Agreement Technologies refer to open computer systems in which autonomous software agents interact with one another, typically on behalf of humans, in order to come to mutually acceptable agreements. With the advance of AI systems in recent years, it has become apparent that such agreements, in order to be acceptable to the involved parties, must remain aligned with ethical principles and moral values. However, this is notoriously difficult to ensure, especially as different human users (and their software agents) may hold different value systems, i.e. they may differently weigh the importance of individual moral values. Furthermore, it is often hard to specify the precise meaning of a value in a particular context in a computational manner. Methods to estimate value systems based on human-engineered specifications, e.g. based on value surveys, are limited in scale due to the need for intense human moderation. In this article, we propose a novel method to automatically learn value systems from observations and human demonstrations. In particular, we propose a formal model of the value system learning problem, its instantiation to sequential decision-making domains based on multi-objective Markov decision processes, as well as tailored preference-based and inverse reinforcement learning algorithms to infer value grounding functions and value systems. The approach is illustrated and evaluated by two simulated use cases.

Aligning AI systems with human values and the value-based preferences of various stakeholders (their value systems) is key in ethical AI. In value-aware AI systems, decision-making draws upon explicit computational representations of individual values (groundings) and their aggregation into value systems. As these are notoriously difficult to elicit and calibrate manually, value learning approaches aim to automatically derive computational models of an agent’s values and value system from demonstrations of human behaviour. Nonetheless, social science and humanities literature suggest that it is more adequate to conceive the value system of a society as a set of value systems of different groups, rather than as the simple aggregation of individual value systems. Accordingly, here we formalize the problem of learning the value systems of societies and propose a method to address it based on heuristic deep clustering. The method learns socially shared value groundings and a set of diverse value systems representing a given society by observing qualitative value-based preferences from a sample of agents. We evaluate the proposal in a use case with real data about travelling decisions.

In this paper, we study the vehicle routing problem (VRP) for a fleet of cooperative autonomous agricultural robots (agribots) equipped with detachable implements, with the goal of efficiently and sustainably completing agricultural tasks in precision crop farming. State of the art in the area of agribot fleet routing with detachable implements is lacking. Consequently, we propose the Capacitated Agriculture Fleet Vehicle Routing Problem with Implements and Limited Autonomy (CAFVRPILA), designed to optimize the agribot fleet's routes across a set of given agricultural tasks while considering implement capacities, agribot-implement compatibilities, and agribots' limited battery autonomies. A heuristic two-phase decomposition approach is proposed for this problem. Simulation experiments show that minimizing travel distances and costs with CAFVRPILA enhances sustainable farming while maximizing productivity and resource use. The results also demonstrate that synchronizing multiple operations improves efficiency, particularly in larger fleets.

The field of value awareness engineering claims that intelligent software agents should be endowed with a set of capabilities related to human values, enabling them to identify value-aligned outcomes and, ultimately, to choose their behaviour in value-aware manner. In this work we develop an ontology that links many of the models and concepts that have been proposed in relation to computational value awareness, so as to be able to formalize in a common language the various heterogeneous research proposals in the field. Specifically, we illustrate its capability for describing multi-agent systems from the value-awareness engineering perspective through several case studies grounded in concrete approaches from literature. The ontology, implemented in OWL and extended with SWRL rules, is evaluated following scenarios of the NeOn Methodology and is interconnected with relevant ontologies in the Semantic Web.

This paper addresses challenges in agricultural cooperative autonomous fleet routing through the proposition, modeling, and resolution of the Dynamic Vehicle Routing Problem with Fair Profits and Time Windows (DVRP-FPTW). The aim is to dynamically optimize routes for a vehicle fleet serving tasks within assigned time windows, emphasizing fair and efficient solutions. Our DVRP-FPTW accommodates unforeseen events like task modifications or vehicle breakdowns, ensuring adherence to task demand, vehicle capacities, and autonomies. The proposed model incorporates mandatory and optional tasks, including optional ones in operational vehicle routes if not compromising the vehicles’ profits. Including asynchronous and distributed column generation heuristics, the proposed Multi-Agent-based architecture DIMASA for the DVRP-FPTW dynamically adapts to unforeseen events. Systematic Egalitarian social welfare optimization is used to iteratively maximize the profit of the least profitable vehicle, prioritizing fairness across the fleet in light of unforeseen events. This improves upon existing dynamic and multi-period VRP models that rely on prior knowledge of demand changes. Our approach allows vehicle agents to maintain privacy while sharing minimal local data with a fleet coordinator agent. We propose publicly available benchmark instances for both static and dynamic VRP-FPTW. Simulation results demonstrate the effectiveness of our DVRP-FPTW model and our multi-agent system solution approach in coordinating large, dynamically evolving cooperative autonomous fleets fairly and efficiently in close to real-time.

In crowdsourced delivery organizations, where individual vehicles with shared common goals may have conflicting individual interests, the preference is for collaboration over competition, provided it is less costly. However, achieving a balance between the efficiency of individual vehicles and the overall fleet poses a challenge. This paper introduces a novel Vehicle Routing Problem with Fair Profits and Time Windows (VRP-FPTW), which aims to meet customer demand and stringent time windows while maximizing the profit of the worst-off vehicle in the fleet. We propose a centralized and distributed vehicle routing model for this problem, both with quality of solution guarantees. The distributed approach is tailored for multi-agent systems relying on a coordination mechanism where each vehicle modeled as an individually rational agent finds its route autonomously in coordination with a fleet coordinator agent, without sharing its private information. The objective of a vehicle agent is to maximize its own profit while following the fleet's norms and regulations based on shared values. Simulation experiments provide compelling evidence of the robustness and scalability of the proposed distributed approach, showcasing significant enhancements in both solution quality and computational efficiency, particularly when dealing with larger vehicle fleets.

Quality of emergency services depend on the effectiveness of emergency vehicle fleets (e.g., police, fire trucks, and ambulances). Traditionally, these fleets are composed of emergency crews that must attend highly stochastic incident demand within a short target arrival time. Incidents must be attended immediately, which prevents that an a priori computed work break schedule is executed as planned and many crews may be left without a break during prolonged periods of time. Extended focused work periods decrease efficiency with related decline of attention and performance. Therefore, break schedule should be regularly updated on the fly to allow frequent and sufficiently long time for rest. In this paper, we propose a dynamic algorithm for on the fly work and break balancing for crews in emergency fleets. Based on the historical intervention data, the algorithm (re)arranges vehicles’ crews’ work breaks as the time evolves considering individual crews’ preferences. Moreover, it dynamically reallocates stand-by vehicles for improved coverage of a region of interest. We show the performance of the proposed algorithm on two simple functional examples.

This work presents an adaptive sanctioning mechanism that can be applied in open multi-agent systems that are regulated through norms. This mechanism tries to identify the attributes of the environment that have some influence on agents decision making and uses such attributes to define sanctions that may prevent norm violations. Our approach adapts sanctions to particular agents and particular environmental states. The mechanism is deployed by using an infrastructure of institutional agents in charge of adapting and applying sanctions to external agents. The proposal is evaluated empirically in a p2p scenario, comparing it with a traditional sanctioning approach for normative systems.

This paper presents a model for rational self-interested agents, which takes into account the possibility of violating norms. The transgressions take place when the expected rewards obtained with the defection from the norms surpass the expected rewards obtained by being norm-compliant. To develop such model, we employ Markov Decision Processes (MDPs). Our approach consists of representing the reactions for norm violations within the MDPs in such a way that the agent is able to reason about how those violations affect her expected utilities and future options.

In this paper we present ARMEDA II, a project designed to integrate distributed heterogeneous medical and genetic databases in support of genomic medicine. In this project, we have followed a "virtual repository" or VR approach. Although VRs are entities that do not contain any data, but metadata, they give users the perception of being working with local repositories that integrate data from different and remote sources. Our approach is based on two basic operators employed to connect new databases to the system: mapping and unification. The mapping process produces what is called the "virtual conceptual schema" of the newly created VR while the unification process provides tools to create an integrated virtual schema for at least two pre-existing VRs. We tested the current implementation of ARMEDA Il using two tumor databases, one containing information from a hospital and the other containing genetic data associated to the tumor samples. The performance of the system was also evaluated using a pre-created set of 30 queries. For all queries the test yielded promising results since the system successfully retrieved the correct information. The ARMEDA II project is the current version of an ongoing project developed in the framework of an European Commission funded project.

The goal of information retrieval (IR) is to provide models and systems that help users to identify the relevant documents to their information needs. Extensive research has been carried out to develop retrieval methods that solve this goal. These IR techniques range from purely syntax-based, considering only frequencies of words, to more semantics-aware approaches. However, it seems clear that there is no single method that works equally well on all collections and for all queries. Prior work suggests that combining the evidence from multiple retrieval experts can achieve significant improvements in retrieval effectiveness. A common problem of expert combination approaches is the selection of both the experts to be combined and the combination function. In most studies the experts are selected from a rather small set of candidates using some heuristics. Thus, only a reduced number of possible combinations is considered and other possibly better solutions are left out. In this paper we propose the use of genetic algorithms to find a suboptimal combination of experts for a document collection at hand. Our approach automatically determines both the experts to be combined and the parameters of the combination function. Because we learn this combination for each specific document collection, this approach allows us to automatically adjust the IR system to specific user needs. To learn retrieval strategies that generalize well on new queries we propose a fitness function that is based on the statistical significance of the average precision obtained on a set of training queries. We test and evaluate the approach on four classical text collections. The results show that the learned combination strategies perform better than any of the individual methods and that genetic algorithms provide a viable method to learn expert combinations. The experiments also evaluate the use of a semantic indexing approach, the context vector model, in combination with classical word matching techniques.

Abstract In the vector space model for information retrieval, term vectors are pair‐wise orthogonal, that is, terms are assumed to be independent. It is well known that this assumption is too restrictive. In this article, we present our work on an indexing and retrieval method that, based on the vector space model, incorporates term dependencies and thus obtains semantically richer representations of documents. First, we generate term context vectors based on the co‐occurrence of terms in the same documents. These vectors are used to calculate context vectors for documents. We present different techniques for estimating the dependencies among terms. We also define term weights that can be employed in the model. Experimental results on four text collections (MED, CRANFIELD, CISI, and CACM) show that the incorporation of term dependencies in the retrieval process performs statistically significantly better than the classical vector space model with IDF weights. We also show that the degree of semantic matching versus direct word matching that performs best varies on the four collections. We conclude that the model performs well for certain types of queries and, generally, for information tasks with high recall requirements. Therefore, we propose the use of the context vector model in combination with other, direct word‐matching methods.