Imitation learning is an approach for generating intelligent behavior when the cost function is unknown or difficult to specify. Building upon work in inverse reinforcement learning (IRL), Generative Adversarial Imitation Learning (GAIL) aims to provide effective imitation even for problems with large or continuous state and action spaces. Driver modeling is one example of a problem where the state and action spaces are continuous. Human driving behavior is characterized by non-linearity and stochasticity, and the underlying cost function is unknown. As a result, learning from human driving demonstrations is a promising approach for generating human-like driving behavior. This article describes the use of GAIL for learning-based driver modeling. Because driver modeling is inherently a multi-agent problem, where the interaction between agents needs to be modeled, this paper describes a parameter-sharing extension of GAIL called PS-GAIL to tackle multi-agent driver modeling. In addition, GAIL is domain agnostic, making it difficult to encode specific knowledge relevant to driving in the learning process. This paper describes Reward Augmented Imitation Learning (RAIL), which modifies the reward signal to provide domain-specific knowledge to the agent. Finally, human demonstrations are dependent upon latent factors that may not be captured by GAIL. This paper describes Burn-InfoGAIL, which allows for disentanglement of latent variability in demonstrations. Imitation learning experiments are performed using NGSIM, a real-world highway driving dataset. Experiments show that these modifications to GAIL can successfully model highway driving behavior, accurately replicating human demonstrations and generating realistic, emergent behavior in the traffic flow arising from the interaction between driving agents.

Automated lane changing is a critical feature for advanced autonomous driving systems. In recent years, reinforcement learning (RL) algorithms trained on traffic simulators yielded successful results in computing lane changing policies that strike a balance between safety, agility and compensating for traffic uncertainty. However, many RL algorithms exhibit simulator bias and policies trained on simple simulators do not generalize well to realistic traffic scenarios. In this work, we develop a data driven traffic simulator by training a generative adverserial network (GAN) on real life trajectory data. The simulator generates randomized trajectories that resembles real life traffic interactions between vehicles, which enables training the RL agent on much richer and realistic scenarios. We demonstrate through simulations that RL agents that are trained on GAN-based traffic simulator has stronger generalization capabilities compared to RL agents trained on simple rule-driven simulators.

Individuality is essential in human society, which induces the division of labor and thus improves the efficiency and productivity. Similarly, it should also be the key to multi-agent cooperation. Inspired by that individuality is of being an individual separate from others, we propose a simple yet efficient method for the emergence of individuality (EOI) in multi-agent reinforcement learning (MARL). EOI learns a probabilistic classifier that predicts a probability distribution over agents given their observation and gives each agent an intrinsic reward of being correctly predicted by the classifier. The intrinsic reward encourages the agents to visit their own familiar observations, and learning the classifier by such observations makes the intrinsic reward signals stronger and the agents more identifiable. To further enhance the intrinsic reward and promote the emergence of individuality, two regularizers are proposed to increase the discriminability of the classifier. We implement EOI on top of popular MARL algorithms. Empirically, we show that EOI significantly outperforms existing methods in a variety of multi-agent cooperative scenarios.

In this paper, we present a new reinforcement learning (RL) algorithm called Distributional Soft Actor Critic (DSAC), which exploits the distributional information of accumulated rewards to achieve better performance. Seamlessly integrating SAC (which uses entropy to encourage exploration) with a principled distributional view of the underlying objective, DSAC takes into consideration the randomness in both action and rewards, and beats the state-of-the-art baselines in several continuous control benchmarks. Moreover, with the distributional information of rewards, we propose a unified framework for risk-sensitive learning, one that goes beyond maximizing only expected accumulated rewards. Under this framework we discuss three specific risk-related metrics: percentile, mean-variance and distorted expectation. Our extensive experiments demonstrate that with distribution modeling in RL, the agent performs better for both risk-averse and risk-seeking control tasks.

In the realm of games research, Artificial General Intelligence algorithms often use score as main reward signal for learning or playing actions. However this has shown its severe limitations when the point rewards are very rare or absent until the end of the game. This paper proposes a new approach based on event logging: the game state triggers an event every time one of its features changes. These events are processed by an Event-value Function (EF) that assigns a value to a single action or a sequence. The experiments have shown that such approach can mitigate the problem of scarce point rewards and improve the AI performance. Furthermore this represents a step forward in controlling the strategy adopted by the artificial agent, by describing a much richer and controllable behavioural space through the EF. Tuned EF are able to neatly synthesise the relevance of the events in the game. Agents using an EF show more robust when playing games with several opponents.

In recent years, on-policy reinforcement learning (RL) has been successfully applied to many different continuous control tasks. While RL algorithms are often conceptually simple, their state-of-the-art implementations take numerous low- and high-level design decisions that strongly affect the performance of the resulting agents. Those choices are usually not extensively discussed in the literature, leading to discrepancy between published descriptions of algorithms and their implementations. This makes it hard to attribute progress in RL and slows down overall progress [Engstrom'20]. As a step towards filling that gap, we implement >50 such ``choices'' in a unified on-policy RL framework, allowing us to investigate their impact in a large-scale empirical study. We train over 250'000 agents in five continuous control environments of different complexity and provide insights and practical recommendations for on-policy training of RL agents.

In model-free reinforcement learning, the temporal difference method and its variants become unstable when combined with nonlinear function approximations. Bellman residual minimization with stochastic gradient descent (SGD) is more stable, but it suffers from the double sampling problem: given the current state, two independent samples for the next state are required, but often only one sample is available. Recently, the authors of Zhu et al. [2020] introduced the borrowing from the future (BFF) algorithm to address this issue for the prediction problem. The main idea is to borrow extra randomness from the future to approximately re-sample the next state when the underlying dynamics of the problem are sufficiently smooth. This paper extends the BFF algorithm to action-value function based model-free control. We prove that BFF is close to unbiased SGD when the underlying dynamics vary slowly with respect to actions. We confirm our theoretical findings with numerical simulations.

We consider the question of learning $Q$-function in a sample efficient manner for reinforcement learning with continuous state and action spaces under a generative model. If $Q$-function is Lipschitz continuous, then the minimal sample complexity for estimating $\epsilon$-optimal $Q$-function is known to scale as ${\Omega}(\frac{1}{\epsilon^{d_1+d_2 +2}})$ per classical non-parametric learning theory, where $d_1$ and $d_2$ denote the dimensions of the state and action spaces respectively. The $Q$-function, when viewed as a kernel, induces a Hilbert-Schmidt operator and hence possesses square-summable spectrum. This motivates us to consider a parametric class of $Q$-functions parameterized by its "rank" $r$, which contains all Lipschitz $Q$-functions as $r \to \infty$. As our key contribution, we develop a simple, iterative learning algorithm that finds $\epsilon$-optimal $Q$-function with sample complexity of $\widetilde{O}(\frac{1}{\epsilon^{\max(d_1, d_2)+2}})$ when the optimal $Q$-function has low rank $r$ and the discounting factor $\gamma$ is below a certain threshold. Thus, this provides an exponential improvement in sample complexity. To enable our result, we develop a novel Matrix Estimation algorithm that faithfully estimates an unknown low-rank matrix in the $\ell_\infty$ sense even in the presence of arbitrary bounded noise, which might be of interest in its own right. Empirical results on several stochastic control tasks confirm the efficacy of our "low-rank" algorithms.

In the optimization of dynamical systems, the variables typically have constraints. Such problems can be modeled as a constrained Markov Decision Process (CMDP). This paper considers a model-free approach to the problem, where the transition probabilities are not known. In the presence of long-term (or average) constraints, the agent has to choose a policy that maximizes the long-term average reward as well as satisfy the average constraints in each episode. The key challenge with the long-term constraints is that the optimal policy is not deterministic in general, and thus standard Q-learning approaches cannot be directly used. This paper uses concepts from constrained optimization and Q-learning to propose an algorithm for CMDP with long-term constraints. For any $\gamma\in(0,\frac{1}{2})$, the proposed algorithm is shown to achieve $O(T^{1/2+\gamma})$ regret bound for the obtained reward and $O(T^{1-\gamma/2})$ regret bound for the constraint violation, where $T$ is the total number of steps. We note that these are the first results on regret analysis for MDP with long-term constraints, where the transition probabilities are not known apriori.

Reinforcement learning (RL) has demonstrated its ability to solve high dimensional tasks by leveraging non-linear function approximators. These successes however are mostly achieved by 'black-box' policies in simulated domains. When deploying RL to the real world, several concerns regarding the use of a 'black-box' policy might be raised. In an effort to make the policies learned by RL more transparent, we propose in this paper a policy iteration scheme that retains a complex function approximator for its internal value predictions but constrains the policy to have a concise, hierarchical, and human-readable structure, based on a mixture of interpretable experts. We show that our proposed algorithm can learn compelling policies on continuous action deep RL benchmarks, matching the performance of neural network based policies, but returns policies that are more amenable to human inspection than neural network or linear-in-feature policies.