DARPA Subterranean Challenge
I was a part of Team CoSTAR participating in the DARPA Subterranean Challenge. While my most significant work has been in risk-aware traversability, I have also worked on several other projects for this competition. Some of my contributions are given below.
We propose STEP (Stochastic Traversability Evaluation and Planning), that pushes the boundaries of the state-of-the-practice to enable safe, risk-aware, and high-speed ground traversal of unknown environments. This framework was the low-level planning stack for the ground robots of Team CoSTAR in the Final circuit of SubT.
This work was published in Robotics: Science and Systems.
Overall summary of the autonomy stack Nebula was published in Field Robotics.
Tracked Vehicles and Stair Climbing
We studied the kinematic modeling of tracked vehicles and methods to synthesize feedback control for climbing stairs. Though simple enough to implement, we show that the differential drive equations do not accurately portray vehicle motion on stairs. We use the power dissipation method to tracked robotic vehicles. This method leads to simple and efficient quasi- static models that can capture the surprisingly subtle effects of grousers on vehicle motion, as well as interaction with stairs. The tracked vehicle was a research platform that was deployed at the STIX event of SubT.
I mentored two undergraduate students, Jake Ketchum and Sasha Bodrova, on the design of an autonomous RC Car. The goal was to have a fast scouting vehicle that could deploy communication nodes. Nikhilesh Alatur and I were also responsible in bringing up the autonomy stack for deployment of the RC car in the Tunnel Circuit of SubT.