pai, “Smooth Pursuit and Gaze Stabilization: an integrated computational model,” Vision Sciences Society Annual Meeting, May 13-18, 2016. Abstract in Journal of Vision, September 2016, Vol.16, 1344. [DOI]
- M. Lesmana, A. Landgren, P.-E. Forssén, and D. K. Pai, “Active Gaze Stabilization,” Indian Conference on Computer Vision, Graphics and Image Processing (ICVGIP ’14) , December 14-17 2014. Best Paper Award. [DOI]
- M. Lesmana and D. K. Pai, “A Biologically Inspired Controller For Fast Eye Movements,” in Proceedings of the 2011 IEEE International Conference on Robotics and Automation, Shanghai, China, May 9-13, 2011. pp. 3670–3675. [DOI]