2019 · Techmation · Harvey Mudd capstone
A senior capstone sponsored by Ningbo Techmation. We built an mmWave radar system that detects and tracks obstacles in real time for indoor autonomous guided vehicles (AGVs).
Dust, glare, low light, and reflective surfaces make cameras and lidar less reliable on a factory floor. Our sponsor, Ningbo Techmation, wanted to know whether a short-range mmWave radar could be the only perception sensor on an indoor AGV, and still let it navigate and avoid both static and moving obstacles.
Since radar was the only sensor, tracking had to be accurate and run in real time on a small embedded computer.
I built a multi-object detection and tracking pipeline that ran on an NVIDIA Jetson TX2. It clustered radar returns with DBSCAN, used gating and data association to match objects between frames, and used an Extended Kalman Filter to estimate their motion.
I tested TI’s IWR1642 and IWR1443 radars against objects of different sizes and radar cross-sections, across different settings and indoor spaces, to see what the sensor could and couldn’t detect.
I wrote a Python tool to parse the radar’s data packets as they streamed in. The rest of the pipeline was built on top of it.
We integrated the system on an indoor robot with ROS, using the radar, an occupancy grid, and a cost map to navigate around static and moving obstacles.
The robot navigated an indoor course using radar as its only sensor. We measured the tracking accuracy and latency. I led the team in the spring semester, and we presented the work at the sponsor’s headquarters in Shanghai.