FEEDBACK FOR STABILIZATION DURING SWIMMING IN LAMPREY AND SALAMANDER ROBOTS
This project aims to study and provide a method of postural stability control for lamprey and salamander robots during swimming. The robot swimming gait is induced by a central pattern generator (CPG) distributed along the spinal cord. Up to now the robot stability was fixed using surfacing lightweight materials. This research has prospectively explored the possibilities to mimic the vestibular vertebrate system integrating acceleration sensing into the CPG network. It investigated the different limbs and tail movements relevant to produce stability and to correct mainly robot rolling and pitch tilt. Different simulations were performed in WebotsTM and then transferred to test the corrective effects of the network on the real robot. Progressive testing has provided sufficient information to discriminate efficient movements. The final implementation was tested during non-perturbed and perturbed swimming. However the last results were not significantly efficient compared to adequate controls. Nevertheless, the tests performed provided interesting information. It demonstrated that stability could be increased using the mentioned postural control method even if the new approach requires more tests and further investigations. The method was also extended to show the possibility of swim guidance in three dimensions. Thus this project has brought relevant approaches and it has opened a few new research questions and application.
This project is concerned with lamprey and salamander robots of the BioRob laboratory. These robots are good anguilliform swimmers. Their swimming gait is generated by a central pattern generator (CPG) distributed in a double chain of oscillatory centres along their spinal cord. The swimming motion and the body oscillations producing the forward motion are planar. However, in case of self-generated or external perturbations (created by forces and torques), inducing rolling, pitch tilt or yaw tilt, the swimming orientation changes and is no more corrected. The goal of the project is to find a method for postural stabilization against the different occurring perturbations. The method must have various characteristics like being adaptive to perturbation situations and inducing the use of a feedback control. The method should also be biologically inspired and based on the real animal shape and locomotion. Up to now, the orientation of the robot was maintained by keeping the robot at the surface plane with sagex pieces on girdles, thus avoiding any orientation change. Consequently the research area was very large and the orientation of the project may be progressively focused to restrain the domain to explore. From the beginning, yaw tilt was left aside to mainly try to counteract rolling and if possible pitch tilt (and yaw tilt is actually not a problem). Rolling is the most annoying perturbation because it prevents in particular tracking using the robot LEDs. This technique is required in some BioRob laboratory researches during swimming. Thus results of this investigation of a natural stabilization of the robot through the CPG will furthermore be directly relevant for further studies. Moreover, stable swimming enables also to decrease useless robot consumption of energy during tests. Depending on the results, the postural control method might also be extended to guide the robot swimming, allowing it to change its swim direction, plunge or surface as required, and thus opening new horizons for its utility.
This study was mainly prospective and has covered many aspects of postural stabilization methods. Based on a literature review of the lamprey postural control and on limbed swimmer movies, stabilization recovery models were implemented. These models had led to different results from basic stability observations to complex reactive movements generated by the CPG network. Even if the final model is not efficient, progressive tests have shown small improvements of the stability. The use of an accelerometer to detect the orientation changes mimicked the natural vestibular system. However numerous tests can still be performed. Due to the setbacks during the project, the final CPG network was not completed and requires new experiments on the real robot to be improved. A less linear implementation could be created into the CPG, weighting directly the limbs movements and uniquely the tail from the front of the robot. The visual part of the vestibular system could be integrated using light sensor on the robot head even if the oscillating movement of the head might be perturbing. Also implanting a angular rate 3-axis gyroscope complementary to the accelerometer could simplify the identification of the perturbations (this system is actually used in above knee prosthesis to control the gait and the leg posture accordingly to the remaining leg). Modification on the robot could still be done. If the ventrally deviated tail brings more stability, it is possible that adding a dorsal or a ventral longitudinal fin will increase the stability against torques and rolling (acting as a drift). Stabilizing the head compared to the body would also reduce the acceleration sensing confusion. As mentioned before, these tests should be made in a deeper pool. That would improve the results quality and allow freer movements in any orientation of the robot.
Finally, even if many approaches could still be investigated, this project reached several objectives like assaying the potential of different postural stabilization controls being adaptive and bio-inspired. It brought many interesting reflexions, it required to work in many aspects of the robotic multidisciplinary domain and asuited well as a translational area Minor Project.
Auke Jan Ijspeert, Alessandro Crespi, and Jean-Marie Cabelguen. From swimming to walking with a salamander robot driven by a spinal cord model. SCIENCE, 315:1416–1420, march 2007.
Auke Jan Ijspeert, Alessandro Crespi, and Jean-Marie Cabelguen. Supporting online material for from swimming to walking with a salamander robot driven by a spinal cord model. SCIENCE, 2007.
A.K. Kozlov, E. Aurell, and S. Grillner. Modeling postural control in the lamprey. Biological Cybernetics, 84:323–330, 2001.
T. G. Deliagina. Vestibular compensation in lampreys: Impairment and recovery of equilibrium control during locomotion. The Journal of Experimental Biology, 200:1459–1471, 1997.
T. G. Deliagina, G.N. Orlovsky, and F. Ullen. Visual input affects the response to roll in reticulospinal neurons of the lamprey. Experimental Brain Research, 95:421–428, 1993.
T. G. Deliagina and E. Pavlova. Modifications of vestibular responses of individual reticulospinal neurons in lamprey caused by unilateral labyrinthectomy. Journal of Neurophysiology, 87:1–14, January 2002.
T. G. Deliagina, P. V. Zelenin, and G.N. Orlovsky. Activity of reticulospinal neurons during locomotion in the freely behaving lamprey. Journal of Neurophysiology, pages 853–863, 2000.
A. Karayannidou, P. V. Zelenin, and T. G. Deliagina. Responses of reticulospinal neurons in the lamprey to lateral turns. Journal of Neurophysiology, 97:512–521, 2007.
M. Mori and Shigeo Hirose. Locomotion of 3d snake-like robots – shifting and rolling control of active cord mechanism acm-r3. Journal of Robotics and Mechatronics, 18(5):521–528, 2006.
G.N. Orlovsky, T.G. Deliagina, and P.Wallen. Vestibular control of swimming in lamprey. i. responses of reticulospinal neurons to roll and pitch. Experimental Brain Research, 90:479–488, 1992.
E. Pavlova. Vestibular Control of Body Orientation in Lamprey. 2004.
F. Ullen, T. G. Deliagina, and S. Grillner. Spatial orientation in the lamprey. i. control of pitch and roll. The Journal of Experimental Biology, 198:665–673, 1995.
P. W. Webb. Stability and maneuverability. Fish Physiology – Fish Biomechanics, 23(8):281–332.
H. Yamada, Makoto Mori, and Shigeo Hirose. Stabilization of the head of an undulating snake-like robot. Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems San Diego, CA, USA, pages 3566–3571, 2007.
Report and presentations