We present an algorithm for learning mixtures of Markov chains and Markov decision processes (MDPs) from short unlabeled trajectories. Specifically, our method handles mixtures of Markov chains with optional control input by going through a multi-step process, involving (1) a subspace estimation step, (2) spectral clustering of trajectories using "pairwise distance estimators," along with refinement using the EM algorithm, (3) a model estimation step, and (4) a classification step for predicting labels of new trajectories. We provide end-to-end performance guarantees, where we only explicitly require the length of trajectories to be linear in the number of states and the number of trajectories to be linear in a mixing time parameter. Experimental results support these guarantees, where we attain 96.6% average accuracy on a mixture of two MDPs in gridworld, outperforming the EM algorithm with random initialization (73.2% average accuracy).

AI can not only outperform people in many planning tasks, but also teach them how to plan better. All prior work was conducted in fully observable environments, but the real world is only partially observable. To bridge this gap, we developed the first metareasoning algorithm for discovering resource-rational strategies for human planning in partially observable environments. Moreover, we developed an intelligent tutor teaching the automatically discovered strategy by giving people feedback on how they plan in increasingly more difficult problems. We showed that our strategy discovery method is superior to the state-of-the-art and tested our intelligent tutor in a preregistered training experiment with 330 participants. The experiment showed that people's intuitive strategies for planning in partially observable environments are highly suboptimal, but can be substantially improved by training with our intelligent tutor. This suggests our human-centred tutoring approach can successfully boost human planning in complex, partially observable sequential decision problems.

We study the problem of reinforcement learning for a task encoded by a reward machine. The task is defined over a set of properties in the environment, called atomic propositions, and represented by Boolean variables. One unrealistic assumption commonly used in the literature is that the truth values of these propositions are accurately known. In real situations, however, these truth values are uncertain since they come from sensors that suffer from imperfections. At the same time, reward machines can be difficult to model explicitly, especially when they encode complicated tasks. We develop a reinforcement-learning algorithm that infers a reward machine that encodes the underlying task while learning how to execute it, despite the uncertainties of the propositions' truth values. In order to address such uncertainties, the algorithm maintains a probabilistic estimate about the truth value of the atomic propositions; it updates this estimate according to new sensory measurements that arrive from the exploration of the environment. Additionally, the algorithm maintains a hypothesis reward machine, which acts as an estimate of the reward machine that encodes the task to be learned. As the agent explores the environment, the algorithm updates the hypothesis reward machine according to the obtained rewards and the estimate of the atomic propositions' truth value. Finally, the algorithm uses a Q-learning procedure for the states of the hypothesis reward machine to determine the policy that accomplishes the task. We prove that the algorithm successfully infers the reward machine and asymptotically learns a policy that accomplishes the respective task.

Connectivity technology has shown great potentials in improving the safety and efficiency of transportation systems by providing information beyond the perception and prediction capabilities of individual vehicles. However, it is expected that human-driven and autonomous vehicles, and connected and non-connected vehicles need to share the transportation network during the transition period to fully connected and automated transportation systems. Such mixed traffic scenarios significantly increase the complexity in analyzing system behavior and quantifying uncertainty for highly interactive scenarios, e.g., lane changing. It is even harder to ensure system safety when neural network based planners are leveraged to further improve efficiency. In this work, we propose a connectivity-enhanced neural network based lane changing planner. By cooperating with surrounding connected vehicles in dynamic environment, our proposed planner will adapt its planned trajectory according to the analysis of a safe evasion trajectory. We demonstrate the strength of our planner design in improving efficiency and ensuring safety in various mixed traffic scenarios with extensive simulations. We also analyze the system robustness when the communication or coordination is not perfect.

We initiate the study of federated reinforcement learning under environmental heterogeneity by considering a policy evaluation problem. Our setup involves $N$ agents interacting with environments that share the same state and action space but differ in their reward functions and state transition kernels. Assuming agents can communicate via a central server, we ask: Does exchanging information expedite the process of evaluating a common policy? To answer this question, we provide the first comprehensive finite-time analysis of a federated temporal difference (TD) learning algorithm with linear function approximation, while accounting for Markovian sampling, heterogeneity in the agents' environments, and multiple local updates to save communication. Our analysis crucially relies on several novel ingredients: (i) deriving perturbation bounds on TD fixed points as a function of the heterogeneity in the agents' underlying Markov decision processes (MDPs); (ii) introducing a virtual MDP to closely approximate the dynamics of the federated TD algorithm; and (iii) using the virtual MDP to make explicit connections to federated optimization. Putting these pieces together, we rigorously prove that in a low-heterogeneity regime, exchanging model estimates leads to linear convergence speedups in the number of agents.

Optimization of human-AI teams hinges on the AI's ability to tailor its interaction to individual human teammates. A common hypothesis in adaptive AI research is that minor differences in people's predisposition to trust can significantly impact their likelihood of complying with recommendations from the AI. Predisposition to trust is often measured with self-report inventories that are administered before interactions. We benchmark a popular measure of this kind against behavioral predictors of compliance. We find that the inventory is a less effective predictor of compliance than the behavioral measures in datasets taken from three previous research projects. This suggests a general property that individual differences in initial behavior are more predictive than differences in self-reported trust attitudes. This result also shows a potential for easily accessible behavioral measures to provide an AI with more accurate models without the use of (often costly) survey instruments.

AI constructivism as inspired by Jean Piaget, described and surveyed by Frank Guerin, and representatively implemented by Gary Drescher seeks to create algorithms and knowledge structures that enable agents to acquire, maintain, and apply a deep understanding of the environment through sensorimotor interactions. This paper aims to increase awareness of constructivist AI implementations to encourage greater progress toward enabling lifelong learning by machines. It builds on Guerin's 2008 "Learning Like a Baby: A Survey of AI approaches." After briefly recapitulating that survey, it summarizes subsequent progress by the Guerin referents, numerous works not covered by Guerin (or found in other surveys), and relevant efforts in related areas. The focus is on knowledge representations and learning algorithms that have been used in practice viewed through lenses of Piaget's schemas, adaptation processes, and staged development. The paper concludes with a preview of a simple framework for constructive AI being developed by the author that parses concepts from sensory input and stores them in a semantic memory network linked to episodic data.

POMDPs capture a broad class of decision making problems, but hardness results suggest that learning is intractable even in simple settings due to the inherent partial observability. However, in many realistic problems, more information is either revealed or can be computed during some point of the learning process. Motivated by diverse applications ranging from robotics to data center scheduling, we formulate a Hindsight Observable Markov Decision Process (HOMDP) as a POMDP where the latent states are revealed to the learner in hindsight and only during training. We introduce new algorithms for the tabular and function approximation settings that are provably sample-efficient with hindsight observability, even in POMDPs that would otherwise be statistically intractable. We give a lower bound showing that the tabular algorithm is optimal in its dependence on latent state and observation cardinalities.

Model-based methods have recently shown great potential for off-policy evaluation (OPE); offline trajectories induced by behavioral policies are fitted to transitions of Markov decision processes (MDPs), which are used to rollout simulated trajectories and estimate the performance of policies. Model-based OPE methods face two key challenges. First, as offline trajectories are usually fixed, they tend to cover limited state and action space. Second, the performance of model-based methods can be sensitive to the initialization of their parameters. In this work, we propose the variational latent branching model (VLBM) to learn the transition function of MDPs by formulating the environmental dynamics as a compact latent space, from which the next states and rewards are then sampled. Specifically, VLBM leverages and extends the variational inference framework with the recurrent state alignment (RSA), which is designed to capture as much information underlying the limited training data, by smoothing out the information flow between the variational (encoding) and generative (decoding) part of VLBM. Moreover, we also introduce the branching architecture to improve the model's robustness against randomly initialized model weights. The effectiveness of the VLBM is evaluated on the deep OPE (DOPE) benchmark, from which the training trajectories are designed to result in varied coverage of the state-action space. We show that the VLBM outperforms existing state-of-the-art OPE methods in general.

Development of autonomous cyber system defense strategies and action recommendations in the real-world is challenging, and includes characterizing system state uncertainties and attack-defense dynamics. We propose a data-driven deep reinforcement learning (DRL) framework to learn proactive, context-aware, defense countermeasures that dynamically adapt to evolving adversarial behaviors while minimizing loss of cyber system operations. A dynamic defense optimization problem is formulated with multiple protective postures against different types of adversaries with varying levels of skill and persistence. A custom simulation environment was developed and experiments were devised to systematically evaluate the performance of four model-free DRL algorithms against realistic, multi-stage attack sequences. Our results suggest the efficacy of DRL algorithms for proactive cyber defense under multi-stage attack profiles and system uncertainties.