Association for the Advancement of Artificial Intelligence: AAAI Publications
Not a member yet
26155 research outputs found
Sort by
Finding All Optimal Solutions in Multi-Agent Path Finding
No full textThe Multi-Agent Path Finding problem (MAPF) aims to find conflict-free paths for a group of agents, leading each agent to its respective goal. MAPF is applicable in navigating autonomous robots and vehicles to their destination. In this paper, we study the requirement of finding all optimal solutions in MAPF. We discuss the representation of all optimal solutions, propose four algorithms for finding them, and perform an extensive empirical evaluation of the proposed algorithms
Lazy Heuristic Search for Solving POMDPs with Expensive-to-Compute Belief Transitions
No full textHeuristic search solvers like RTDP-Bel and LAO* have proven effective for computing optimal and bounded sub-optimal solutions for Partially Observable Markov Decision Processes (POMDPs), which are typically formulated as belief MDPs. A belief represents a probability distribution over possible system states. Given a parent belief and an action, computing belief state transitions involves Bayesian updates that combine the transition and observation models of the POMDP to determine successor beliefs and their transition probabilities. However, there is a class of problems, specifically in robotics, where computing these transitions can be prohibitively expensive due to costly physics simulations, raycasting, or expensive collision checks required by the underlying transition and observation models, leading to long planning times. To address this challenge, we propose Lazy RTDP-Bel and Lazy LAO*, which defer computing expensive belief state transitions by leveraging Q-value estimation, significantly reducing planning time. These algorithms are specific instantiations of the broader idea of lazy search for POMDPs. We demonstrate the superior performance of the proposed lazy planners in domains such as contact-rich manipulation for pose estimation, outdoor navigation in rough terrain, and indoor navigation with a 1-D Lidar sensor. Additionally, we discuss practical Q-value estimation techniques for commonly encountered problem classes that our lazy planners can leverage. Our results show that lazy heuristic search methods dramatically improve planning speed by postponing expensive belief transition evaluations while maintaining solution quality
Learning Heuristic Functions with Graph Neural Networks for Numeric Planning (Extended Abstract)
No full textIn this paper, we investigate the application of heuristics based on Graph Neural Networks (GNNs) to lifted numeric planning problems, an area that has been relatively unexplored. Building upon the GNN approach for learning general policies proposed by Staahlberg et al., we extend the architecture to make it sensitive to the numeric components inherent in the planning problems we address. We achieve this by observing that, although the state space of a numeric planning problem is infinite, the finite subgoal structure of the problem can be incorporated into the architecture, allowing for the construction of only a finite number of nodes. Instead of learning general policies, we train our models to function as a heuristic within a best-first search algorithm. We explore various configurations of this architecture and demonstrate that the resulting heuristics are highly informative and, in certain domains, offer a better trade-off between guidance and computational cost compared to other inductive and deductive heuristics
Bidirectional Heuristic Search in Longest Path Problems (Extended Abstract)
No full textBidirectional heuristic search has the potential to decrease search time in combinatorial search problems amenable to backward search. To date, bidirectional search has been limited to minimization or shortest path problems. This paper extends the notion of bidirectional heuristic search to (constrained) longest path problems, which turns out to be non-trivial due to the path necessarily being part of the state and the inapplicability of standard bidirectional heuristic search techniques such as meet-in-the-middle (MM) and BAE*.
We present a basic bidirectional heuristic search for longest simple path (LSP) in undirected graphs, and prove its correctness. We then suggest several refinements and optimizations, as well as a generalization to other types of longest path problems Coil-in-a-box (CIB).
Empirical evaluation shows that, as with many forms of bidirectional search, sometimes unidirectional search wins, but for a sizable chunk of problem instance types, bidirectional search performs better by expanding fewer nodes and achieves a shorter runtime despite the increased overhead per expansion
LSRP*: Scalable and Anytime Planning for Multi-Agent Path Finding with Asynchronous Actions (Extended Abstract)
No full textMulti-Agent Path Finding (MAPF) seeks collision-free paths for multiple agents from their respective starting locations to their respective goal locations while minimizing path costs. Although many MAPF algorithms were developed, most of them rely on a common assumption on synchronized actions, where the actions of all agents start at the same time and always take a time unit. This assumption may limit use of MAPF planners in practice. To get rid of this assumption, recently, an algorithm called Loosely Synchronized Rule-Based Planning (LSRP) is proposed, which can find sub-optimal solutions for many agents. However, LSRP often finds poor quality solutions due to its unbounded sub-optimality. This paper develops a new anytime planner called LSRP* that can keep improving solution quality after the initial solution is obtained until the runtime budget depletes. We analyze the properties of LSPR* and test it against several baselines with up to 1000 agents in various maps. LSRP* can handle up to 25% more agents than LSRP and can reduce up to 40% of the solution cost found by LSRP
Frontmatter
No full textThis frontmatter introduces the proceedings of the Eighteenth International Symposium on Combinatorial Search (SoCS 2025), held from August 12–15, 2025, in Scotland, United Kingdom. It includes a preface by the conference co-chairs—Maxim Likhachev, Hana Rudová, and Enrico Scala—along with details on the Best Paper Awards, the organizing and program committees, and the sponsors of this edition
Blockchain-Enhanced Machine Learning for Dynamic Routing and Secure Communications in Autonomous Vehicle Networks
No full textThe advent of autonomous vehicles (AVs) marks a significant milestone in urban transportation, promising to enhance safety, reduce congestion, and improve environmental sustainability. However, deploying AVs on a mass scale comes with critical challenges related to secure and efficient vehicular communication. This research work proposes a novel framework that combines the security features of blockchain technology with the adaptive capabilities of machine learning (ML) to address these major challenges. Integrating a blockchain-based protocol ensures tamper-proof and transparent communication within AV networks, protecting against a wide array of cyber threats. Concurrently, ML algorithms are employed to optimize real-time routing decisions based on comprehensive traffic data and environmental conditions. Through simulation in realistic urban scenarios, our framework demonstrates a significant improvement in communication security and routing efficiency, indicating a promising avenue for achieving scalable and reliable AV networks. Operational cost assessments further reveal the economic viability of the proposed model, underscoring its potential to deliver long-term savings through enhanced efficiency and reduced human intervention. Thus an efficient solution in terms of security, dynamic routing, and scalability with respect to traditional models
Leveraging Isomorphism for Developing AI Focused Innovation Systems
No full textInnovation systems have long been a valuable tool for understanding innovation, using a systems perspective to explain its emergence. While useful, the specific type of innovation these systems create is often overlooked. Given AI's importance, it warrants focused attention. Using the innovation ecosystem model, the following agents can be considered: universities, industry, government, infrastructure, financing, research institutions, and entrepreneurship, aiming for radical or incremental AI innovation.
To foster AI effectively, it is crucial to focus on designing a system that creates and competes in such a field. Leveraging isomorphic properties, the AI-innovation system can attract new ventures and companies, legitimizing them through aligning themselves to the innovation system’s theme, and therefore achieving specialization. Following this, growth can be achieved via a related expansion strategy, targeting AI's complementary industries.
By positioning the government as the innovation system’s articulator, resources can be allocated to maximize agent conditions and promote AI-related innovation, by designing public policies for each agent under the umbrella of an AI-innovation system development.
In conclusion, an AI-focused innovation system is a powerful way for regions to develop AI capabilities. It not only fosters AI development, but can also grow to use AI to develop other industries, creating a positive feedback loop. This approach allows for a specialized system that can grow and expand, creating a virtuous cycle of innovation
Seeing Safety: Computer Vision for Real-Time PPE Monitoring in the Middle East Construction Sector
No full textWorkplace safety is still a major concern worldwide, especially in high-risk settings like construction sites where manual PPE monitoring is insufficient. As employers want to move towards a more efficient and safer worksite, context awareness is becoming increasingly important in the construction industry. Besides, conventional PPE compliance checks can be ineffective, costly and prone to errors. This paper presents a novel approach to enhance workplace safety in construction environments through the development of an automated Personal Protective Equipment (PPE) detection system utilizing advanced computer vision techniques. YOLOv5 object detection models were used in this study to address the limitations of manual PPE compliance monitoring by achieving real-time detection under diverse environmental conditions, such as variable lighting and occlusion. The methodology encompasses comprehensive dataset preparation, annotation, and model training, achieving a mean Average Pre-cision (mAP) exceeding 70%, with YOLOv5m(a) attain-ing an [email protected] of 0.841. This research contributes to reducing workplace hazards by improving monitoring efficiency and lays a foundation for future advancements in industrial safety systems
Strategic Integration of Artificial Intelligence in Healthcare: Theoretical Frameworks, Adoption, Enablers, and Barriers — A Scoping Review
No full textHealthcare organizations increasingly leverage Artificial Intelligence (AI) to enhance clinical decision-making, operational efficiency, and strategic positioning. However, existing research on human-AI collaboration in healthcare has not fully explored how strategic management theories intersect with us-er-centered design principles, interpretability, and ethical considerations essential for building reliable AI partners. This scoping review aimed to (i) map the current landscape of AI-enabled knowledge sharing in healthcare organizations; (ii) identify theoretical frameworks, including human-in-the-loop and user acceptance models; (iii) examine both organizational and user-level enablers and barriers, such as ethical concerns, transparency, and digital literacy; and (iv) propose an integrated strategic management perspective for more robust, inclusive, and ethically grounded AI adoption. Eligible studies addressed AI-enabled interventions (e.g., ma-chine learning, deep learning, natural language processing) in diverse healthcare settings (resource-limited, public, and private institutions), with no date restrictions. Only English-language publications were included. A comprehensive search across SCOPUS, PubMed, and EB-SCOhost-Web of Science yielded 327 articles, with 297 screened for relevance using Covidence software. Five studies met eligibility criteria and were thematically synthesized using NVivo. Analytical categories spanned organizational readiness, stakeholder acceptance, digital/ethical infrastructure, and collaborative AI design features. Key facilitators of successful AI adoption included leadership endorsement, specialized training, robust institutional sup-port, and perceived utility of AI solutions. Frequently em-ployed frameworks (Technology Acceptance Model, Unified Theory of Acceptance and Use of Technology, Diffusion of Innovations, Sociotechnical Systems) addressed individual-level behaviors but rarely accounted for deeper strategic management factors. Ethical concerns related to patient privacy, data security, and algorithmic bias underscored the need for transparent and explainable AI, particularly in high-stakes healthcare contexts. Current research on healthcare AI adoption predominantly emphasizes user acceptance without fully integrating strategic management and collaborative design principles. Future inquiry should incorporate human-in-the-loop approaches, interpretability methodologies, and strategic management theories to enhance AI sustainability and transparency, foster trust, and safeguard ethical standards. By coupling these dimensions with organizational strategy, healthcare systems can more effectively harness AI for sustainable competitive advantage, elevated clinical outcomes, and responsible innovation