Evryon

Project Description

/webdav/site/biorob/users/191978/public/website_images/picto_rehab_large.pngEvryon: EVolving morphologies for human-Robot sYmbiotic interactiON

2009-Feb-01 to 2012-May-31

The goal of the EVRYON project is to develop a novel approach for the design of wearable robots, where robot morphology and control are co-evolved and optimized together. The outcomes of this project are intended for a range of applications from rehabilitation to augmentation.

Using CPGs to control exoskeletons

Exoskeletons are particular rehabilitation robots that are worn like an outer shell of the body. In collaboration with the ARTS lab at Scuola Superiore Sant’Anna (Pisa, Italy), we developed an innovative control method for an elbow exoskeleton (the NEUROExos) based on adaptive Central Pattern Generators. Conceptually, the elbow moving back and forth (continuous flexion/extension) can be viewed as an oscillator. The adaptive CPG comes to synchronize with this oscillator, and is able to learn its amplitude and frequency. In turn, the adaptive CPG feeds back some torque to the user through the joint actuator, therefore providing movement assistance.

 
Subject with NEUROExos
Human subject wearing the NEUROExos and performing cyclical elbow movements with the forearm like an inverted pendulum
(courtesy ARTS lab, SSSA, Italy).
   Setup diagram

Sketchy diagram of the interaction between the elbow and the adaptive CPG.

 
Preliminary results illustrated a clear effect of the assistive strategy. Intact subjects were tested, and we recorded a significant decrease in their EMG profiles (biceps/triceps) as the level of assistance increased.
EMG profile

Co-optimization structure and control

Developing wearable exoskeletons poses many challenges in term of energy efficiency, robustness, control, etc… For that reason, we also investigate advanced methods to optimize both the morphology (position of the links, number of actuators, …) and the control of a lower-limb exoskeleton. Ideally, both should be co-optimized, such that the device would embed a sort of “structural intelligence” resulting in efficient dynamical interactions between the user, the mechanics, and the controller.
 

co-optimization example
We developed a new software package based on parallel computing in order to handle these multi-variable optimization problems. This software can execute different optimization algorithms and is based on a multi-layer architecture.
 
Software architecture