On the Sample Efficiency of Abstractions and Potential-Based Reward Shaping in Reinforcement Learning

The use of Potential Based Reward Shaping (PBRS) has shown great promise in the ongoing research effort to tackle sample inefficiency in Reinforcement Learning (RL). However, the choice of the potential function is critical for this technique to be effective. Additionally, RL techniques are usually constrained to use a finite horizon for computational limitations. This introduces a bias when using PBRS, thus adding an additional layer of complexity. In this paper, we leverage abstractions to automatically produce a "good" potential function. We analyse the bias induced by finite horizons in the context of PBRS producing novel insights. Finally, to asses sample efficiency and performance impact, we evaluate our approach on four environments including a goal-oriented navigation task and three Arcade Learning Environments (ALE) games demonstrating that we can reach the same level of performance as CNN-based solutions with a simple fully-connected network.

On Computing Plans with Uniform Action Costs

In many real-world planning applications, agents might be interested in finding plans whose actions have costs that are as uniform as possible. Such plans provide agents with a sense of stability and predictability, which are key features when humans are the agents executing plans suggested by planning tools. This paper adapts three uniformity metrics to automated planning, and introduce planning-based compilations that allow to lexicographically optimize sum of action costs and action costs uniformity. Experimental results both in well-known and novel planning benchmarks show that the reformulated tasks can be effectively solved in practice to generate uniform plans.

Generalising Planning Environment Redesign

In Environment Design, one interested party seeks to affect another agent's decisions by applying changes to the environment. Most research on planning environment (re)design assumes the interested party's objective is to facilitate the recognition of goals and plans, and search over the space of environment modifications to find the minimal set of changes that simplify those tasks and optimise a particular metric. This search space is usually intractable, so existing approaches devise metric-dependent pruning techniques for performing search more efficiently. This results in approaches that are not able to generalise across different objectives and/or metrics. In this paper, we argue that the interested party could have objectives and metrics that are not necessarily related to recognising agents' goals or plans. Thus, to generalise the task of Planning Environment Redesign, we develop a general environment redesign approach that is metric-agnostic and leverages recent research on top-quality planning to efficiently redesign planning environments according to any interested party's objective and metric. Experiments over a set of environment redesign benchmarks show that our general approach outperforms existing approaches when using well-known metrics, such as facilitating the recognition of goals, as well as its effectiveness when solving environment redesign tasks that optimise a novel set of different metrics.

Contrastive Explanations of Multi-agent Optimization Solutions

In many real-world scenarios, agents are involved in optimization problems. Since most of these scenarios are over-constrained, optimal solutions do not always satisfy all agents. Some agents might be unhappy and ask questions of the form ``Why does solution $S$ not satisfy property $P$?''. In this paper, we propose MAoE, a domain-independent approach to obtain contrastive explanations by (i) generating a new solution $S^\prime$ where the property $P$ is enforced, while also minimizing the differences between $S$ and $S^\prime$; and (ii) highlighting the differences between the two solutions. Such explanations aim to help agents understanding why the initial solution is better than what they expected. We have carried out a computational evaluation that shows that MAoE can generate contrastive explanations for large multi-agent optimization problems. We have also performed an extensive user study in four different domains that shows that, after being presented with these explanations, humans' satisfaction with the original solution increases.

Fairness in Multi-Agent Planning

In cooperative Multi-Agent Planning (MAP), a set of goals has to be achieved by a set of agents. Independently of whether they perform a pre-assignment of goals to agents or they directly search for a solution without any goal assignment, most previous works did not focus on a fair distribution/achievement of goals by agents. This paper adapts well-known fairness schemes to MAP, and introduces two novel approaches to generate cost-aware fair plans. The first one solves an optimization problem to pre-assign goals to agents, and then solves a centralized MAP task using that assignment. The second one consists of a planning-based compilation that allows solving the joint problem of goal assignment and planning while taking into account the given fairness scheme. Empirical results in several standard MAP benchmarks show that these approaches outperform different baselines. They also show that there is no need to sacrifice much plan cost to generate fair plans.

Inapplicable Actions Learning for Knowledge Transfer in Reinforcement Learning

Reinforcement Learning (RL) algorithms are known to scale poorly to environments with many available actions, requiring numerous samples to learn an optimal policy. The traditional approach of considering the same fixed action space in every possible state implies that the agent must understand, while also learning to maximize its reward, to ignore irrelevant actions such as $\textit{inapplicable actions}$ (i.e. actions that have no effect on the environment when performed in a given state). Knowing this information can help reduce the sample complexity of RL algorithms by masking the inapplicable actions from the policy distribution to only explore actions relevant to finding an optimal policy. This is typically done in an ad-hoc manner with hand-crafted domain logic added to the RL algorithm. In this paper, we propose a more systematic approach to introduce this knowledge into the algorithm. We (i) standardize the way knowledge can be manually specified to the agent; and (ii) present a new framework to autonomously learn these state-dependent action constraints jointly with the policy. We show experimentally that learning inapplicable actions greatly improves the sample efficiency of the algorithm by providing a reliable signal to mask out irrelevant actions. Moreover, we demonstrate that thanks to the transferability of the knowledge acquired, it can be reused in other tasks to make the learning process more efficient.

Anticipatory Counterplanning

In competitive environments, commonly agents try to prevent opponents from achieving their goals. Most previous preventing approaches assume the opponent's goal is known a priori. Others only start executing actions once the opponent's goal has been inferred. In this work we introduce a novel domain-independent algorithm called Anticipatory Counterplanning. It combines inference of opponent's goals with computation of planning centroids to yield proactive counter strategies in problems where the opponent's goal is unknown. Experimental results show how this novel technique outperforms reactive counterplanning, increasing the chances of stopping the opponent from achieving its goals.

Explaining Preference-driven Schedules: the EXPRES Framework

Scheduling is the task of assigning a set of scarce resources distributed over time to a set of agents, who typically have preferences about the assignments they would like to get. Due to the constrained nature of these problems, satisfying all agents' preferences is often infeasible, which might lead to some agents not being happy with the resulting schedule. Providing explanations has been shown to increase satisfaction and trust in solutions produced by AI tools. However, it is particularly challenging to explain solutions that are influenced by and impact on multiple agents. In this paper we introduce the EXPRES framework, which can explain why a given preference was unsatisfied in a given optimal schedule. The EXPRES framework consists of: (i) an explanation generator that, based on a Mixed-Integer Linear Programming model, finds the best set of reasons that can explain an unsatisfied preference; and (ii) an explanation parser, which translates the generated explanations into human interpretable ones. Through simulations, we show that the explanation generator can efficiently scale to large instances. Finally, through a set of user studies within J.P. Morgan, we show that employees preferred the explanations generated by EXPRES over human-generated ones when considering workforce scheduling scenarios.

Multi-tier Automated Planning for Adaptive Behavior (Extended Version)

A planning domain, as any model, is never complete and inevitably makes assumptions on the environment's dynamic. By allowing the specification of just one domain model, the knowledge engineer is only able to make one set of assumptions, and to specify a single objective-goal. Borrowing from work in Software Engineering, we propose a multi-tier framework for planning that allows the specification of different sets of assumptions, and of different corresponding objectives. The framework aims to support the synthesis of adaptive behavior so as to mitigate the intrinsic risk in any planning modeling task. After defining the multi-tier planning task and its solution concept, we show how to solve problem instances by a succinct compilation to a form of non-deterministic planning. In doing so, our technique justifies the applicability of planning with both fair and unfair actions, and the need for more efforts in developing planning systems supporting dual fairness assumptions.