Engineering Co-op: Amazon Robotics Spring 2024

Worked on joint space and task space trajectory generation and motion planning for high DOF serial link robotic arms. More specifically, I created a MATLAB tool that allows users to import any serial link robot configuration (URDF), control the robot, visualize it, set waypoints, suggest JS waypoints, find singularities, and also rapidly generate smooth time-optimal trajectories in either joint space or task space with or without obstacles. My method rapidly computes these trajectories through strategic interpolation, inverse kinematics initialization, and intelligently dealing with joint acceleration and angle limits.

Co-authored a paper published to the 2024 Amazon Machine Learning Conference (under NDA), detailing my successful research. This tool has already proved essential to the team particularly with testing new robots and experimenting with different configurations.

As part of this project, I researched and implemented a custom bi-directional RRT-connect algorithm for serial link arms, iteratively improving its speed and accuracy while displaying my ability to read, synthesize and utilize academic robotics papers.

Also worked on mechanical design for a novel robot, leading to successful demonstration of the project and its prioritization within Amazon. In particular I worked on heatsinks, electronics enclosures, links, and did some work experimenting with on varying types of robot mechanical configurations.