Hierarchical Decomposition of Nonlinear Dynamics and Control for System Identification and Policy Distillation

The control of nonlinear dynamical systems remains a major challenge for autonomous agents. Current trends in reinforcement learning (RL) focus on complex representations of dynamics and policies, which have yielded impressive results in solving a variety of hard control tasks. However, this new sophistication and extremely over-parameterized models have come with the cost of an overall reduction in our ability to interpret the resulting policies. In this paper, we take inspiration from the control community and apply the principles of hybrid switching systems in order to break down complex dynamics into simpler components. We exploit the rich representational power of probabilistic graphical models and derive an expectation-maximization (EM) algorithm for learning a sequence model to capture the temporal structure of the data and automatically decompose nonlinear dynamics into stochastic switching linear dynamical systems. Moreover, we show how this framework of switching models enables extracting hierarchies of Markovian and auto-regressive locally linear controllers from nonlinear experts in an imitation learning scenario.

A Nonparametric Off-Policy Policy Gradient

Reinforcement learning (RL) algorithms still suffer from high sample complexity despite outstanding recent successes. The need for intensive interactions with the environment is especially observed in many widely popular policy gradient algorithms that perform updates using on-policy samples. The price of such inefficiency becomes evident in real-world scenarios such as interaction-driven robot learning, where the success of RL has been rather limited. We address this issue by building on the general sample efficiency of off-policy algorithms. With nonparametric regression and density estimation methods we construct a nonparametric Bellman equation in a principled manner, which allows us to obtain closed-form estimates of the value function, and to analytically express the full policy gradient. We provide a theoretical analysis of our estimate to show that it is consistent under mild smoothness assumptions and empirically show that our approach has better sample efficiency than state-of-the-art policy gradient methods.

A Nonparametric Offpolicy Policy Gradient

Reinforcement learning (RL) algorithms still suffer from high sample complexity despite outstanding recent successes. The need for intensive interactions with the environment is especially observed in many widely popular policy gradient algorithms that perform updates using on-policy samples. The price of such inefficiency becomes evident in real-world scenarios such as interaction-driven robot learning, where the success of RL has been rather limited. We address this issue by building on the general sample efficiency of off-policy algorithms. With nonparametric regression and density estimation methods we construct a nonparametric Bellman equation in a principled manner, which allows us to obtain closed-form estimates of the value function, and to analytically express the full policy gradient. We provide a theoretical analysis of our estimate to show that it is consistent under mild smoothness assumptions and empirically show that our approach has better sample efficiency than state-of-the-art policy gradient methods.

Receding Horizon Curiosity

Sample-efficient exploration is crucial not only for discovering rewarding experiences but also for adapting to environment changes in a task-agnostic fashion. A principled treatment of the problem of optimal input synthesis for system identification is provided within the framework of sequential Bayesian experimental design. In this paper, we present an effective trajectory-optimization-based approximate solution of this otherwise intractable problem that models optimal exploration in an unknown Markov decision process (MDP). By interleaving episodic exploration with Bayesian nonlinear system identification, our algorithm takes advantage of the inductive bias to explore in a directed manner, without assuming prior knowledge of the MDP. Empirical evaluations indicate a clear advantage of the proposed algorithm in terms of the rate of convergence and the final model fidelity when compared to intrinsic-motivation-based algorithms employing exploration bonuses such as prediction error and information gain. Moreover, our method maintains a computational advantage over a recent model-based active exploration (MAX) algorithm, by focusing on the information gain along trajectories instead of seeking a global exploration policy. A reference implementation of our algorithm and the conducted experiments is publicly available.

Stochastic Optimal Control as Approximate Input Inference

Optimal control of stochastic nonlinear dynamical systems is a major challenge in the domain of robot learning. Given the intractability of the global control problem, state-of-the-art algorithms focus on approximate sequential optimization techniques, that heavily rely on heuristics for regularization in order to achieve stable convergence. By building upon the duality between inference and control, we develop the view of Optimal Control as Input Estimation, devising a probabilistic stochastic optimal control formulation that iteratively infers the optimal input distributions by minimizing an upper bound of the control cost. Inference is performed through Expectation Maximization and message passing on a probabilistic graphical model of the dynamical system, and time-varying linear Gaussian feedback controllers are extracted from the joint state-action distribution. This perspective incorporates uncertainty quantification, effective initialization through priors, and the principled regularization inherent to the Bayesian treatment. Moreover, it can be shown that for deterministic linearized systems, our framework derives the maximum entropy linear quadratic optimal control law. We provide a complete and detailed derivation of our probabilistic approach and highlight its advantages in comparison to other deterministic and probabilistic solvers.

Self-Paced Contextual Reinforcement Learning

Generalization and adaptation of learned skills to novel situations is a core requirement for intelligent autonomous robots. Although contextual reinforcement learning provides a principled framework for learning and generalization of behaviors across related tasks, it generally relies on uninformed sampling of environments from an unknown, uncontrolled context distribution, thus missing the benefits of structured, sequential learning. We introduce a novel relative entropy reinforcement learning algorithm that gives the agent the freedom to control the intermediate task distribution, allowing for its gradual progression towards the target context distribution. Empirical evaluation shows that the proposed curriculum learning scheme drastically improves sample efficiency and enables learning in scenarios with both broad and sharp target context distributions in which classical approaches perform sub-optimally.

Model-Free Trajectory-based Policy Optimization with Monotonic Improvement

Many of the recent trajectory optimization algorithms alternate between linear approximation of the system dynamics around the mean trajectory and conservative policy update. One way of constraining the policy change is by bounding the Kullback-Leibler (KL) divergence between successive policies. These approaches already demonstrated great experimental success in challenging problems such as end-to-end control of physical systems. However, the linear approximation of the system dynamics can introduce a bias in the policy update and prevent convergence to the optimal policy. In this article, we propose a new model-free trajectory-based policy optimization algorithm with guaranteed monotonic improvement. The algorithm backpropagates a local, quadratic and time-dependent \qfunc~learned from trajectory data instead of a model of the system dynamics. Our policy update ensures exact KL-constraint satisfaction without simplifying assumptions on the system dynamics. We experimentally demonstrate on highly non-linear control tasks the improvement in performance of our algorithm in comparison to approaches linearizing the system dynamics. In order to show the monotonic improvement of our algorithm, we additionally conduct a theoretical analysis of our policy update scheme to derive a lower bound of the change in policy return between successive iterations.