Event Location: GHC 4405

Abstract: The neural controls of human and animal locomotion have been studied over centuries. However, much of our knowledge about the locomotion control of complex species, especially humans, still relies on extrapolating from what is known in simpler animals. One barrier for better understanding the control of human locomotion is that we cannot directly measure the overall control structure and the role of individual neural pathways at the behavioral level through current experimental techniques. As an alternative, researchers resort to simulation studies to test control hypotheses. By testing a hypothesized controller in a physically and physiologically appropriate simulation environment, one can evaluate the plausibility of the controller and propose experimental studies which can further investigate the hypothesis.

This proposal investigates how much of human locomotion control can be explained by spinal reflexes using neuromuscular physics simulation. It is known that the spinal cord control is essential in generating locomotion behaviors in humans and other vertebrates. However, the functional contributions of the individual underlying neural mechanisms, reflexes and central pattern generators, are not fully understood. This proposal presents a reflex-based neural control model that can generate diverse human locomotion behaviors. The control is based on neurophysiologically plausible muscle-reflex pathways realizing limb functions essential to legged systems. With different sets of control parameters, the model can generate walking and running, acceleration and deceleration, slope and stair negotiation, turning, and deliberate obstacle avoidance. The plausibility of the control model is also discussed.

The latter part of the proposal provides some exemplary studies of using the neuromuscular human locomotion model. In the first study, the neuromuscular model serves as a simulation test-bed to explore the functionality of the compliance of the feet in locomotion. The simulation model suggests that the compliant feet incurs more energetic costs than stiff feet do in walking, which agrees with preliminary experimental results. The second study transforms the neuromuscular model into a robotic controller for the bipedal robot ATRIAS. The robotic controller, or the virtual neuromuscular controller, emulates the neuromuscular model online to calculate the desired joint torque for robust walking. With the controller, ATRIAS can walk on a rough terrain with ground height changes between +-20 cm in a 2D simulation environment.

In the future work, I will complete the hierarchical structure of the neural controller, which is currently limited to the swing leg control. With the full hierarchical structure, the controller is expected to generate diverse human locomotion behaviors and transitions between the behaviors by changing a small set of high-level control parameters. In addition, I plan to enrich the exemplary studies of using the neuromuscular model, including applying the virtual neuromuscular controller to the ATRIAS hardware.

Committee:Hartmut Geyer, Chair

Christopher G. Atkeson

Stelian Coros

Auke J. Ijspeert, EPFL