Logic-based technologies for multi-agent systems: a systematic literature review

“Public institutions have followed, too: for instance, the European Commission already invested 1.5 billion euros in AI for 2018–2020, and many more are planned beyond 2020.1 Furthermore, the “American Artificial Intelligence Initiative: Year One Annual Report”, delivered on February 2020 [334], witnesses the record am…”

“This is where SLR comes at a hand: by providing a technical tool to produce surveys that can be verified and reproduced, since it requires to make the methods used to build the survey explicit and repeatable.”

“So, when including/excludingMAS technologies in/from our survey, we refer to the wellknown definition of agents by Wooldridge and Jennings [356], or, to a basically-equivalent”

“ASPIC+ & Dung-o-Matic (2008) & TOAST (2012) Primary references: [348] Other references describing / exploiting the technology: [39,40] URL: http://tweetyproject.org/downloads/index.html http://toast.arg.tech http://www.arg-tech.org/index.php/projects/dung-o-matic MASperspective TheASPIC+ argumentation framework by Pra…”

“Most of the surveys around, however, while useful and interesting, typically suffer from the authors’ bias [206], in particular when large research topics are addressed: as such, they typically fail in the essential aspects of reproducibility that should characterise any scientific work [291].”

“Table 4 Main technologies excluded from the survey, along with the motivation spanning between (i) “general purpose”, (ii) “no logic”, (iii) “technology not found” Motivation for exclusion Technology General purpose 4QL [243], CCALC [163], CDF-Rules [164], Clingo [292], DR-Prolog [37], Inter4QL [199], InterProlog [62]…”

“Table 4 Main technologies excluded from the survey, along with the motivation spanning between (i) “general purpose”, (ii) “no logic”, (iii) “technology not found” Motivation for exclusion Technology General purpose 4QL [243], CCALC [163], CDF-Rules [164], Clingo [292], DR-Prolog [37], Inter4QL [199], InterProlog [62]…”

“Multi-agent systems facilitated coordination among distributed entities, exemplified by auction-based resource allocation in supply chain management [16–18].”

“…have long understood that building effective AI agents requires integrating multiple disciplines, including formal logic, cognitive science, machine learning, and software engineering—so that these agents can not only make autonomous decisions but also behave reliably and explainably [114].”

“This is, for instance, what happened with the Multi-Agent System community, where Prolog and LP have been extensively exploited in the last decades as the technological or conceptual basis for tens of agent-oriented technologies (Calegari et al., 2021).”

“4) Symbolic Artificial Intelligence: Related to our focus on logical relationships, learning logical relationships to represent expert knowledge has been traditionally embodied by the field of expert systems [29], multi-agent systems [30], and knowledge query manipulation language [31].”

“, those that respond to stimuli without complex symbolic reasoning to reason about future actions) to deliberative ones (those that maintain a symbolic world model for reasoning about plans and decision making) [29].”

“Calegari et al [3] presented a systematic review of MAS but mainly focused on logical technologies.”

“Multi-agent systems (MAS) offer a promising alternative to address these limitations through task decomposition and collaboration (Calegari et al. 2021; Cardoso and Ferrando 2021).”

“As shown by recent surveys [7], it is quite common for the MAS community to represent knowledge and agents’ behaviours symbolically both at the conceptual and technological level—often relying upon computational logic [8].”

“AI [2,3]: CL has enabled the development of the former in the past, and it is now pushing the latter towards interpretability and explainability.”

“These systems utilize various methodologies, including logical inference [38], game theory [39], bio-inspired algorithmic search [40], and machine learning [41].”

“This deliberative reasoning process makes BDI models suitable for applications requiring goal-oriented behavior and planning, such as sophisticated financial advisory systems or compliance monitoring agents that need to reason about regulatory obligations [90–93].”

“The complex and opaque nature of LLMs makes it challenging to directly analyze and improve their behavior and outputs (Calegari et al., 2020).”

“In the study of MAGs, the relationship between agents can be described as a network topology [32].”

“Early agent-based systems (Calegari et al., 2021; Cardoso and Ferrando, 2021) focused on coordination and had limited problem-solving capabilities often within the bounds of predefined and narrow operational contexts.”

“These agents possess capabilities for environmental perception (Leal Filho et al., 2022), rea - soning, and control (Calegari et al., 2021), allowing them to actively collect and analyze environmental information and take corresponding actions (Seng et al., 2022).”

“(Calegari et al., 2020) recently reviewed logic-based technologies for MAS and provide a nice overview of the available, and actively developed, agent technologies including means of agent verification.”

“Explainability and transparency are very critical as well [2].”

“Prior to the widespread availability of advanced foundation models, intelligent systems operated within highly constrained parameters, relying primarily on rule-based logic, symbolic reasoning, and predetermined response patterns [4], [5].”

“Because agents in the BDI model can make choices regarding what they want and what they believe, this model has received the most focus and has become very popular in the construction of autonomous agents in both theoretical and applied robotics and distributed systems [1].”

“The collaborative efforts of multiple intelligent entities leverage their respective strengths to collectively accomplish complex tasks, thereby improving production efficiency and system performance in various domains[2].”

“IoA is a novel technique that emerged with an aim to mitigate the communication deficiency in terms of reasoning and social abilities found in IoT scenarios, with the help of software agents (Calegari et al., 2021).”

“From a recent review of logic-based approaches for multi-agent systems by (Calegari et al., 2020), it is clear that model-checking has retained its position as the primary tool for practical verification of multi-agent systems.”

“In multi-agent interactions, the use of logic-based systems can be supported in the literature to enforce contract, organisational norms, trace distributed decisions, and managing conflicts [2].”

“First, an MAS enables the representation of heterogeneous compu‑ tational intelligence across different agent types, reflecting the diverse cognitive models that shape discourse contributions [41].”

“In the end, we forward the interested readers to the original SLR for the full details [7].”

“agent to sense its environment and trigger its behavior when a change is detected [7].”

“Depending on whether MASs were exploited to study systems’ complexity, or to tame it, many solutions have been proposed to either use MAS abstractions to simulate system behaviour [5] or for designing and deploying real-world systems [6].”

“Hence, we build upon the recent work by Calegari et al. [24], which surveys the state-of-the-art of logic-based agent-oriented technologies, and we select the ones aimed to support general-purpose BDI agents programming.”

“However, many cognitive agent models and their implementation platforms (see [10] and [36]) are often resource heavy for real-time applications as demanding as eye-tracking.”

“…BDI proposal, these notions have significantly evolved in computational logics, and many logic-based agent-oriented languages and frameworks to specify agents and multi-agent systems have been defined (for a survey of these languages and architectures, the reader may refer to [46], [47], [48]).”

“Instead of following hardcoded procedures (as in traditional rule-based automation [6]), agentic AI systems can perceive context, reason about goals, generate multi-step plans, execute actions through external tools, and revise decisions using feedback-driven adaptation [7].”

“Building upon the foundational characteristics, scholars such as Nami & Malekpour (2008) and Calegari et al. (2020) propose additional dimensions that define how virtual organizations operate in practice.”

“It is based on three different artificial intelligent tools specifically: • The fuzzy Ontologies which are defined as a quintuple OF = {I, C, R, F, A} where I is a set of individuals, C is a set of concepts, R is a set of relations, F is a set of fuzzy relations and A is the set of axioms [1-4].”

“This because a logic basis provides (at least potentially) verifiability and explainability, and thus trustworthiness [7] 3 .”

“Several agent-oriented programming languages and systems exist, many of them based upon computational logic (cf., e.g., [4, 5, 14] for recent surveys on such languages).”

“MAS have been at the core of the design of intelligent systems since their very beginning, and their long-term connection with logic-based technologies, which characterised their early days, might open new ways to engineer explainable intelligent systems [2].”

“These agents are capable of sensing, reasoning and acting either independently or cooperatively [12].”

“Such constraints and task specifications can be encoded through logic formulation and piece-wise nonlinear functions [14].”

“On mainstream programming platforms, (such as the Java Virtual Machine (JVM) [12], used for the implementation of several BDI frameworks), computational autonomy may be achieved by mapping each agent onto a control-flow-related primitive: a thread, a processes, or an event loop.”

“…paper) [9], 2012 (4 paper) [3], [4], [10], [12], 2013 (2 paper) [7], [8], 2014 (4 paper) [5], [11], [14], [19], 2015 (1 paper) [6], 2020 (1 paper) [23], 2021 (3 paper) [20], [22], [21] 2022 (2 paper) [22], [29] ) [20], agriculture (2 paper) [9], [19], tourism (1 paper) [3], education (1 paper)…”

“Jade [11] x x x x Jason [14] x x x x Agentsheets [15] x x x x Cormas [16] x x x x x Ingenias [17] x x x x x Repast [19] x x x x Madkit [20] x x Netlogo [21] x x x Simpack [22] Starlogo [23] x x x x Framework de SySAge x x x x x x x x”

“Symbolic agents, also known as deliberative or logical agents, operate based on explicit representations of the rules of knowledge and reasoning (de Silva et al., 2020; Calegari et al., 2021).”

“Recent advances in multi-agent systems (Cardoso and Ferrando 2021; Calegari et al. 2021) provide a promising foundation for such role-specialized co-ordination.”

“Another class is given by declarative approaches [7, 15], often logic-based, that specify what are the goals (rather than how they should be pursued).”

“As mentioned by [5, 6], logical approaches, especially fuzzy logic, play a prominent role in both rule-based control and uncertainty handling, despite the widely used basic form.”

“While logic programming has been applied to reactive control (IndiGolog [3] and LTL synthesis [7]), surveys of logic-based MAS technologies [2] confirm most approaches operate at high abstraction levels, leaving low-level control to imperative code.”

“Recently, along with the rapid development of modern subsymbolic approaches to AI, symbolic approaches of classical AI are gaining momentum again as more and more researchers exploit their potential to make AI more understandable, explain-able and therefore trustworthy [3, 4].”

“Multi-agent systems (MAS), as a key technology for collaborative warfare, enable multiple agents to achieve efficient cooperation through means such as information sharing, task allocation, and resource scheduling [1-9], thereby optimizing decision-making and execution.”

“The influence of the belief-desire-intention (BDI) and related reasoning formalisms on logic-based technologies for MAS and the needs for tools and frameworks to build production-grade agent systems have also been surveyed [10, 12].”

“According to [7,8], fractal network processes have a long-term dependence, expressed in an infinite correlation interval.”

“As a popular and mature research field, multi-agent system (MAS) makes arrangements for a group of agents to accomplish a common problem or task in a centralized or distributed manner [3,4].”

“The emergence of hybrid systems [34], [35], [36], which blend ontological knowledge [37], collaborative filtering, and multi-agent frameworks, represents a logical progression, but frequently at the cost of greater system complexity.”