Hierarchical Task Network (HTN) planners generate plans using a decomposition process with extra domain knowledge to guide search towards a planning task. Domain experts develop such domain knowledge through recipes of how to decompose higher level tasks, specifying which tasks can be decomposed and under what conditions. In most realistic domains, such recipes contain recursions, i.e., tasks that can be decomposed into other tasks that contain the original task. Such domains require that either the domain expert tailor such domain knowledge to the specific HTN planning algorithm, or an algorithm that can search efficiently using such domain knowledge. By leveraging a three-stage compiler design we can easily support more language descriptions and preprocessing optimizations that when chained can greatly improve runtime efficiency in such domains. In this paper we evaluate such optimizations with the HyperTensioN HTN planner, winner of the HTN IPC 2020 total-order track.

The quality and efficiency of road maintenance greatly impact on transportation performance. There is an urgent need to automate the process of decision-making for road maintenance, especially for reactive maintenance. However, automating this process is challenging because (1) the expert knowledge required in this process is implicit and difficult to formalize and (2) the defect reports from frontline inspectors are in unstructured human language. This paper proposes a method based on natural language processing (NLP) and knowledge graph reasoning. This method combines the semantic understanding capability of NLP with the structural reasoning capability of the knowledge graph, enabling nuanced and logically rigorous road maintenance decisions. Real-world highway defect reports are used for validation. The experimental results demonstrate a significant improvement of the proposed model over baseline models, with up to 13.5% higher accuracy on the prediction of key tasks. This approach enables a traceable decision-making process, avoiding a black-box model.

In annotating real-life large, noisy and domain-specific images for digitising infrastructure, substantial human effort persists despite past advancements. This research provides practical and interpretable scores for human annotators, enabling flexible annotation strategies, improving automation and reducing the effort required to create and correct image labels. The authors present the CRAAC solution: Consistency Regularised Active learning and Automatic Corrections, which builds on Mask R-CNN with three additional modules: consistency regularisation, scoring modules for active learning and automatic corrections. Experiments on our pavement image dataset, recorded with a low silhouette score of 0.146 and qualitative annotation inconsistencies, reduce the human effort of mouse clicks by 5-11% and improve the quality metrics of mAP and AR by approx. 40% from the original Mask R-CNN. The automatic correction further reduces the performance variation.

Abstract Hierarchical Task Network (HTN) planners generate plans using a decomposition process with extra domain knowledge to guide search towards a planning task. Domain experts develop such domain knowledge through recipes of how to decompose higher level tasks, specifying which tasks can be decomposed and under what conditions. In most realistic domains, such recipes contain recursions, i.e., tasks that can be decomposed into other tasks that contain the original task. Such domains require that either the domain expert tailor such domain knowledge to the specific HTN planning algorithm, or an algorithm that can search efficiently using such domain knowledge. By leveraging a three-stage compiler design we can easily support more language descriptions and preprocessing optimizations that when chained can greatly improve runtime efficiency in such domains. In this paper we evaluate such optimizations with the HyperTensioN HTN planner, winner of the HTN IPC 2020 total-order track.

Road maintenance involves manually classifying a large volume of textual data necessary for downstream applications such as raising a maintenance job order. Automation can not only bring significant time and cost savings, but it can also facilitate digitalization efforts like the Road Digital Twin (DT). However, as is the case with many Architecture, Engineering and Construction (AEC) applications, annotated data availability is low, which demands exploration of specialized techniques for resource-constrained settings that have not been focused on in engineering. This work bridges this gap by proposing a data-efficient similarity-based text classifier that aims at effectively utilizing existing domain knowledge and pre-training knowledge of Large Language Models (LLMs) to enable rapid domain adaptation. It reformulates text classification as a similarity comparison task, using semantics directly as a classification criterion. Through a case study on classifying road inspection comments, the proposed classifier outperformed both traditionally fine-tuned and few-shot learning approaches. It attained an f 1 score of 0.46 with just one example per class, equivalent to the value for Sentence Transformer Fine-Tuning (SetFit) with 4 examples and Llama3 with 10. Additionally, it is able to keep up with traditional fine-tuning methods when trained with more than 300,000 total examples, achieving an a c c u r a c y of more than 95% and f 1 of around 0.9. These results indicate that the proposal is competitive against traditionally fine-tuned and few-shot models across all levels of data availability. This versatility significantly elevates the feasibility of deploying an automated text classification pipeline in a complex engineering field like road maintenance.

The CAMHighways dataset is presented, built from mobile mapping data that surveyed over 40 km of UK Highways. The dataset consists of textured meshes for road assets (including the pavement, traffic signs, and road furniture), segmented and classified point clouds, orthomosaics generated from pavement images, defect label annotations and shapefiles, and ground penetrating radar point clouds. All modalities are georeferenced and can be integrated into game engines and/or GIS software. The main aim of this work is to facilitate and automate the building of a Digital Twin (DT), a digital representation of the highway, in order to streamline inspection and maintenance through virtual reality, robotics simulation, and DT- and AI-driven data analysis. It also serves as a valuable source for other applications, such as training semantic scene understanding and defect detection algorithms. This paper introduces the dataset and outlines the data preparation process, including novel automation methods developed for this purpose, as well as integration guidelines and possible applications. • The new CAMHighways dataset is presented, spanning 42.8 km of UK highways. • Mobile mapping data is prepared for building a road DT for inspection & maintenance. • 3D meshes, point clouds, pavement orthomosaics, labels, and GPR data are included. • Several aspects of the DT generation process are automated. • The dataset is integrated into game engines and GIS software.

Significant funds have been allocated to maintain road networks each year in developed countries. Performance prediction is crucial for pavement management systems to adjust working plans and budget allocation. As a Bayesian nonparametric method, Gaussian process regression (GPR) is powerful in predicting nonlinear time series and quantifying uncertainty. However, it remains computationally intensive and fails to adapt to the time-varying characteristics. To address such issues, a dynamic GPR model is proposed for probabilistic prediction of the International Roughness Index (IRI) for flexible pavements. A moving window strategy is developed to substantially shrink the size of training data, which effectively alleviates computational cost and thus leads to a dynamic GPR. A genetic algorithm is then adopted to determine the optimal window size by considering the trade-off between computational efficiency and accuracy. A dataset acquired from Long-Term Pavement Performance (LTPP) is used to demonstrate the feasibility of the dynamic GPR. Its performance is compared to traditional GPR as well as dynamic and static Bayesian linear regression (BLR) models. The comparison results indicate that the proposed dynamic GPR can increase the accuracy by 0.86, 1.52, and 2.27 times for dynamic BLR, static GPR, and static BLR, respectively. It exhibits the best results in terms of accuracy and uncertainty metrics due to its nonlinear modelling and time-varying ability.

Agent programming languages have proved useful for formally modelling implemented systems such as PRS and JACK, and for reasoning about their behaviour. Over the past decades, many agent programming languages and extensions have been developed. A key feature in some of them is their support for the specification of ‘concurrent’ actions and programs. However, their notion of concurrency is still limited, as it amounts to a nondeterministic choice between (sequential) action interleavings. Thus, the notion does not represent ‘true concurrency’, which can more naturally exploit multi-core computers and multi-robot manufacturing cells. This paper provides a true concurrency operational semantics for a BDI agent programming language, allowing actions to overlap in execution. We prove key properties of the semantics, relating to true concurrency and to its link with interleaving.

The BDI model forms the basis of much of the research on symbolic models of agency and agent-oriented software engineering. While many variants of the basic BDI model have been proposed in the literature, there has been no systematic review of research on BDI agent architectures in over 10 years. In this paper, we survey the main approaches to each component of the BDI architecture, how these have been realised in agent programming languages, and discuss the trade-offs inherent in each approach.

This paper concerns the development of low-cost solutions to address challenges in digital manufacturing (DM). Service Oriented Architectures (SOAs) are a promising approach for addressing the requirements of a low-cost DM architecture. Interaction between services in a SOA is facilitated by a connectivity technology, i.e., a framework for interoperable data exchange between heterogeneous participants. We review a variety of connectivity technologies according to their suitability for use in an SME manufacturer's production environment, and we assess how they have been integrated into past architectures. We then provide insights into an incremental and modular architecture for manufacturing SMEs.

The Procedural Reasoning System (PRS) is arguably the first implementation of the Belief--Desire--Intention (BDI) approach to agent programming. PRS remains extremely influential, directly or indirectly inspiring the development of subsequent BDI agent programming languages. However, perhaps surprisingly given its centrality in the BDI paradigm, PRS lacks a formal operational semantics, making it difficult to determine its expressive power relative to other agent programming languages. This paper takes a first step towards closing this gap, by giving a formal semantics for a significant fragment of PRS. We prove key properties of the semantics relating to PRS-specific programming constructs, and show that even the fragment of PRS we consider is strictly more expressive than the plan constructs found in typical BDI languages.

Determining the most appropriate means of producing a given product, i.e., which manufacturing and assembly tasks need to be performed in which order and how, is termed process planning. In process planning, abstract manufacturing tasks in a process recipe are matched to available manufacturing resources, e.g., CNC machines and robots, to give an executable process plan. A process plan controller then delegates each operation in the plan to specific manufacturing resources. In this paper we present an approach to the automated computation of process plans and process plan controllers. We extend previous work to support both non-deterministic (i.e., partially controllable) resources, and to allow operations to be performed in parallel on the same part. We show how implicit fairness assumptions can be captured in this setting, and how this impacts the definition of process plans.

There is a rising demand for customised products with a high degree of complexity. To meet these demands, manufacturing lines are increasingly becoming autonomous, networked, and intelligent, with production lines being virtualised into a manufacturing cloud, and advertised either internally to a company, or externally in a public cloud. In this paper, we present a novel approach to two key problems in such future manufacturing systems: the realisability problem (whether a product can be manufactured by a set of manufacturing resources) and the control problem (how a particular product should be manufactured). We show how both production recipes specifying the steps necessary to manufacture a particular product, and manufacturing resources and their topology can be formalised as labelled transition systems, and define a novel simulation relation which captures what it means for a recipe to be realisable on a production topology. We show how a controller that can orchestrate the resources in order to manufacture the product on the topology can be extracted from the simulation relation, and give an algorithm to compute a simulation relation and a controller.

Hierarchically structured agent plans are important for efficient planning and acting, and they also serve (among other things) to produce “richer” classical plans, composed not just of a sequence of primitive actions, but also “abstract” ones representing the supplied hierarchies. A crucial step for this and other approaches is deriving precondition and effect “summaries” from a given plan hierarchy. This paper provides mechanisms to do this for more pragmatic and conventional hierarchies than in the past. To this end, we formally define the notion of a precondition and an effect for a hierarchical plan; we present data structures and algorithms for automatically deriving this information; and we analyse the properties of the presented algorithms. We conclude the paper by detailing how our algorithms may be used together with a classical planner in order to obtain abstract plans.

For intelligent Human-Robot Interaction (HRI), a robot should be equipped with some core reasoning capabilities such as perspective taking, effort analysis, and affordance analysis. This paper starts to explore how a robot equipped with such reasoning abilities could be evaluated. To this end, inspired by the Turing test, we design a game involving a human-robot competition scenario. Interestingly, the participants' subjective feedback, which tended to compare the robot's abilities with their own, points toward potential criteria for developing benchmark scenarios and evaluation matrices.

HTN planners have generally relied on specialised languages for domain and problem representations. To facilitate adoption by other communities such as robotics, however, and integration with real world applications written in standard programming languages, we need HTN planners that are based on more familiar concepts from structured programming, and that come ready with features supporting integration. In this paper, we demonstrate how the HATP (Hierarchical Agent-based Task Planner) HTN planner offers such “syntactic sugar” and some of these features. Moreover, since it has a conceptually distinct syntax compared to traditional HTN planners, we also develop a formalism to unambiguously capture HATP's syntax and an important subset of its semantics, which we then use to compare against the formalism of a well understood family of HTN planners and to show that the former is sound. Finally, we demonstrate that despite quite possibly using “heavier” data structures to naturally capture HATP's syntax/semantics, and thereby facilitate extensions to HATP and integration with other applications, the implementation still performs acceptably.

Bridging the gap between symbolic and geometric planning has received much attention in recent years. An important issue in some of the works that combine the two approaches is finding the right balance between backtracking at the symbolic level versus at the geometric planning level. We present in this work a new approach to interleaved backtracking, where the symbolic planner backtracks to try alternative action branches that naturally map to different geometric solutions. This eliminates the need to “protect” certain symbolic conditions when backtracking at the geometric level, and addresses a completeness issue in our previous approach to interleaved backtracking. We discuss a concrete, non-trivial symbolic-geometric planning example in the context of Human-Robot Interaction, a full implementation of the combined planning technique, and an evaluation of performance as well as the effect of increasing the symbolic-action branching factor.

While symbolic planners work with an abstract representation of the real world, allowing plans to be constructed relatively quickly, geometric planning - although more computationally complex - is essential for building symbolic plans that actually work in the real world. To combine the two types of systems, we present in this paper a meaningful interface, and insights into a methodology for developing interwoven symbolic-geometric domains. We concretely present this “link” between the two approaches with algorithms and data structures that amount to an intermediate layer that coordinates symbolic-geometric planning. Since both planners are capable of “backtracking” at their own levels, we also investigate the issue of how to interleave their backtracking, which we do in the context of the algorithms that form the link. Finally, we present a prototype implementation of the combined system on a PR2 robot.