The original goal of this project was to use a robotic arm to provide assistance during rehabilitation tasks, and in order to achieve this goal, we had to develop several models for predicting the intention of the participant, implement planning algorithm along the intention of motion. Besides that, we also had to figure out some technical details, such as how to transform the information obtained from the camera to the robot, how to properly measure the EMG activation, and how to make all the components communicate with teach other efficiently.
After two years of work in this project, it was time to perform a validation study, testing not only each individial component, but also how everything works together. These experiments, even though delayed during to the CoVID-19 pandemic, took eventually place at the Mechanical Engineering department of UTCN, with the kind help of Paul Tucan.
We performed measurements with 11 volunteers, using the following experimental protocol:
- Attach electrodes on the appropriate muscles
- Perform motion without the assistance of the robot
- Train the neural network model of intention prediction
- Perform the same motion with the assistance of the robot
The validation for the success of the experiments was the quantification of the interaction force between the volunteer and the robot. This force can help us quantify the amount of assistance the robot is providing to the volunteer, which is the goal of this implementation. The experimental setup, methodology, and results were submitted for publication at the IEEE Transactions on Biomedical Engineering.
Below, you can see a sample of the measurements. In the video you see in parallel a real view of the experiment (volunteer being assisted by the robot), and a simulated view as captured by the depth camera and the robot. The interaction force between the robot and the person is also presented.
