• Agent-Based RAG Architecture with Event-Driven Data for Explainable Robots. • A scalable, efficient, decoupled, and adaptive solution for robot explanations. • Review of XAI in robotics, LLMs in XARs, and LLM evaluation approaches. • Empirical results showing accurate robot explanations in various scenarios. • Future work on grounded explanations for autonomous agents. Effective communication in Human-Robot Interaction (HRI) is essential for building trust in autonomous systems. Robotic agents must provide clear, factual explanations that help non-expert users understand their decisions and actions, thereby promoting transparency and acceptance. However, generating structured, contextually relevant, and well-reasoned explanations remains a significant challenge, especially in dynamic environments, where rapidly changing circumstances make it difficult to ensure accuracy, consistency, and timeliness. To address this problem, we propose an architecture that generates natural language explanations grounded in accountable agent data. A distributed event streaming platform captures and processes high-volume system data in real time, which is then used by an agent-based Retrieval-Augmented Generation (RAG) approach to produce accurate and context-aware explanations. By decoupling explanation generation from the robot’s onboard resources, the architecture enables scalable and efficient reasoning while minimizing computational overhead. Experiments on robotic navigation tasks demonstrate that the system achieves high performance across quantitative metrics, with Context Recall consistently above 85%, Faithfulness over 78%, and Semantic Similarity near 96%. Criteria-based evaluations show high levels of Correctness (97.5-100%), with scores for Understandability, Informativeness, and Coherence exceeding 4.0 on a 5-point scale. These results provide strong evidence that integrating curated real-time data with agent-based reasoning enhances the interpretability, reliability, and user trust in autonomous robot behavior.

Self-localization is a fundamental capability that mobile robot navigation systems integrate to move from one point to another using a map. Thus, any enhancement in localization accuracy is crucial to perform delicate dexterity tasks. This paper describes a new localization algorithm that maintains several populations of particles using the Monte Carlo Localization (MCL) algorithm, always choosing the best one as the system's output. As novelties, our work includes a multi-scale map-matching algorithm to create new MCL populations and a metric to determine the most reliable. It also contributes the state of the art implementations, enhancing recovery times from erroneous estimates or unknown initial positions. The proposed method is evaluated in ROS2 in a module fully integrated with Nav2 and compared with the current state-of-the-art Adaptive AMCL solution, obtaining good accuracy/recovery times.

Cyberattacks are a growing concern for companies and public administrations. The literature shows that analyzing network-layer traffic can detect intrusion attempts. However, such detection usually implies studying every datagram in a computer network. Therefore, routers routing a significant volume of network traffic do not perform an in-depth analysis of every packet. Instead, they analyze traffic patterns based on network flows. However, even gathering and analyzing flow data has a high-computational cost, and therefore routers usually apply a sampling rate to generate flow data. Adjusting the sampling rate is a tricky problem. If the sampling rate is low, much information is lost and some cyberattacks may be neglected, but if the sampling rate is high, routers cannot deal with it. This paper tries to characterize the influence of this parameter in different detection methods based on machine learning. To do so, we trained and tested malicious-traffic detection models using synthetic flow data gathered with several sampling rates. Then, we double-check the above models with flow data from the public BoT-IoT dataset and with actual flow data collected on RedCAYLE, the Castilla y León regional academic network.

Motion Capture systems are crucial in many fields, and Mobile Robotics is one of them. This paper describes an Open Source robotic framework to standardize the use of motion capture systems called MOCAP4ROS2. This framework features a layered architecture that allows building applications that use Motion Capture systems regardless of the specific system model/vendor. The challenges are technical and social: on the one hand, resolving synchronization and representation issues; on the other hand, involving the community to reach a consensus on the necessary interfaces. MOCAP4ROS2 has been implemented in ROS2 and already has drivers (we understand a driver for MOCA4ROS2 as a ROS2 node that publish the MOCAP system information) for today’s main commercial systems.

Abstract Understanding why a robot's behaviour was triggered is a growing concern to get human‐acceptable social robots. Every action, expected and unexpected, should be able to be explained and audited. The formal model proposed here deals with different information levels, from low‐level data, such as sensors' data logging; to high‐level data that provide an explanation of the robot's behaviour. This study examines the impact on the robot system of a custom log engine based on a custom ROS logging node and investigates pros and cons when used together with a NoSQL database locally and in a cloud environment. Results allow to characterize these alternatives and explore the best strategy for offering a fully log‐based accountability engine that maximizes the mapping between robot behaviour and robot logs.

Facing human activity-aware navigation with a cognitive architecture raises several difficulties integrating the components and orchestrating behaviors and skills to perform social tasks. In a real-world scenario, the navigation system should not only consider individuals like obstacles. It is necessary to offer particular and dynamic people representation to enhance the HRI experience. The robot's behaviors must be modified by humans, directly or indirectly. In this paper, we integrate our human representation framework in a cognitive architecture to allow that people who interact with the robot could modify its behavior, not only with the interaction but also with their culture or the social context. The human representation framework represents and distributes the proxemic zones' information in a standard way, through a cost map. We have evaluated the influence of the decision-making system in human-aware navigation and how a local planner may be decisive in this navigation. The material developed during this research can be found in a public repository (https://github.com/IntelligentRoboticsLabs/social_navigation2_WAF) and instructions to facilitate the reproducibility of the results.

Autonomous robots need to plan the tasks they carry out to fulfill their missions. The missions’ increasing complexity does not let human designers anticipate all the possible situations, so traditional control systems based on state machines are not enough. This paper contains a description of the ROS2 Planning System (PlanSys2 in short), a framework for symbolic planning that incorporates novel approaches for execution on robots working in demanding environments. PlanSys2 aims to be the reference task planning framework in ROS2, the latest version of the de facto standard in robotics software development. Among its main features, it can be highlighted the optimized execution, based on Behavior Trees, of plans through a new actions auction protocol and its multi-robot planning capabilities. It already has a small but growing community of users and developers, and this document is a summary of the design and capabilities of this project.

Abstract Context awareness in ambient assisted living programmes for the elderly is a cornerstone in the current scenario of noncustomized service robots distributed around the world. This research proposes a context‐awareness system for a human–robot scene interpretation based on seven primary contexts and the American Occupational Therapy Association. The context‐awareness system defined here proposes an inference mechanism for the activity recognition supported on hierarchical Bayesian networks. However, when the information from sensors increases, the computational cost associated also increases. Thus, an evaluation of different Bayesian network models is necessary for decreasing its impact over the robot performance. Two topological models have been modelled and tested using OpenMarkov application: a two‐level approach of an input–observations layer and the activity recognition layer, and a three‐layer model setting apart a primary contexts layer, the input–observations layer, and the activity recognition layer. The qualitative and quantitative results presented here show better performance in terms of memory and memory in a three‐layer model. Besides, its effect on a hybrid architecture of a robotic platform is presented.

The ability to move around the environment is one of the most important capabilities of a mobile robot. Although navigation is considered an already achieved capacity, there is still much work to be done to integrate navigation with high level reasoning and acting. Navigate in indoor environments also involve complex actions, such as opening doors, use elevators, and many others. We propose a topological navigation system based on Artificial Intelligence (AI)Planning. Starting from a symbolic representation of the environment, navigation tasks are divided into phases, in which different actions are required. This approach has demonstrated to be very effective to plan the operations of a robot at indoor environments. The final result is method compact, efficient and scalable. Our system has been successfully tested at European Robotics League in the humanoid robot Pepper.

This article exposes the localization issues faced during the implementation of a humanoid goalkeeper to take part in the RoboCup SPL standardized soccer robot competition. For this task, we have used BICA, a state-driven, component-based architecture created by our counterparts in the URJC, to allow a much easier behavior design process. The use of BICA let us choose different self-localization methods in configuration time, and the possibility of using none in running time if some type of error appears.

When facing the problem of teaching the basis of robot control programming to computer science students, apart from the syllabus of the course, some other requirements have to be considered, such as which is the most appropriate robot and which are the right tools for learning how to control it. In this paper, we describe the tools we have chosen for teaching robotics, focusing on an environment that supports practical assignments. We also analyze the reasons that made us choose each tool, giving special emphasis to the Libre software requirement that we have imposed on every tool we are using. Finally, we present the results and opinions we have obtained from our students and the lessons we have learned by using this Libre software approach.