| dc.description.abstract | Transfemoral (Above-knee) amputation of the leg of an individual as a result of traumatic injury or due to complications arising out of diabetes or vascular disorders is a common occurrence worldwide. Following the surgical amputation procedure, the subject is fitted with prosthetic leg to help regain mobility. Prosthetic sockets are designed to transfer the body weight to the leg during locomotion. During normal human gait, the lower limbs perform four major functions: balance, positioning, support, and power. Prosthetic legs currently available in the market are mostly passive devices that provide limited support and functionality during walking. These devices also have limited adaptability during walking or to enable a more active lifestyle. The common problems of the existing above-knee prosthesis for the unilateral amputees include asymmetry between motion of the prosthetic leg with the intact leg, reduced speed along with increased energy expenditure. Not only that, but there are also different types of forces, counter forces and errors associated with gait which was ignored in some active prosthesis designs. If these technical problems are left un-addressed, they may end up with secondary medical issues requiring further surgery. While it is desirable for the prosthetic limb to have similar or close efficiency or tracking to the intact limb, it is more important for the prosthetic leg to be able to replicate the movement of a normal human leg as much as possible. Most of the studies earlier were limited to pathological gait tests in laboratory environments using inertial sensor/motion trackers which restricted the mobility of the individuals. Recently, smarter data acquisition systems are designed to capture the human locomotion in an easier and effective way. Combination of these factors result in greater advancement of prosthetic research.
Prior research in lower-limb amputee gait has focused mostly trans-tibial (below knee) amputees as they are the highest in number. In general, available prostheses for people with lower limb amputation are primarily passive devices whose performance cannot be adjusted or optimized to meet the requirements of different users. The adverse complications of wearing poorly functioning prosthetic devices include asymmetric gait, increased metabolic energy consumption, limited blood flow, instability, sores, and joint pain. The amputees might have to undergo further joint (knee/hip) replacement procedure and that increases the chance of the increased number of trans femoral amputee in the long run. There exists a high and increasing demand for an advanced prosthetic foot that is comfortable and able to replicate the function of the biological foot. Trans-femoral amputees are the second highest and the research is more challenging as the amputees lost two of their vital joints (ankle and knee). So, to design an efficient prosthetic ankle-knee system, (including all the challenges for transtibial amputees) it is very important to consider the coupling effects of the two joints and different associated errors, or force associated with the gait like ground reaction force.
Currently available prosthetic knees are either simple mechanical hinges or sophisticated computer controlled. Development of active powered prosthetic knees (focused on the control with little emphasis) results in uncomfortable, low efficient, low energy consuming device. The inherent nonlinearities of the actuators make it difficult to control. Again, interaction forces between residual limb and the socket are dynamic in nature and are a result of gait pattern of individuals, interaction of the feet with the terrain, and the transfer of rest of the body weight during gait. These factors made the prosthetic device control and design advancement challenging for researchers.
Earlier literatures address assessing gait symmetry, movement of the healthy joints, activities of the residual muscles and the metabolic energy consumption in individuals who had undergone traditional amputation. There were research studies done showing considerable residual muscle activity in the transtibial and transfemoral amputees and minimal or random muscle activity based on the co-relation between residuum socket interface (RSI) force and EMG to the type of gait. These forces are a source of interest for researchers to investigate for better controlling. Adaptive controllers like PD, PID and combinations are used in the development of active prosthetic devices. But PID and other traditional adaptive controllers cannot handle these nonlinearities and challenges of human locomotion properly. Moreover, most of the designs do not have consistent performance over the total gait cycle or consecutive steps.
All prostheses require some sort of stability mechanism, either manual or a weight-activated locking system. The main joints made of mechanical hinges should control the flexion and extension motion to mimic human gait. For unilateral amputee, the development of Artificial optimized neural network controller is important in this regard as it can train the neurons with the input data from the intact leg and mimic similar trajectory for the residual limb to follow.
This dissertation addresses the limitations of traditional controllers in an orderly fashion by building a strong platform to develop intelligent knee-ankle prosthesis system. The following are the key steps adopted in this dissertation.
• First, a mathematical model will be developed for a leg movement during normal gait. Algorithms for gait analysis will be developed to study the gait of people with above-knee amputation in real time during work-related activities. Simulations will be done to observe the performance of the controller.
• A more reliable and realistic learning-based control strategy will be developed to adaptively compensate for the unknown, changing ankle-knee dynamics and drive the prosthetic ankle-knee joint along the desired trajectories. Different combinations of control parameters will be changed to see the performance improvement and error reduction. Comparative results will be shown for different controllers.
• Finally, a framework for experimental transfemoral amputee gait study will be proposed to collect data using force sensors and EMG sensors attached to the residual limbs and muscles during work related activities and normal gait.
It is anticipated that the learning capabilities of the control strategies will enable the prosthetic ankle-knee joints to not only replicate the movement of the healthy knee-ankle system, but also improve the stability of the gait and optimize the performance to a great extent. Learning-based control of the prosthetic ankle-knee joint algorithms used here consider the ankle-knee dynamics, foot-ground interaction, and the movement of the rest of the body to make it appropriate to be used for transfemoral unilateral amputee. The first strategy uses an artificial neural network-based controller to learn the unknown and changing dynamics of the ankle-knee joint and to track a desired ankle knee displacement profile. In the subsequent strategies, the neural dynamic programming-based controller is improvised by increasing the number of neurons and other parameters, comparative performance was shown for two joints also. Later a centralized controller is used to control both the joints. Additional PID is used and comparative analysis between controller schemes are presented to have a balanced and better control. Actual gait data (obtained from the healthy human subjects) of this dissertation is used to study the effectiveness of the controller. It will be interesting to see the performance of the adaptive neural network controller for unilateral transfemoral amputee with changes in terrain and in user requirements. It is anticipated that the strategy developed in this dissertation will help build an intelligent prosthetic system that can significantly improve the mobility and long-term health of people with lower limb amputation followed by proper rehabilitation. | en_US |
