Few Shot System Identification for Reinforcement Learning
Karim Farid, Nourhan Sakr
TL;DR
The paper tackles the challenge of robust, data-efficient dynamics learning for reinforcement learning by proposing few-shot online system identification through probabilistic latent dynamics embeddings. It introduces two variational, neural-ODE based models, VCNODETI (time-invariant) and VCNODET (time-variant), that encode system dynamics from limited recent history and predict future states for model-based control. By operating in a latent, linearly-structured space and using model predictive control, the framework achieves high sample efficiency and robustness across variations in model parameters. Experimental results on toy datasets and OpenAI gym benchmarks with domain randomization demonstrate accurate trajectory prediction and effective latent-space MPC, supporting potential for simulation-to-real transfer. The work advances online adaptation in MBRL by separating dynamics encoding from state representation and leveraging variational inference to capture parameter uncertainty.
Abstract
Learning by interaction is the key to skill acquisition for most living organisms, which is formally called Reinforcement Learning (RL). RL is efficient in finding optimal policies for endowing complex systems with sophisticated behavior. All paradigms of RL utilize a system model for finding the optimal policy. Modeling dynamics can be done by formulating a mathematical model or system identification. Dynamic models are usually exposed to aleatoric and epistemic uncertainties that can divert the model from the one acquired and cause the RL algorithm to exhibit erroneous behavior. Accordingly, the RL process sensitive to operating conditions and changes in model parameters and lose its generality. To address these problems, Intensive system identification for modeling purposes is needed for each system even if the model dynamics structure is the same, as the slight deviation in the model parameters can render the model useless in RL. The existence of an oracle that can adaptively predict the rest of the trajectory regardless of the uncertainties can help resolve the issue. The target of this work is to present a framework for facilitating the system identification of different instances of the same dynamics class by learning a probability distribution of the dynamics conditioned on observed data with variational inference and show its reliability in robustly solving different instances of control problems with the same model in model-based RL with maximum sample efficiency.