Dzeladini, F.

Locomotion prediction of bipedal walking in monkeys



As a consequence of the development of locomotion rehabilitation and the great interest

of extracting information directly from brain activity, a lot of work have be done regard-

ing the prediction of locomotion from brain records (see pionner wark of N.Fitzsimmons).

Currently, accurate real-time prediction of joints in a Cartesian coordinate system

are possible. Such predictions are expected to have possible applications, such as assist-

ing old people walking or, in a larger horizon, opening the way to the development of

efficient exoskeleton. Movement prediction is crucial for the development of tools that

could assists disabled, since it will allow anticipation of the movements and thus permit

more fluidity in the movements.


In this project, we started developing a new way of representing the movement dur-

ing locomotion, based on joint angle instead of expressing joints in a Cartesian coordi-

nate system. The method uses a specific oscillator, developed by J. Kieboom, capable

of generating an arbitrary waveform to dynamically reconstruct the joint angles. On the

one hand, this will permit to reduce amount of information needed to accurately predict

the joint position and on the other hand, by the intrinsic use of differential equations,

this will offers stability regarding perturbations, and could easily be extended to take

into account crucial information, such as feedback. Howeve, the method described has

certain drawbacks that will be extensively discussed in chapter 4.


The project is separated in two parts. In the first part, we will present the method and

apply it to predict the locomotion of a monkey walking on a treadmill. The parameters

used for the reconstruction will be directly extracted from the walking, and not from

brain records. In the second part, we will analyze the relationships between the different

extracted parameters, and assess their relative importance in locomotion modulation.