Developing reinforcement learning agents that can generalise effectively to new tasks is one of the main challenges in AI research. This paper introduces Fracture Cluster Options (FraCOs), a multi-level hierarchical reinforcement learning method designed to improve generalisation performance. FraCOs identifies patterns in agent behaviour and forms temporally-extended actions (options) based on the expected future usefulness of those patterns, enabling rapid adaptation to new tasks. In tabular settings, FraCOs demonstrates effective transfer and improves performance as the depth of the hierarchy increases. In several complex procedurally-generated environments, FraCOs consistently outperforms state-of-the-art deep reinforcement learning algorithms, achieving superior results in both in-distribution and out-of-distribution scenarios.

Reinforcement learning has achieved remarkable success in complex decision-making environments, yet its lack of transparency limits its deployment in practice, especially in safety-critical settings. Shapley values from cooperative game theory provide a principled framework for explaining reinforcement learning; however, the computational cost of Shapley explanations is an obstacle for their use. We introduce FastSVERL, a scalable method for explaining reinforcement learning by approximating Shapley values. FastSVERL is designed to handle the unique challenges of reinforcement learning, including temporal dependencies across multi-step trajectories, learning from off-policy data, and adapting to evolving agent behaviours in real time. FastSVERL introduces a practical, scalable approach for principled and rigourous interpretability in reinforcement learning.

Efficient exploration remains one of the most important open problems in reinforcement learning. Discovering novel states or transitions requires policies that efficiently direct the agent away from the regions of the state space that are already well explored. We introduce Novel Exploration via Orthogonality (NEO), an approach that automatically uncovers not only which regions of the environment are novel but also how to reach them by leveraging Laplacian representations. NEO uses the eigenvectors of a modified graph Laplacian to induce gradient flows from states that are frequently visited (less novel) to states that are seldom visited (more novel). We show that NEO's modified Laplacian yields eigenvectors whose extreme values align with the most novel regions of the state space. We provide bounds for the eigenvalues of the modified Laplacian; and we show that the smoothest eigenvectors with real eigenvalues below certain thresholds provide guaranteed gradients to novel states for both undirected and directed graphs. In an empirical evaluation in online, incremental settings, NEO outperformed related state-of-the-art approaches, including eigen-options and cover options, in a large collection of undirected and directed environments with varying connectivity structures.

Station-keeping tasks for high-altitude balloons show promise in areas such as ecological surveys, atmospheric analysis, and communication relays. However, identifying the optimal time and position to launch a latex high-altitude balloon is still a challenging and multifaceted problem. For example, tasks such as forest fire tracking place geometric constraints on the launch location of the balloon. Furthermore, identifying the most optimal location also heavily depends on atmospheric conditions. We first illustrate how reinforcement learning-based controllers, frequently used for station-keeping tasks, can exploit the environment. This exploitation can degrade performance on unseen weather patterns and affect station-keeping performance when identifying an optimal launch configuration. Valuing all states equally in the region, the agent exploits the region’s geometry by flying near the edge, leading to risky behaviours. We propose a modification which compensates for this exploitation and finds this leads to, on average, higher steps within the target region on unseen data. Then, we illustrate how Bayesian Optimisation (BO) can identify the optimal launch location to perform station-keeping tasks, maximising the return from a given rollout. We show BO can find this launch location in fewer steps compared to other optimisation methods. Results indicate that, surprisingly, the most optimal location to launch from is not commonly within the target region. Please find further information about our project at https://sites.google.com/view/bo-lauch-balloon/.

For reinforcement learning systems to be widely adopted, their users must understand and trust them. We present a theoretical analysis of explaining reinforcement learning using Shapley values, following a principled approach from game theory for identifying the contribution of individual players to the outcome of a cooperative game. We call this general framework Shapley Values for Explaining Reinforcement Learning (SVERL). Our analysis exposes the limitations of earlier uses of Shapley values in reinforcement learning. We then develop an approach that uses Shapley values to explain agent performance. In a variety of domains, SVERL produces meaningful explanations that match and supplement human intuition.

High altitude balloons have proved useful for ecological aerial surveys, atmospheric monitoring, and communication relays. However, due to weight and power constraints, there is a need to investigate alternate modes of propulsion to navigate in the stratosphere. Very recently, reinforcement learning has been proposed as a control scheme to maintain balloons in the region of a fixed location, facilitated through diverse opposing wind-fields at different altitudes. Although air-pump based station keeping has been explored, there is no research on the control problem for venting and ballasting actuated balloons, which is commonly used as a low-cost alternative. We show how reinforcement learning can be used for this type of balloon. Specifically, we use the soft actor-critic algorithm, which on average is able to station-keep within 50 km for on average 25% of the flight, consistent with state-of-the-art. Furthermore, we show that the proposed controller effectively minimises the consumption of resources, thereby supporting long duration flights. We frame the controller as a continuous control reinforcement learning problem, which allows for a more diverse range of trajectories, as opposed to current state-of-the-art work, which uses discrete action spaces. Furthermore, through continuous control, we can make use of larger ascent rates which are not possible using air-pumps. The desired ascent-rate is decoupled into desired altitude and time-factor to provide a more transparent policy, compared to low-level control commands used in previous works. Finally, by applying the equations of motion, we establish appropriate thresholds for venting and ballasting to prevent the agent from exploiting the environment. More specifically, we ensure actions are physically feasible by enforcing constraints on venting and ballasting.

What is a useful skill hierarchy for an autonomous agent? We propose an answer based on a graphical representation of how the interaction between an agent and its environment may unfold. Our approach uses modularity maximisation as a central organising principle to expose the structure of the interaction graph at multiple levels of abstraction. The result is a collection of skills that operate at varying time scales, organised into a hierarchy, where skills that operate over longer time scales are composed of skills that operate over shorter time scales. The entire skill hierarchy is generated automatically, with no human input, including the skills themselves (their behaviour, when they can be called, and when they terminate) as well as the dependency structure between them. In a wide range of environments, this approach generates skill hierarchies that are intuitively appealing and that considerably improve the learning performance of the agent.

Learning from small data sets is difficult in the absence of specific domain knowledge. We present a regularized linear model called STEW that benefits from a generic and prevalent form of prior knowledge: feature directions. STEW shrinks weights toward each other, converging to an equal-weights solution in the limit of infinite regularization. We provide theoretical results on the equal-weights solution that explains how STEW can productively trade-off bias and variance. Across a wide range of learning problems, including Tetris, STEW outperformed existing linear models, including ridge regression, the Lasso, and the non-negative Lasso, when feature directions were known. The model proved to be robust to unreliable (or absent) feature directions, still outperforming alternative models under diverse conditions. Our results in Tetris were obtained by using a novel approach to learning in sequential decision environments based on multinomial logistic regression.

We examined the sequence of decision problems that are encountered in the game of Tetris and found that most of the problems are easy in the following sense: One can choose well among the available actions without knowing an evaluation function that scores well in the game. This is a consequence of three conditions that are prevalent in the game: simple dominance, cumulative dominance, and noncompensation. These conditions can be exploited to develop faster and more effective learning algorithms. In addition, they allow certain types of domain knowledge to be incorporated with ease into a learning algorithm. Among the sequential decision problems we encounter, it is unlikely that Tetris is unique or rare in having these properties.

Simple decision heuristics are models of human and animal behavior that use few pieces of information---perhaps only a single piece of information---and integrate the pieces in simple ways, for example, by considering them sequentially, one at a time, or by giving them equal weight. It is unknown how quickly these heuristics can be learned from experience. We show, analytically and empirically, that only a few training samples lead to substantial progress in learning. We focus on three families of heuristics: single-cue decision making, lexicographic decision making, and tallying. Our empirical analysis is the most extensive to date, employing 63 natural data sets on diverse subjects.

Many attempts to understand the success of simple decision heuristics have examined heuristics as an approximation to a linear decision rule. This research has identified three environmental structures that aid heuristics: dominance, cumulative dominance, and noncompensatoriness. Here, we further develop these ideas and examine their empirical relevance in 51 natural environments. We find that all three structures are prevalent, making it possible for some simple rules to reach the accuracy levels of the linear decision rule using less information.

We present a characterization of a useful class of skills based on a graphical representation of an agent's interaction with its environment. Our characterization uses betweenness, a measure of centrality on graphs. It may be used directly to form a set of skills suitable for a given environment. More importantly, it serves as a useful guide for developing online, incremental skill discovery algorithms that do not rely on knowing or representing the environment graph in its entirety.

How should a reinforcement learning agent act if its sole purpose is to efficiently learn an optimal policy for later use? In other words, how should it explore, to be able to exploit later? We formulate this problem as a Markov Decision Process by explicitly modeling the internal state of the agent and propose a principled heuristic for its solution. We present experimental results in a number of domains, also exploring the algorithm's use for learning a policy for a skill given its reward function---an important but neglected component of skill discovery.

We present a new subgoal-based method for automatically creating useful skills in reinforcement learning. Our method identifies subgoals by partitioning local state transition graphs---those that are constructed using only the most recent experiences of the agent. The local scope of our subgoal discovery method allows it to successfully identify the type of subgoals we seek---states that lie between two densely-connected regions of the state space while producing an algorithm with low computational cost.

We describe an application of relational knowledge discovery to a key regulatory mission of the National Association of Securities Dealers (NASD). NASD is the world's largest privatesector securities regulator, with responsibility for preventing and discovering misconduct among securities brokers. Our goal was to help focus NASD’s limited regulatory resources on the brokers who are most likely to engage in securities violations. Using statistical relational learning algorithms, we developed models that rank brokers with respect to the probability that they would commit a serious violation of securities regulations in the near future. Our models incorporate organizational relationships among brokers (e.g., past coworker), which domain experts consider important but cannot easily use otherwise. The learned models were subjected to an extensive evaluation using more than 18 months of data unseen by the model developers and comprising over two person weeks of effort by NASD staff. Model predictions were found to correlate highly with the subjective evaluations of experienced NASD examiners. Furthermore, in all performance measures, our models performed as well as or better than the handcrafted rules that are currently in use at NASD. 1.

We present a new method for automatically creating useful temporal abstractions in reinforcement learning. We argue that states that allow the agent to transition to a different region of the state space are useful subgoals, and propose a method for identifying them using the concept of relative novelty. When such a state is identified, a temporally-extended activity (e.g., an option) is generated that takes the agent efficiently to this state. We illustrate the utility of the method in a number of tasks.