Hierarchical Controller

Dinh Binh Khanh

Figure 1: sensors and hierarchical controller for the elbow exosuit. A load cell, monitoring the tension in the suit's tendons, and a flex sensors, recording the joint angle, are used to implement an adaptive hierarchical controller that captures the subjects' intention to move (high-level), compensates for backlash in the transmission (mid-level) and accounts for friction in the Bowden cables (low-level).

While the low profi le, lightweight and compliance of soft wearable devices make them very appealing solutions for assisting human movements on a daily basis, their intrinsic soft nature poses unquestionable control challenges: deformation of the stretchable materials, friction in the Bowden cables and the viscoelastic properties of human soft tissues make a simple feedback control inadequate for achieving a reasonable tracking accuracy. Moreover, understanding the the intentions of the wearer is a key but challenging task.

We propose a hierarchical cascade controller, schematised in Figure 1.b, to achieve accurate and intuitive control of our soft exosuit using only a load cell on the tendons and a flex sensor to measure the elbow joint's position (Figure 1.a). The proposed framework does so using a three-layered architecture: the highest layer decodes the user's intention and tunes the level of assistance accordingly; the middle layer compensated for backlash phenomena deriving from shifts of the fabric and Bowden cables upon the application of forces; the lowest layer accounts for friction in the transmission.

Using the exosuit to lift a 1kg load with the proposed control paradigm (Video 1) results in a 50% reduction in muscular effort of the biceps brachii when compared to a no-suit baseline, while accurately tracking the user's desired trajectory.  

Video 1: testing the elbow exosuit with the adaptive hierarchical controller. Subjects were asked to freely perform flexion/extension movements of the elbow while holding a 1kg weight in their hand. We used Electromyography on their biceps brachii to monitor their muscular effort and a flex sensor, sewn in the fabric, to track their joint's angle. 

Publications

Binh Khanh Dinh, Michele Xiloyannis, Chris Wilson Antuvan, Leonardo Cappello, and Lorenzo Masia, "Hierarchical Cascade Controller for Assistance Modulation in a Soft Wearable Arm Exoskeleton", Robotics and Automation Letters (RAL), 2017.

Binh Khanh Dinh, Michele Xiloyannis, Chris Wilson Antuvan, Leonardo Cappello, Shih-Cheng Yen and Lorenzo Masia, "Adaptive Backlash Compensation in Upper Limb Soft Wearable Exoskeletons", Robotics and Autonomous Systems (RAS), 2017.

Michele Xiloyannis, Khanh D. Binh, Leonardo Cappello, Chris W. Antuvan and Lorenzo Masia, "​Design guidelines for upper limb soft wearable exosuits", Wearable Technology for Medicine and Healthcare, Elsevier, in press.

Lorenzo Masia, Irfan Hussain, Michele Xiloyannis, Claudio, Pacchierotti, Leonardo Cappello, Monica Malvezzi, Giovanni Spagnoletti, Chris Wilson Antuvan, Dinh Binh Khanh, Maria Pozzi and Domenico Prattichizzo, "Soft Wearable Assistive Robotics: Exosuits and Supernumerary Limbs", Springer, in press.

Binh Khanh Dinh, Michele Xiloyannis, Chris Wilson Antuvan, Leonardo Cappello, and Lorenzo Masia, "Hierarchical Cascade Controller for Assistance Modulation in a Soft Wearable Arm Exoskeleton", International Conference on Robotics and Automation (ICRA), Singapore, 2017.

Binh Khanh Dinh, Leonardo Cappello, Michele Xiloyannis and Lorenzo Masia, "Position Control using Adaptive Backlash Compensation for Bowden Cable Transmission in Soft Wearable Exoskeleton",  IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Daejeon, South Korea,  2016.