Background

Research interest in soft robotic manipulators comes from the potential capability of soft manipulators to perform complex tasks in an unstructured environment and safely around humans. Bio-inspiration is often provided by soft-bodied creatures, such as octopuses, that have evolved to solve complex motion control problems like reaching, grasping, fetching, crawling, or swimming. The exceptional coordination abilities of these marine animals have naturally motivated efforts to gain a deeper understanding of the biophysical principles underlying their distributed neuromuscular control.

Research description

We have an ongoing inter-disciplinary project together with biologists and roboticists to better understand the sensorimotor control of an octopus arm.  We refer to the simulation platform as Cyberoctopus.  

Modeling.  The dynamics of a single octopus arm are modeled using the Cosserat theory of elastic rods.  In contrast to typical rigid link models of classical robotics, Cosserat rod models capture, through linear and angular momentum balances, the (one-dimensional) continuum and distributed nature of elastic slender bodies deforming in space. These models account for all modes of deformation – bend, twist, stretch, shear – induced by external and internal forces and couples.

Control.  The control problem is to activate the internal muscles to solve motion control tasks, e.g. reaching and grasping.  In a path-breaking paper, our group provided the first such demonstration of the energy shaping control methodology to solve this control problem (see videos below).  Shading indicates activation of internal muscles (see webpage for additional details on control-oriented modeling of internal muscles).

This proposed procedure provides three main features. First, it yields a simple yet stable feedback control law that is integratable and implementable with realistic simulation and physical platform. Second, the modified potential energy and the controlled Hamiltonian contains meaningful physical interpretation of the stress-strain relationship. Lastly, it offers a benchmark for other more sophisticated forms of control methodology, including optimal control, reinforcement learning and/or dynamic neural network control. 

Presentations

Slides from the prelim exam of Heng-Sheng Chang for the Ph.D. program, University of Illinois at Urbana Champaign, 2023

Slides from the talk to Department of Power Mechanical Engineering, National Tsing Hua University, Taiwan, January 5, 2023 (with recording in Mandarin

Slides from the Coordinate Science Laboratory Student Conference, University of Illinois at Urbana Champaign, U.S., February 28, 2020

Publications

Chang, H. S., Halder, U., Shih, C. H., Naughton, N., Gazzola, M., and Mehta, P. G., 2023, February. Energy-shaping control of a muscular octopus arm moving in three dimensions. Proceedings of the Royal Society A, 479(2270), 20220593.

Chang, H.S., Halder, U., Tekinalp, A., Gribkova, E., Gazzola, M. and Mehta, P.G., 2021, December. Controlling a CyberOctopus soft arm with muscle-like actuation. In 2021 60th IEEE Conference on Decision and Control (CDC). IEEE.

Chang, H.S., Halder, U., Shih, C.H., Tekinalp, A., Parthasarathy, T., Gribkova, E., Chowdhary, G., Gillette, R., Gazzola, M. and Mehta, P.G., 2020, December. Energy shaping control of a cyberoctopus soft arm. In 2020 59th IEEE Conference on Decision and Control (CDC) (pp. 3913-3920). IEEE.

Acknowledgements

Financial support from the ONR MURI N00014-19-1-2373, NSF/USDA #2019-67021-28989, and NSF EFRI C3 SoRo #1830881 is gratefully acknowledged.