Stiffness characterisation of a soft exoskeleton system for fall recovery

• Main Author: Eng, Gerald
• Other Authors: Ang Wei Tech
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Mechanical engineering
• Online Access: http://hdl.handle.net/10356/75337
• Institution: Nanyang Technological University

The amazing mobility that human beings have is due to the complex biological systems that allow for bipedalism. However, there is a group of people who are exceptionally prone to falling due to impairment in their sensory, effector and processing systems. A large proportion of this group of people would comprise of the elderly. A solution to help this growing group of people is to design a soft exoskeleton system that will prevent falls by exerting a counteracting force on the user when a fall is detected. With a fall initiating and ending in about 500ms, an actuator needs to be high force, high bandwidth and high stroke. In addition, the actuator can only direct assistive force via the subject’s body and the exosuit. Since the human body is highly compliant, it affects the efficiency of the force transfer significantly. To improve the efficiency and repeatability of the system, there needs to be a reliable method to estimate the required actuator force and the exosuit tightness. An experimental procedure was developed to study the relationship between the force and displacement experienced by the exosuit. Different characteristics of the subject’s body (eg: BMI, body fat) were also considered when interpreting the relationship. 4 methods of analysis were used: • The first method models the exosuit system with a linear stiffness spring • The second method models the exosuit system with a non-linear stiffness spring (2nd order polynomial) • The third method models the exosuit system with a variable stiffness spring, taking into account the different stiffness of the bone and muscle structure in the human body. • The fourth method forgoes the spring model and uses a trained neural network to predict the force and displacement of the exosuit Amongst the first three methods, all of which model the exosuit system as a spring system, the second method which used a 2nd order polynomial fitted the experimental data of 10 subjects the best. The R2-value was 0.880 as compared to 0.869 and 0.814 for the first and third method respectively. Hence the 2nd order polynomial model was used for further comparison with the fourth method. The fourth method, which used the experimental results of an additional 8 subjects to train the neural network, had a mean output force error of 12.7N for the data set used for training. As a test, the network was used to predict the force requirement for 2 new subjects and had a mean output force error of 16.0N. The neural network was superior to the 2nd order polynomial model, which resulted in a higher error of 23.0N for the same test subjects. With a larger sample size, the network will be able predict the measurements with more accuracy for a population. These predictions could then be used to design an actuator for the exoskeleton system which matches the force-displacement characteristics of a subject with decent reliability.