Performance Improvement for Particle Filter SLAM

In practice, it is sometimes difficult to well-model the motion and the observation because of some unpredictable disturbances. Particle filter is a sampling method to approximate these two models. However, the number of particles is a critical factor that influences the speed and the accuracy of particle filter SLAM. In this project, with the dataset generated with the THOR robot, I implemented the particle filter SLAM and improved its performance. Some modifications are proposed. Link

Scene 0	Scene 1	Scene 2