Back Extension Exoskeleton
Overview of the project
As an individual Final Year Project, I built a back extension exoskeleton under the supervision of Professor Ling Shi. The project focus on developing an affordable exoskeleton system design to effectively provide assistance for elders and workers in logistics and material handling industry, thus reducing back stress and reduce back pain.
Currently the exoskeleton prototype was developed up to the second generation, which was capable of assisting the user in weight-lifting and stair-climbing tasks. However, the quantitative measurements of assistive torque, improvements in muscle stress,etc, are not yet performed due to time and resource limitations.
A brief demonstration video is here for your reference:
Below Sections showcase the development process of the exoskeleton prototype
First Generation
The first generation structure of the exoskeleton is finished and successful structural functionality tests finished. The test video is displayed here.
The video demonstrated the ability of the user to perform tasks such as walking, squatting, bending and going over stairs. It showed that the exoskeleton is able to move with the human body in working conditions without creating interference with the user.
Test with actuation - Passive Actuator
We also tested with an elastic passive actuator that can exert ~90 kgf during full stretch.
Two sandbags weighing a total of ~15 kg was lifted multiple times with different distance to the standing point to test the effect of the exoskeleton in different moment arm settings.
Comparing with the control group without the exoskeleton, the effects are not significant with short moment arms, but the stress (not quantified yet as the test is still at an very early stage) on the lower back region is significantly reduced with assistance from the exoskeleton during lifting with longer moment arms.
The system structure is planned as below:
The User bends his/her neck backward to control the exoskeleton, further the bend, larger the assistive torque.
The control signal is converted to required torque, which is send to the motor torque controller after being corrected by the geometry and pose of the exoskeleton.
Below are some detailed pictures of the first prototype
Front View
Right View
Back View
Left View
Front View
Side View
Back View
Second Generation
With lessons learned from the first generation, the second generation was improved in multiple areas such as weight, comfort, and power.
the entire frame was redesigned and lighter materials was used to reduce the overall weight of the exoskeleton to about 14 kg, half of the weight of the first generation.
Load simulations on critical structural components were performed in SolidWorks to make sure the new frame design is strong enough.
Frame modifications
Apart from lighter structure, the frame was also modified in the second generation design to be more comfortable. The structure near the shoulder area was changed to include a cutout where the user's shoulders can fit in and no longer have to be pushed against the vertical beam in the first generation.
A view of the new frame design is shown here:
Inclusion of harnesses
To further improve comfort and usability of the exoskeleton, medical harnesses were installed for all of the interaction points of the exoskeleton. This allows the user to comfortably fit in the exoskeleton and also evenly distributes the assistive force to reduce the pressure on the user's legs.
These harnesses were originally used for hip joint and neck fixtures after injury, they were repurposed and modified then installed on the corresponding sections of the exoskeleton.
Actuation Upgrade
Comparing with the first generation which is only passively actuated with a rubber band, the second generation have an improved actuation system consisting of 3 IMU sensors, 2 BLDC motors and a central control unit in the STM32 family, it is the first "smart" version of the exoskeleton.
Based on the direction of gravitational acceleration, the sensors read out tilting angles(through a built-in KF) of the trunk and each leg of the exoskeleton, which were then converted to relative angles and angular speeds to determine the motion intention of the user. After compensating for the geometry of the exoskeleton, a desired motor torque will be calculated and sent to the ESC via CAN for each leg.
Below are some other pictures of the second prototype
Exoskeleton Front View
Exoskeleton Side View
User Front View
User Side View
The second generation prototype already have the essential features and functions, however there is still a huge room for improvement, in both hardware and software aspect.
Future plans with the project includes:
1. Quantitatively model the exoskeleton system, then develop control algorithms and strategies to improve accuracy and responsiveness of the assistive torque
2. Develop control algorithms to improve the "transparency" of the exoskeleton in IDLE mode.
3. Implement software filters to process tension sensor data and then integrate the tension data into closed-loop control of output assistive torque.
