Autonomy Shifting for Autonomous Wheelchairs
Often individuals with mobility impairments would like help when controlling their assistive devices but, not too much help. This project seeks to determine when to automatically shift between different, discrete levels-of-autonomy for users driving an autonomous wheelchair. Each level-of-autonomy provides different levels of assistance to the user to assist them in navigating different obstacle courses. For this project, a study was run on 16 human subjects. The subjects could request changes in assistance while they navigated different obstacle courses. As they navigated the courses, the wheelchair measured information about the user's control command quality and the environment. These measurements, as well as the user's requests, are being used to train a machine learning model to classify when to shift between different levels of autonomous assistance.
I worked on this project while I was a graduate student at Northwestern University working as a member of the argallab. The argallab seeks to apply robotics autonomy to assistive technology to improve the lives of the differently abled. This lab resides within the Shirley Ryan AbilityLab, the nation's perimer rehabilitaion hospital. This ongoing project will be the topic of my master's thesis.
Skills and Tools Used
- Python, C++, ROS
- Design of Human Trials, IRB
- Time Series Data Engineering
- Classical Machine Learning (scikit-learn)
- Deep Neural Networks, KERAS, TensorFlow
- NSF Graduate Research Fellowship
- US DOD National Defense Science and Engineering Graduate Fellowship