Active Above-Knee Prosthesis: A Guide to a Smart Prosthetic Leg

By Zlata Jelacic, Remzo Dedic and Haris Dindo

Active Above-Knee Prosthesis: A Guide to a Smart Prosthetic Leg presents original research and development results, providing a firsthand overview of idea generation and prototype production. The book gives insights into the problem of stair ascent for people with above-knee amputation and offers a solution in the form of a physical prototype of an active above-knee prosthesis with an actuated ankle. The book's authors have developed and tested a physical prototype of an active above-knee prosthesis, giving anyone who is researching and designing prosthetic devices firsthand knowledge on how to build on, and continue with, work that has already been done.

• Presents state-of-the-art technology in powered prosthetics
• Helps readers evaluate design options and create new developments
• Provides guidance on the evolution of advanced prosthetic design

• Language: English
• Publisher: Academic Press
• Release date: Jun 16, 2020
• ISBN: 9780128186848

Zlata Jelacic

Dr. Zlata Jelacic is assistant professor, Faculty of Mechanical engineering, University of Sarajevo, Bosnia and Herzegovina

Book preview

Active Above-Knee Prosthesis

A Guide to a Smart Prosthetic Leg

Zlata Jelačić

Assistant professor, Department of Mechanics and Control, Faculty of Mechanical Engineering, University of Sarajevo, Bosnia and Herzegovina

Remzo Dedić

Professor, Department of Design and Product Development, Faculty of Mechanical Engineering, Computing and Electrical Engineering, University of Mostar, Bosnia and Herzegovina

Haris Dindo

Assistant professor, Department of Engineering, University of Palermo, Italy

Table of Contents

Cover image

Title page

Copyright

Dedication

Acknowledgment

Chapter 1. The challenges of prosthetic design and control

Abstract

1.1 The problem of robot–environment interaction

1.2 The effect of motor redundancy on cooperative interactions

1.3 Overview of the current situation in rehabilitation robotics

1.4 Impact of amputation on the kinetic chain and proprioception

References

Chapter 2. Human motor system

Abstract

2.1 Human motor control

2.2 Motor redundancy and optimization

2.3 Adaptability of the human motor system

2.4 Motor memory and learning

2.5 Postural control

2.6 Biomechanical analysis of movement

References

Chapter 3. Hydraulic power and control system

Abstract

3.1 Parameter definition and design of the hydraulic linear actuator for mechanization of the above-knee prosthesis

3.2 Defining the constructive concept of a linear actuator

3.3 Defining the global hydraulic system for linear actuators

3.4 Power supply selection for hydraulic power unit

3.5 Hydraulic control of an intelligent active robotic prosthesis

3.6 Designing a mobile hydraulic power unit

3.7 Conclusions

References

Chapter 4. Prosthetic modelling and simulation

Abstract

4.1 General procedure for simulations

4.2 Modelling biologically inspired systems

4.3 Analytical model of the above-knee prosthesis

4.4 Model of hydraulic actuator for knee and ankle joints

4.5 Modelling of the DC engine

4.6 Robotic manipulator control techniques

4.7 Robust control theory based on the passivity principle

4.8 Simulation results of the dynamic model and controller

4.9 Conclusion

References

Chapter 5. Prosthetic design and prototype development

Abstract

5.1 Introduction

5.2 Research background

5.3 SmartLeg overview

5.4 Artificial foot

5.5 Prototype development

5.6 Experimental investigation into the kinematics of the above-knee prosthesis

5.7 Motion analysis and finite element analysis

5.8 Foot pressure research

References

Chapter 6. Dynamics-based action recognition for motor intention prediction

Abstract

6.1 Introduction

6.2 Related work

6.3 Wearable motion capture system

6.4 Machine learning

6.5 Action representation

6.6 Experimental results

6.7 Conclusions

Acknowledgements

References

Chapter 7. Experimental validation of the prosthetic leg

Abstract

7.1 Problem definition

7.2 Adaptive changes in motor patterns

7.3 Amputation

7.4 Testing of the hydraulic actuator

7.5 Measurements on subjects with and without amputation

7.6 Testing the first prototype with actuated knee and ankle joints

7.7 Prototype with actuated knee and ankle joints

References

Index

Appendix

Copyright

Academic Press is an imprint of Elsevier

125 London Wall, London EC2Y 5AS, United Kingdom

525 B Street, Suite 1650, San Diego, CA 92101, United States

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom

Copyright © 2020 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

ISBN: 978-0-12-818683-1

For Information on all Academic Press publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Mara Conner

Acquisitions Editor: Fiona Geraghty

Editorial Project Manager: Ana Claudia A. Garcia

Production Project Manager: Poulouse Joseph

Cover Designer: Miles Hitchen

Typeset by MPS Limited, Chennai, India

Dedication

To my mother Zinaida and my son Maximilian, with love.

Thank you for always loving and supporting me!

–Zlata Jelačić

To my mother Camila.

–Remzo Dedić

To my parents and my son Evan, for all the love and care.

–Haris Dindo

Acknowledgment

Writing a complex book like this is harder than we thought and more rewarding than we could ever have imagined. It would probably not have been possible without the help of many people and organizations. First of all, we want to thank Mr. Miso Hudjec, Dr. Adisa Vucina, Dr. Alma Ziga, Dr. Vesna Raspudic and Dr. Miljan Rupar who – throughout the past years and different projects – worked on the development of parts of the (bio)mechanical model of the prosthesis presented in this book. We would also like to thank Mr. Nihad Subasic from the Orthopedic laboratory Miracle for his help on conducting experiments with lower-limb amputees.

We would like to thank the Ministry of Education and Science of the Federation of Bosnia and Herzegovina for recognizing the importance of the project and awarding it with numerous research grants, as well as the University of Glasgow (Sports lab, Dr. Kenneth J. Hunt), University of Strathclyde (Glasgow, Dr. Norman Govan) and Carnegie Mellon University (Robotic Institute, Medical Robotic Lab, Dr Branko Jaramaz) for providing the fertile ground for early-stage prototypes of the idea. Finally, we would like to thank Dr. Liliana Lo Presti, Dr. Antonio Chella and Dr. Marco La Cascia for their help in developing the intention recognition model and Dr. Giuseppe La Tona for his help on collecting the experimental data from the sensors. We would also like to thank the Robotics Lab at the Faculty of Engineering, University of Palermo, for providing the necessary experimental equipment, as well as ST Microelectronics for providing the iNemo IMU modules used in developing the wearable motion capture system.

Chapter 1

The challenges of prosthetic design and control

Abstract

During the last decade, there has been significant interest – both in academia and in industry – in devising technologically advanced solutions for the improvement of mobility of people with a lower-limb amputation. This is partly due to the fact that the number of lower-limb amputees is constantly increasing. The majority of current prosthetic solutions are energetically passive devices, meaning that these devices can only react, while an active one can both act and react. Hence, they are unable to restore full mobility to lower-limb amputees. Many common everyday activities, such as walking up a slope or ascending and descending stairs, require the exertion of large forces and moments that passive devices cannot afford. This is due to the lack of externally powered joints that could substitute a large number of missing muscles and provide a gait with kinematics and dynamics similar to that of nonamputees. Current research and development efforts are concentrated on the introduction of externally powered joints that go beyond the variable-dampening characteristics of today’s microprocessor control by generating their own powered movements, as natural as users’ own gait patterns.

Keywords

Human–machine interface; actuated joints; prosthetic control

During the last decade, there has been significant interest – both in academia and in industry – in devising technologically advanced solutions for the improvement of mobility of people with a lower-limb amputation. This is partly due to the fact that the number of lower-limb amputees is constantly increasing.

A recent study estimated the current number of amputees to double by the year 2050, reaching a total of 3.6 million in the United States alone, with a trend of 185,000 persons undergoing an amputation of an upper or lower limb each year [1]. The main causes of lower-limb amputations are dysvascular diseases (80%) and trauma (15%). Other minor causes include congenital deformities, cancer, and war-related causes (especially due to landmines that affect mainly civilians and are present long after military conflicts are over).

The majority of current prosthetic solutions are energetically passive devices, meaning that these devices can only react, while an active one can both act and react. Hence, they are unable to restore full mobility to lower-limb amputees. Recent achievements, such as the wide adoption of microprocessor-regulated above-knee prostheses able to adaptively adjust the stance and swing phases according to particular usage and needs – and thus to decrease amputees’ efforts – are an important step towards that goal. Such ‘intelligent’ prostheses are capable of altering in real time their responses to situational needs, and they enable amputees to perform various activities almost naturally, such as walking on level and inclined ground, running, descending stairs, riding a bicycle, and even swimming and diving.

However, many common everyday activities, such as walking up a slope or ascending and descending stairs, require the exertion of large forces and moments that passive devices cannot afford. This is due to the lack of externally powered joints that could substitute a large number of missing muscles and provide a gait with kinematics and dynamics similar to that of nonamputees. Indeed, current efforts are concentrated on the introduction of externally powered joints that go beyond the variable-dampening characteristics of today’s microprocessor control by generating their own powered movements, as natural as users’ own gait patterns.

Designing prosthetic devices that can physically interact with patients for the purpose of jointly performing motor assignments or physical assistance in achieving motor goals or helping to rehabilitate presents a major challenge to prosthetic design and control. Actuated, prosthetic devices have the potential to improve the rehabilitation and motor learning of amputees through the initiation of the treatment process before it can be established through conventional methods, which increases the intensity of training, creates an enriched environment that allows simulation of the actual conditions, and performance of motor tasks that patients are otherwise unable to perform alone.

1.1 The problem of robot–environment interaction

The second half of the 20th century witnessed the rise of robotic and automated systems in industry. In the last 10 years, with the advancement of computers and the reduced costs of hardware and software, robots have found application in different fields such as agriculture, underwater systems, and recently within the home. The robots that are used in industry were initially able to perform simple tasks (e.g. so-called ‘pick and place’) due to the lack of advanced sensor capabilities. Thanks to the progress in sensor technologies, robotic systems are becoming more intelligent and environments in which they work are gradually changing from purely static to dynamic ones. Examples of such sensors are tactile sensors and sensors for measuring displacement, as well as visual systems (e.g. cameras).

As robotic systems become widespread in different domains, they are expected to execute diverse and challenging tasks. In order to successfully accomplish these, such systems should be robust and safe, and they should demonstrate a sufficient level of flexibility. A robotic or automated system is considered to be robust, if it is capable of operating under varying operating conditions without changing its initial structure.

Robots used in industrial applications should be safe in the sense that they should neither damage themselves nor the objects present in their environments. Furthermore, for applications concerning robots working close to humans, the safety of the humans should be guaranteed. The flexibility of a robotic system is its ability to be reassigned quickly and easily in the case of changing manufacturing demands. One particular class of the aforementioned systems is robotic manipulators, which are mechanisms composed of a chain of rigid bodies (i.e. links) connected by joints [2].

An important part of the manipulation tasks which include the physical interaction between the manipulator and the environment is the so-called contact tasks. In order to successfully perform these tasks, manipulators should improve the sensory capabilities (e.g. registration of the force during interaction).

1.1.1 Control of contact tasks

Following the successful applications of manipulators in tasks where physical interaction with the environment is not the main intention, such as spot-welding, spray painting and palletizing, it has become logical to start investigation of robot applications in contact tasks.

Control of contact tasks has been investigated in the last three decades [3,4], with a particular desire to enhance autonomy of manipulators operating in unstructured (or semistructured) environments. In a structured environment, configuration of objects with which the robot interacts is known precisely, unintended collisions do not occur, the ambient conditions (e.g. lighting, temperature) do not vary significantly, etc. Applications like spot-welding and spray painting can be executed using preplanned motion profiles. Robot control strategies that consider only desired motion profiles are less suitable for use in unstructured environments. This is due to the fact that the success of this strategy depends heavily on accurate modelling of the manipulator and the environment. Any modelling error or uncertainty eventually results in less accurate motion planning and consequently unexpected contact forces/moments may arise. In classical motion control, high-bandwidth servo control designs are used to increase robustness against modelling and parameter uncertainties and disturbances. However, when both the manipulator and the environment are very stiff, the contact forces/moments can reach very high values, causing damage to either or both.

The risk of damage can be reduced if the manipulator can comply with the environment, that is if it can modify its response based on the contact force/moments. Compliant behaviour can be achieved either by mechanical design, analogue/digital control or both. Grinding, polishing, deburring and mechanical assembly are examples of industrial applications that require a manipulator to be in contact with the environment most of the time. Manipulators in domestic applications (i.e. home robotics), that have attracted ever-increasing attention in recent years, are also used to execute contact tasks such as wiping surfaces and opening doors.

Contact forces can be actively controlled in two different ways, indirectly or directly [5]. Indirect schemes use motion control as an implicit means to regulate the contact forces, whereas direct schemes utilize explicit force feedback loops [3]. Indirect techniques are impedance (or admittance) control [6] and stiffness (or compliance) control (a simplified type of impedance control) [5]. The direct techniques include hybrid motion/force control [7], inner/outer motion/force control [8] and parallel position/force control [9]. Detailed modelling of the environment can be avoided if indirect schemes are used, however the position-tracking performance can deteriorate [3]. Among the direct methods, hybrid motion/force control is quite common, where its success depends on whether explicit constraint equations defining the environment geometry exist [4]. Another challenge for hybrid controllers is to establish contact with the environment in a stable way [10].

In a complete contact task there are three phases, free motion, contact motion and the transition phases. As the name suggests, the first refers to the case where the manipulator moves in spaces free of obstacles, the second is related to motion along certain surfaces, whereas the last one considers the transitions to and from free and contact motion phases which involve impact phenomena [11]. An important problem associated with the control of contact tasks is the transition phase in which the manipulator comes from free motion into contact with the environment. Successfully completing the transition phase is important to executing a complete contact task.

There are many factors that affect the robustness and performance of a manipulator in contact tasks. These are in general related to the availability of different sensory feedback data, and knowledge of models of the manipulator and the environment. Many variants of indirect and direct contact force control algorithms can be found in the control literature. Control algorithms for contact tasks can be classified as nonmodel (or model-free) based, model-based, adaptive, robust and robust-adaptive schemes. This classification is made based on several criteria. Firstly, the properties of the dynamic model of the manipulator used in the stability analysis are determined. Whether, how and to what extent such a model is used in the control design is investigated. Secondly, the mechanical properties of the end-effector used in this model are determined. This is done by checking whether the effect of compliance, be it due to a force sensor or another source such as a soft cover, is included in the model. Thirdly, they are classified into categories depending on the way they define the desired trajectories (known/distorted/modified) and/or they decompose the task space (estimated/measured online/identified online) [12]. Next, the mechanical and geometric properties of the environment the manipulator is supposed to make contact with are classified. This classification is done based on whether the environment is modelled as a compliant or (idealized) rigid one. After that, special attention is given to the rotational parameterizations. This is important for contact tasks that consider not only contact forces but also contact moments (or torques) since characterization of rotational contact parameters (e.g. stiffness) is not as straightforward as for translational ones [13]. Next, considerations include the type of measurements/estimations and whether the effect of measurement noise or estimation error is taken into account in the control design and stability analysis. The achieved stability results are classified regarding free and contact motion and transition phases of the contact task.

In the last decade, there has been increasing interest on intentionally introducing mechanical compliance in the design of manipulators for service applications. This is driven by the desire to increase safety, to damp the impact forces and to provide a better force/torque transmission to the manipulators’ joints by reducing the effects of backlash, dry friction, etc. Examples of such designs that can be found in the literature are series elastic actuators and variable stiffness/damping/impedance actuators (see Ref. [14]). These devices usually have additional internal control loops to regulate the torques delivered to the joints, or joint stiffness/damping/impedances.

1.1.2 Cooperative manipulation

Multiarm robotic systems have been a popular subject of active research in recent years [15,16]. These systems are required due to the limited payload capacity of single-arm systems in certain tasks and the need for additional equipment (e.g. fixtures) besides the single-arm manipulators. Practical examples are heavy payload transportation and fixtureless multipart assembly in industry and in space, or folding of cloths and preparing meals in the domestic domain. Cooperative manipulators can have significant advantages compared to a single robot. If multiple manipulators are used to carry a heavy or large payload, for example, the weight can be distributed among several smaller and cheaper robots and the payload can be handled more safely. Mechanical assembly, an important process in many industries (e.g. automotive), can be performed faster and flexibly. Special fixtures, whose main purpose is to support certain parts of the assembly, are often used in this process. With the help of multiple manipulators where one or more play the role of the fixture, the number of special fixtures can be reduced or ultimately their use can be eliminated completely. In many cooperative tasks, the manipulators grasp a common object and also bring it into contact with the environment. Some examples are scribing, painting, grinding, polishing, contour following and object aligning. Although in cooperative manipulation usually a commonly grasped object is considered, in Ref. [17] a distinction is made between noncoordinated (i.e. each arm performs different tasks), coordinated (i.e. each arm performs a different part of the same task) and bimanual (in the case of two manipulators) manipulation. The analysis and control of the first class is the same as in the case of individual manipulators for which there is an abundance of literature (Fig. 1.1).

Figure 1.1 The idea behind cooperative manipulation.

Cooperative manipulation tasks can be categorized, according to the grasp points, as fixed and nonfixed [17]. In the first case, it is assumed that the object is rigidly attached to the manipulators, thus the contact constraints are bilateral. In the latter case, relative motion between the object and the manipulators is possible, thus the contact constraints are unilateral. Although contact cannot be broken when considering bilateral constraints, it is still possible to model holding an object using grippers with fixed grasp points if the object has specific features (e.g. the ear of a coffee mug). In such a case, contact can be broken if the grippers are opened. Different control laws have been developed for cooperative manipulators such as master/slave, hybrid position/force, input–output and input-state linearization, impedance and passivity-based control. In master/slave control, one manipulator (master) is motion controlled and in charge of imposing the desired motion of the object, whereas the others (slaves) are force controlled and required to follow the motion imposed by the master. Problems such as the requirement for the slave(s) to be sufficiently compliant and how to assign the roles of master and slave(s) to the manipulators dynamically for certain tasks are commonly found in the literature [18]. Hybrid position/force control is one of the first nonmaster/slave control algorithms used for cooperative manipulation. It considers transforming the motion and force variables of the end effectors of the manipulators into object motion and internal/external forces, such that they can be controlled separately. The drawbacks of this controller are related to the incorrect use of orthogonality and contact compliance. Input–output and input-state linearization are model-based compensation methods which realize a decoupled linear system that can be controlled using well-known linear techniques. Using this technique, controllers have been designed in the joint space and in the operational (or task) space. In Ref. [19] a reduced order dynamical model is obtained by constraint elimination which is used to design a controller that decouples the force and motion-controlled degrees of freedom.

Impedance control is another very widely used technique for controlling cooperative manipulation tasks. Its use in cooperative manipulation can be categorized into three groups; by enforcing an impedance relationship between the grasped object and the external environment, by enforcing an impedance relationship between the grasped object and the manipulators or a combination of both. The approaches in the first category consider controlling the external forces that arise from the contact of the grasped object with the environment and usually require the knowledge of object accelerations (either by measurement or by estimation) and an object’s inertial parameters. For the second type, the emphasis is on controlling the internal forces of the grasped object instead of controlling the external forces. Only very few geometric parameters of the object are required for the operation of this type of controllers, knowledge of object inertial parameters is usually not required.

The last category considers controlling both the internal forces of the object and contact forces between the object and an external environment. The impedance controllers for regulating internal forces and external forces are combined in Refs [20,21]. Besides the previously mentioned approaches, passivity-based or adaptive or robust control algorithms for cooperative manipulators also exist in the literature.

One of the most demanding examples of rehabilitation robotics is active prosthetic devices that replace amputated extremities. The design of this type of rehabilitation robot is more complex because of their everyday use and therefore has more complex requirements in the fields of stability and security. Development of a management strategy should take into account the time of patient adjustment that is crucial in the selection of prosthetic devices and frequency its use.

1.1.3 The interaction problem in rehabilitation

However, in order for actuated prosthetic devices to satisfy such roles in rehabilitation, the interaction between robots and patients should be intuitive and natural. Many