Autonomous Robots
Above: A comparison of our final proposed Particle Filter with EKF algorithm performance with other localization methods.
Below: Visual tools used for debugging and comparing the 2D Gaussian Distributions of the four localization methods.
This introductory class focused on perception and planning of mobile robots. This class was by far the most demanding class I have taken; all programming, implementation, debugging, and tuning occurred in rigorous 2 week intervals. This class taught me a lot about the importance of visualizations tools, especially for debugging.
The left figure shows me and my team's implementation of a Particle Filter used in tandem with an Extended Kalman Filter that utilizes the newest sensor data to get a more accurate estimation of its location.
Here is the car autonomously navigating to the goal while replanning in the midst of unmapped obstacles as it operates at a speed of 2m/s. You can see the Particle Filter and obstacle avoidance algorithms at work in the simulation window.
This is half of an implementation of Simultaneous Localization and Mapping (SLAM) called Correlative Scan Matching (CSM). Instead of using successive transforms and covariances between scans to solve for the map, CSM aligns the previous scan with the next scan to create the map.
Here is a link to the repository containing all of the code written for this class:
