Projects

Modular library for safe RL, providing baselines for a number of constraints using a variety of algorithms.

Probabilistic shielding for decentralised multiagent RL with dynamics induced through world and opponent modelling.

Adaptive shielding using GR(1) specifications and Inductive Logic Programming to repair specifications.

Probabilistic shielding for constrained Markov Decision Processes with unknown safety dynamics.

Write formal Reinforcement Learning reward specifications in Quantitative Linear Temporal Logic on finite traces.

Latent shielding for safe RL, including continuous dynamics using DreamerV3.