Berkeley, California, United States
Project Description:
(1) Developed a robust architecture for vision based localization, consisting of camera calibration, object detection, image retrieval and state estimation using a particle filter (PF)
(2) Successfully demonstrated the end-to-end implementation on the testing vehicle in a controlled environment
Responsibilities:
• State Estimation
* Developed and implemented an optimized PF algorithm in real-time using ROS 2 Python, and conducted offline simulation testing to evaluate the algorithm's performance
* Employed a data-driven approach and incorporated Gaussian mixture models (GMMs) to model the measurement noise within the PF algorithm. Demonstrated a substantial average improvement of 60% in position accuracy, reducing the overall error from 2.78 m to 0.99 m.
