I'm an MS student at Texas A&M University. My research focuses on design, development and testing of algorithms to enable deployment of AV's in rural settings.
Skills: C++ / Python / ROS / Tensorflow / Pytorch / Sensor Fusion/ Localization
Some of my projects:
-Road Surface Segmentation in 3D point clouds: Used DeepLabv3 along with a Range Image based representation of Lidar point clouds to segment road surfaces. Model was trained on rural road data collected around Bryan-College Station area.
-Global Pose estimation using LIDAR and Open Steet Maps: Implemented a Particle filter based localization methodology using Lidar Road Detection as measurements. The motivation was to help deal with driving scenarios where performance of positioning sensors such as GPS is inconsistent. The algorithm was able to recover pose within a search space as large as 36 sq.km with a mean error les than 2 meters after convergence.
-Object Detection in Images: Used the Tensoflow Object Detection API to detect objects in Camera images. I used an SSD architecture with a ResNet50 backbone, along with modidifications such as scale and Aspect Ratio tuning for anchor boxes to optimize performace on the Waymo Open Dataset.
-LiDAR-Camera Senosr fusion for object tracking: Fused 2D detections from camera measurements along with 3D object detections on LIDAR using an Extended Kalman Filter. Currently working on improving tracking performance as well as integration with ROS.
-CARLA Simulator: LiDAR-GPS fusion for localization: Fused detections from noisy GPS masurements and pose estimates from LIDAR scan matching using a Kalman Filter.
- 🔭 I’m currently working on a Lidar-Camera fusion based perception system
- 📫 How to reach me: [email protected]
