A running collection of robot experiments I’ve run, each with a video and a short description.
Search-Then-Commit: Multitask Belief Space Planning Applied to Robotic Manipulation
A successful execution of our belief space planning algorithm on our challenge task. More info on the GitHub page.
MIT ACL Outdoor AGV Exploration Mission
Demo of outdoor, GPS-denied, mesh comms, decentralized exploration and mapping mission.
Multi-Agent SLAM and Planning
Multiple robots simultaneously localizing, building a shared map, and planning their motion through the environment.
3D Printed Arm & Table Cleanup Policy
An arm I built from 3D printed components and trained to clean up pins on my apartment table.
6 UGV Swap with Obstacle Avoidance
Six UGVs swapping positions while avoiding obstacles, using WarpWeaver, the decentralized real-time multi-agent planner from a paper I co-authored, Dispersive Tree Agreements for Multi-Robot Coordination.
10 UGV Swap with Deconfliction
Ten UGVs swapping positions while deconflicting their paths with each other, also using WarpWeaver.
Fun Obstacle Course Teleoperation of UGV I Built
Teleoperating a UGV I built through an obstacle course, just for fun.
Drone-Based Infrastructure Inspection Demo
A demo of using a drone to inspect infrastructure.
UGV Dynamic Obstacle Avoidance with MIGHTY and Custom Mapping
This is a UGV I built demonstrating dynamic obstacle avoidance using MIGHTY, a planning algorithm developed in our lab, and a custom mapper I created. My mapper combines ground elevation mapping to assess terrain traversability with 3D voxel mapping to capture obstacles above the ground, including overhangs and moving objects. The fused map helps the planner navigate traversable terrain while avoiding obstacles.
Planner paper: “MIGHTY: Hermite Spline-based Efficient Trajectory Planning” by Kota Kondo et al.