The CoMMA Lab at ICRA!
The CoMMA Lab (which included Zak, Alexiy, Pranav, and Akshaya) were at ICRA 2025!
Papers Presented
Three papers were presented in the main conference:
“Nearest-Neighbourless Asymptotically Optimal Motion Planning with Fully Connected Informed Trees (FCIT*)” with Tyler Wilson, Jonathan Gammell, Wil Thomason, and Lydia E. Kavraki
This work opens new doors in almost-surely asymptotically optimal (ASAO) planning and shows with SIMD-accelerated edge validation and no nearest neighbors datastructure we can achieve real-time ASAO planning for high-dimensional manipulators!
“CaStL: Constraints as Specifications through LLM Translation for Long-Horizon Task and Motion Planning” with Weihang Guo and Lydia E. Kavraki
This work presents an LLM-based natural language-to-specification framework that goes beyond goal reachability and can specify constraints such as action ordering, blocking, and more! We show using constraints is more effective than direct PDDL translation or task decomposition.
“Constrained Nonlinear Kaczmarz Projection on Intersections of Manifolds for Coordinated Multi-Robot Mobile Manipulation” with Akshaya Agrawal, Parker Mayer, and Geoffrey Hollinger
This work presents a new method for finding valid configurations of multi-robot teams under complex non-linear constraints (e.g., as in collaborative transport) - handling systems where Newton-Raphson fails!
Workshops
Zak helped organize the workshop “RoboARCH: Robotics Acceleration with Computing Hardware and Systems”, which brought together computer architects, roboticists, and software engineers to discuss advances in hardware acceleration for robotics.
The lab also presented three papers at RoboARCH and one at AQ²UASIM:
“AORRTC: Finding Optimal Paths with AO-x and RRT-Connect” with Tyler Wilson, Wil Thomason, Zachary Kingston and Jonathan Gammell.
This work presents AORRTC, an almost-surely asymptotically-optimal variant of RRT-Connect building on the AO-X meta-planner. AORRTC reaches the empirical optimum solution in milliseconds and finds initial solutions as fast as RRT-Connect.
“Faster Behavior Cloning with Hardware-Accelerated Motion Planning” with Alexiy Buynitsky and Zachary Kingston.
This work presents some initial results on using VAMP to accelerate imitation learning methods, demonstrating significantly faster data generation and training than prior.
“pRRTC: GPU-Parallel RRT-Connect for Fast, Consistent, and Low-Cost Motion Planning” with Chih H. Huang, Pranav Jadhav, Brian Plancher and Zachary Kingston.
This work presents pRRTC, a GPU-parallel implementation of RRT-Connect that achieves the best performance on the hardest problems compared to other CPU/GPU planners.
“Underwater Multi-Robot Simulation and Motion Planning in Angler” with Akshaya Agrawal, Evan Palmer, Zachary Kingston, and Geoffrey A. Hollinger.
This work presents an extension to the underwater simulation framework Angler for multi-robot simulation and motion planning.
Zak also gave a talk at the workshop on Language and Semantics of Task and Motion Planning and was on a panel at the end of the day. You can watch the talk on YouTube.
