Conformal predictive safety filters that: 

1.) Predict the other agents’ trajectories

2.) Use statistical techniques to provide uncertainty intervals around these predictions

3.) Learn an additional safety filter that closely follows the RL controller but avoids the uncertainty intervals

We use conformal prediction to learn uncertainty-informed predictive safety filters, which make no assumptions about the agents’ distribution. 

K. J. Strawn, N. Ayanian, and L. Lindemann. "Conformal Predictive Safety Filter for RL Controllers in Dynamic Environments" in IEEE Robotics and Automation Letters, October 2023. [IEEE, arxiv]

We track the rewards the agent sees during training to automatically avoid symmetric actions that lead to suboptimal performance in zero-shot cooperative scenarios.

K. J. Strawn and N. Ayanian. "Symmetry Agnostic Learning for Multi-Robot Zero-Shot Coordination", in AAMAS Workshop on Autonomous and Multi-Robot Systems, May 2022. [PDF Preprint]

A multi-robot consensus system with Byzantine fault-tolerance, building upon blockchain for a MAPF problem with adversarial or unknown agents that is competitive against current distributed task and path planning algorithms.

K. Strawn and N. Ayanian. "Byzantine Fault Tolerant Consensus for Multi-Robot Pickup and Delivery ", in International Symposium on Distributed Autonomous Robotic Systems (DARS), June 2021. [PDF Preprint]

We use blockchain to provide Byzantine fault tolerance to multi-robot problems, such as multi-robot pickup and delivery. 

K. Strawn and N. Ayanian. "Byzantine Fault Tolerant Multi-Robot Planning for Online Pickup and Delivery", in ICRA Workshop, Foundational Problems In Multi-Robot Coordination Under Uncertainty And Adversarial Attacks, June 2020.