Pose Estimator
Particle filter pose estimation with C++ and CUDA
Particle Filter Simulation
Overview
Inference-oriented pose estimation built around a particle filter with 3D object detection of landmarks. The C++ implementation is the primary runtime CMake manages the build. Deployed on the 2022 RAPID REACT robot to compute 3D shot solutions to the hub from real-time sensor input.
My contribution
Designed and implemented the particle filter from scratch in C++, wrote the CUDA kernels for likelihood evaluation, and integrated the system on the robot. The pose output fed directly into the drive controller for autonomous shot calculations.