This paper presents a logic programming-based framework for policy-aware autonomous agents that can reason about potential penalties for non-compliance and act accordingly. While prior work has primarily focused on ensuring compliance, our approach considers scenarios where deviating from policies may be necessary to achieve high-stakes goals. Additionally, modeling non-compliant behavior can assist policymakers by simulating realistic human decision-making. Our framework extends Gelfond and Lobo's Authorization and Obligation Policy Language (AOPL) to incorporate penalties and integrates Answer Set Programming (ASP) for reasoning. Compared to previous approaches, our method ensures well-formed policies, accounts for policy priorities, and enhances explainability by explicitly identifying rule violations and their consequences. Building on the work of Harders and Inclezan, we introduce penalty-based reasoning to distinguish between non-compliant plans, prioritizing those with minimal repercussions. To support this, we develop an automated translation from the extended AOPL into ASP and refine ASP-based planning algorithms to account for incurred penalties. Experiments in two domains demonstrate that our framework generates higher-quality plans that avoid harmful actions while, in some cases, also improving computational efficiency. These findings underscore its potential for enhancing autonomous decision-making and informing policy refinement.

Open-ended AI agents need to be able to learn efficiently goals of increasing complexity, abstraction and heterogeneity over their lifetime. Beyond sampling efficiently their own goals, autotelic agents specifically need to be able to keep the growing complexity of goals under control, limiting the associated growth in sample and computational complexity. To adress this challenge, recent approaches have leveraged hierarchical reinforcement learning (HRL) and language, capitalizing on its compositional and combinatorial generalization capabilities to acquire temporally extended reusable behaviours. Existing approaches use expert defined spaces of subgoals over which they instantiate a hierarchy, and often assume pre-trained associated low-level policies. Such designs are inadequate in open-ended scenarios, where goal spaces naturally diversify across a broad spectrum of difficulties. We introduce HERAKLES, a framework that enables a two-level hierarchical autotelic agent to continuously compile mastered goals into the low-level policy, executed by a small, fast neural network, dynamically expanding the set of subgoals available to the high-level policy. We train a Large Language Model (LLM) to serve as the high-level controller, exploiting its strengths in goal decomposition and generalization to operate effectively over this evolving subgoal space. We evaluate HERAKLES in the open-ended Crafter environment and show that it scales effectively with goal complexity, improves sample efficiency through skill compilation, and enables the agent to adapt robustly to novel challenges over time.

This paper addresses the challenge of navigation in large, visually complex environments with sparse rewards. We propose a method that uses object-oriented macro actions grounded in a topological map, allowing a simple Deep Q-Network (DQN) to learn effective navigation policies. The agent builds a map by detecting objects from RGBD input and selecting discrete macro actions that correspond to navigating to these objects. This abstraction drastically reduces the complexity of the underlying reinforcement learning problem and enables generalization to unseen environments. We evaluate our approach in a photorealistic 3D simulation and show that it significantly outperforms a random baseline under both immediate and terminal reward conditions. Our results demonstrate that topological structure and macro-level abstraction can enable sample-efficient learning even from pixel data.

Discussions on causal relations in real life often consider variables for which the definition of causality is unclear since the notion of interventions on the respective variables is obscure. Asking 'what qualifies an action for being an intervention on the variable X' raises the question whether the action impacted all other variables only through X or directly, which implicitly refers to a causal model. To avoid this known circularity, we instead suggest a notion of 'phenomenological causality' whose basic concept is a set of elementary actions. Then the causal structure is defined such that elementary actions change only the causal mechanism at one node (e.g. one of the causal conditionals in the Markov factorization). This way, the Principle of Independent Mechanisms becomes the defining property of causal structure in domains where causality is a more abstract phenomenon rather than being an objective fact relying on hard-wired causal links between tangible objects. We describe this phenomenological approach to causality for toy and hypothetical real-world examples and argue that it is consistent with the causal Markov condition when the system under consideration interacts with other variables that control the elementary actions.

This article describes an action rule induction algorithm based on a sequential covering approach. Two variants of the algorithm are presented. The algorithm allows the action rule induction from a source and a target decision class point of view. The application of rule quality measures enables the induction of action rules that meet various quality criteria. The article also presents a method for recommendation induction. The recommendations indicate the actions to be taken to move a given test example, representing the source class, to the target one. The recommendation method is based on a set of induced action rules. The experimental part of the article presents the results of the algorithm operation on sixteen data sets. As a result of the conducted research the Ac-Rules package was made available.

Playing repeated matrix games (RMG) while maximizing the cumulative returns is a basic method to evaluate multi-agent learning (MAL) algorithms. Previous work has shown that UCB, M3, S or Exp3 algorithms have good behaviours on average in RMG. Besides, hedging algorithms have been shown to be effective on prediction problems. An hedging algorithm is made up with a top-level algorithm and a set of basic algorithms. To make its decision, an hedging algorithm uses its top-level algorithm to choose a basic algorithm, and the chosen algorithm makes the decision. This paper experimentally shows that well-selected hedging algorithms are better on average than all previous MAL algorithms on the task of playing RMG against various players. S is a very good top-level algorithm, and UCB and M3 are very good basic algorithms. Furthermore, two-level hedging algorithms are more effective than one-level hedging algorithms, and three levels are not better than two levels.

Complex mission or task specification languages play a fundamentally important role in human/robotic interaction. In realistic scenarios such as emergency response, specifying temporal, resource and other constraints on a mission is an essential component due to the dynamic and contingent nature of the operational environments. It is also desirable that in addition to having a formal semantics, the language should be sufficiently expressive, pragmatic and abstract. The main goal of this paper is to propose a mission specification language that meets these requirements. It is based on extending both the syntax and semantics of a well-established formalism for reasoning about action and change, Temporal Action Logic (TAL), in order to represent temporal composite actions with constraints. Fixpoints are required to specify loops and recursion in the extended language. The results include a sound and complete proof theory for this extension. To ensure that the composite language constructs are adequately grounded in the pragmatic operation of robotic systems, Task Specification Trees (TSTs) and their mapping to these constructs are proposed. The expressive and pragmatic adequacy of this approach is demonstrated using an emergency response scenario.

In this paper, we confront the problem of applying reinforcement learning to agents that perceive the environment through many sensors and that can perform parallel actions using many actuators as is the case in complex autonomous robots. We argue that reinforcement learning can only be successfully applied to this case if strong assumptions are made on the characteristics of the environment in which the learning is performed, so that the relevant sensor readings and motor commands can be readily identified. The introduction of such assumptions leads to strongly-biased learning systems that can eventually lose the generality of traditional reinforcement-learning algorithms. In this line, we observe that, in realistic situations, the reward received by the robot depends only on a reduced subset of all the executed actions and that only a reduced subset of the sensor inputs (possibly different in each situation and for each action) are relevant to predict the reward. We formalize this property in the so called 'categorizability assumption' and we present an algorithm that takes advantage of the categorizability of the environment, allowing a decrease in the learning time with respect to existing reinforcement-learning algorithms. Results of the application of the algorithm to a couple of simulated realistic-robotic problems (landmark-based navigation and the six-legged robot gait generation) are reported to validate our approach and to compare it to existing flat and generalization-based reinforcement-learning approaches.

In this paper, we confront the problem of applying reinforcement learning to agents that perceive the environment through many sensors and that can perform parallel actions using many actuators as is the case in complex autonomous robots. We argue that reinforcement learning can only be successfully applied to this case if strong assumptions are made on the characteristics of the environment in which the learning is performed, so that the relevant sensor readings and motor commands can be readily identified. The introduction of such assumptions leads to strongly-biased learning systems that can eventually lose the generality of traditional reinforcement-learning algorithms. In this line, we observe that, in realistic situations, the reward received by the robot depends only on a reduced subset of all the executed actions and that only a reduced subset of the sensor inputs (possibly different in each situation and for each action) are relevant to predict the reward. We formalize this property in the so called 'categorizability assumption' and we present an algorithm that takes advantage of the categorizability of the environment, allowing a decrease in the learning time with respect to existing reinforcement-learning algorithms. Results of the application of the algorithm to a couple of simulated realistic-robotic problems (landmark-based navigation and the six-legged robot gait generation) are reported to validate our approach and to compare it to existing flat and generalization-based reinforcement-learning approaches.