Kuri
My Contribution Highlights
- Implemented person tracker using a custom systematic importance resampling particle filter to fuse point clouds from depth sensor with visual bounding boxes from MobileNet object detector
- Implemented person follower using a PID controller and a feedback linearization approach
- Integrated person following feature into robot with voice command
A demonstration of the person following feauture I implemented with an integrated voice command.
A visualization of the person tracker in rviz. Blue points near the tracked_person coordinate frame are particles from the particle filter; other points are 3D points from the depth sensor.
Block diagram of the perception for the person follower. I implemented the Person Tracker section.
Visualization of the robot telemetry in rviz. Blue points near the tracked_person coordinate frame are particles from the particle filter; other points are 3D points from the depth sensor.
Project Info
- Robot Type: Two-wheeled Mobile Indoor Robot
- Application: Personal Home Robot
- Organization: Mayfield Robotics
- Personal Role: Computer Vision Intern, June 2018 – August 2018