This survey explores the transformative impact of robotics on intralogistics, driven by supply chain complexities and evolving consumer demands. It reviews various robotic systems alongside foundational algorithms for their operation, such as simultaneous localization and mapping, diverse path planning strategies, and advanced perception/manipulation techniques. An important focus is put on multirobot system coordination, task allocation, and fleet management in logistics. The article then examines human–robot collaboration and relevant safety standards. Finally, it identifies key challenges for future development, including interoperability, advanced AI integration, scalability, robustness in dynamic environments, and economic barriers to adoption.
The increasing use of neural networks (NNs) in control systems for robots, autonomous systems, and other safety-critical applications will demand confidence in their safety, stability, and robustness. The aim of this article is to review recently developed NN model structures (both static and dynamic) with built-in certification of robustness in various forms, including bounded sensitivity to adversarial disturbances (Lipschitzness), dynamical stability, and robust invertibility. In doing so, we aim to help strengthen connections between control theory and machine learning. We present a unifying overview of tools to guarantee such robustness properties while retaining model expressivity and computational efficiency of training and inference. Expressivity is maintained via sophisticated certification methods building on the theory of integral quadratic constraints in robust control, while computational efficiency is maintained via direct parameterizations, which enable learning of robust models via standard unconstrained optimization methods such as gradient descent, without any auxiliary constraints or projections. We then show how such robust NN models can be used in building blocks within architectures for learning control system components such as physics-informed nonlinear observers (state estimators) with guaranteed convergence, control policy parameterizations with guaranteed stability and robustness, and Lyapunov functions and their variants, such as storage and value functions.
This survey reviews recent developments in fault diagnosis for both linear and nonlinear dynamical systems, covering model-based and data-driven approaches as well as passive and active detection and estimation methods. A central focus is placed on the geometric interpretation of diagnosis filters and their connection to the concept of behavioral sets, providing an intuitive view of their performance. We also review optimization-based techniques that enhance the robustness of linear filters when applied to nonlinear or uncertain systems. Furthermore, we point out recent progress in active fault diagnosis, where input design plays a key role in improving detectability and estimation accuracy. To bridge theory and practice, we include a set of real-world industrial applications that demonstrate the implementation and effectiveness of these methods in realistic settings.
Robot contests serve as catalysts for scientific advancement and innovation in robotics. While traditional robotics research focuses mainly on subsystem optimization or theoretical analysis and design, robot contests provide a crucial platform for benchmarking complete robot systems by encouraging integrated system development and novel solutions to realistic problems common to all participants. This article explains how robot competitions and challenges designed according to best practices foster progress across various domains. It highlights their significant impact on education, including their promotion of teamwork and problem-solving skills, and their role in technology transfer, exemplified by successful spin-off companies. Popular robot contests based on more traditional approaches are surveyed as sources of a diversity of approaches to scoring performance and the development of good practices. Alternative concepts, such as cooperative competitions to foster transferability across robotic platforms and tasks as well as performance assessments using benchmarking metrics, are introduced in the last part of the article.
Teach and repeat (T&R) navigation has gained popularity over the past 15 years for its reliable path tracking in global navigation satellite system (GNSS)–denied environments. By using topometric navigation, it blends the scalability of graphs to connect distant places of interest, with metric estimates of a robot's state that allow control algorithms to correct path-tracking errors. Robots have been field-tested in off-road environments and found commercial applications in mining, cleaning, and agriculture. We present a summary of the different interpretations of T&R, introduce the relative strengths of various sensors in T&R, highlight methods that boost the reliability of T&R on robot hardware, and propose that the effectiveness of local maps in topometric navigation comes from a sensor-dependent bias–variance trade-off.
We comprehensively survey mixed crowd navigation, focusing on the integration of robotic agents with volitional crowds (humans or human-driven vehicles) to achieve system-wide benefits. The survey is organized following the perception–interaction–planning–control pipeline, examining four core components: (a) perceiving global crowd behavior from local robot observations through nonparticipant and participant observation methods; (b) modeling volitional agent responses via rule-based and data-driven interaction frameworks; (c) predicting crowd dynamics across microscopic, mesoscopic, and macroscopic scales using both traditional and machine learning approaches; and (d) synthesizing control policies that guide crowds toward desired states. Wed address critical challenges such as complex interaction modeling under partial observability, constrained robotic influence, and the need for multiscale behavioral consistency. Key applications span pedestrian crowd management and mixed traffic control. We also highlight emerging trends in mixed crowd navigation, including the use of deep reinforcement learning and foundation models, while identifying persistent challenges in human irrationality modeling, compliance prediction, and privacy-preserving algorithms for real-world deployment.
This article presents an overview of operator theory, operator-splitting methods, and their applications. It shows that many existing algorithms in optimization and games can be abstracted to an operator-theoretic problem that often requires only minimal conditions. The main advantage is that an operator-theoretic approach is unifying and can simplify complicated convergence proofs. The article reviews some of the applications of this approach, with a focus on game-theoretic algorithms.
The use of socially assistive robots (SARs) in autism spectrum disorder (ASD) therapy has seen rapid advancements, with increasing evidence supporting their role in fostering engagement and skill acquisition. However, much of the existing research has focused on short-term outcomes without fully integrating a developmental perspective on learning. This article repositions SARs as cognitive scaffolds—external supports that guide children with ASD in acquiring, practicing, and internalizing social cognition skills in a longer-term developmental perspective. Bridging technical innovation with psychological and clinical views, we propose a structured framework for robot-assisted therapy that transitions from robot-led engagement to human-mediated generalization. Furthermore, we discuss design principles for optimizing SARs, address real-world application challenges, and outline key ethical considerations. We conclude with recommendations for future research and deployment strategies to ensure that SARs serve as dynamic, adaptive tools that empower individuals with ASD rather than replacing traditional therapeutic interactions.
The ability to induce desired shape and topographical changes in matter on demand is key to future advances in materials science and robotics. An important class of shape-morphing solids is characterized by spatially programmable anisotropic deformations that can be activated through a variety of stimuli, including light, heat, and pressure. We use the word nematic to describe such solids, borrowing a term from liquid crystal theory. We highlight the diversity of nematics that have been developed in recent years and review their distinguishing geometric and mechanical features, with a focus on thin shape-morphing sheets. We showcase how shape-morphing nematic systems can be designed to execute mechanical tasks. We end with an exploration of the challenges and opportunities involving the control of shape-morphing systems, which will be necessary for more advanced applications in robotics and is thus a key frontier in the field.
Medical robots capable of autonomously performing interventional and surgical procedures are becoming a reality. Autonomous medical robots promise enhanced accuracy, reduced medical errors, improved accessibility to specialized care, and lower healthcare costs through efficient, minimally invasive procedures. This review examines motion planning as a fundamental building block enabling autonomous medical robots. Motion planning aims to compute high-quality and safe motions for robotic instruments to accomplish interventional and surgical procedures while considering anatomical constraints and physical limitations. We categorize medical robot motion planning into navigation planning (maneuvering instruments to targets via intratissue or endoluminal paths) and manipulation planning (tissue interaction through contact or contact-free approaches). We frame motion planning as a type of AI guidance that can enable eyes-on/hands-off automation. We review state-of-the-art methods in navigation and manipulation planning for medical robots, discuss challenges to clinical adoption, and explore future opportunities for autonomous medical robots.
In this tutorial article, we expose the mechanisms underlying the design of optimization algorithms based on so-called dynamic integral quadratic constraints. These tools from robust control allow one to systematically construct accelerated first-order optimization algorithms with optimal guaranteed convergence rates by solving small-sized semidefinite programs. This is possible even if the information flow from and to the gradient is subject to nontrivial dynamics such as delays. Numerical experiments not only illustrate how to recover accelerated gradient algorithms by design but also unveil the flexibility of this approach gained from its embedding into systems theory and controller design, relying on the generalized plant framework.
This article presents an overview of geometric approaches to facilitate the acquisition and transfer of robot skills. It focuses on three complementary geometric frameworks: signed distance fields, geometric algebra, and Riemannian geometry, which provide representations facilitating learning, planning, control, and optimization problems in robotics. The first consists of representing shapes in an implicit manner through the use of a distance function, where different approaches can be used to encode and learn this function. The second, geometric algebra, is linked to Clifford algebra and allows basic geometric primitives to be treated in a unified manner, including 6D poses, planes, lines, circles, and spheres, which can represent various forms of constraints in robot applications. The third leverages the use of Riemannian manifolds to extend models and algorithms originally developed for standard Euclidean data to curved spaces. These manifolds can represent a variety of geometric objects in robotics, not only for structured objects such as spheres, matrices, and subspaces, but also for more generic smooth manifolds described by a Riemannian metric to measure distances. The article discusses the distinctions and connections between these geometric approaches and shows how they can contribute to various problems in robotics, with a focus on manipulation tasks.
Place recognition, the ability to identify previously visited locations, is critical for both biological navigation and autonomous systems. This review synthesizes findings from robotic systems, animal studies, and human research to explore how different systems encode and recall place. We examine the computational and representational strategies employed across artificial systems, animals, and humans, highlighting convergent solutions such as topological mapping, cue integration, and memory management. Animal systems reveal evolved mechanisms for multimodal navigation and environmental adaptation, while human studies provide unique insights into semantic place concepts, cultural influences, and introspective capabilities. Artificial systems showcase scalable architectures and data-driven models. We propose a unifying set of concepts by which to consider and develop place recognition mechanisms and identify key challenges such as generalization, robustness, and environmental variability. This review aims to foster innovations in artificial localization by connecting future developments in artificial place recognition systems to insights from both animal navigation research and human spatial cognition studies.
Remarkable capabilities have been achieved by robotics and AI, mastering complex tasks and environments. Yet, humans often remain passive observers, fascinated but uncertain how to engage. Robots, in turn, cannot reach their full potential in human-populated environments without effectively modeling human states and intentions and adapting their behavior. To achieve a synergistic human-robot collaboration (HRC), a continuous information flow should be established: humans must intuitively communicate instructions, share expertise, and express needs. In parallel, robots must clearly convey their internal state and forthcoming actions to keep users informed, comfortable, and in control. This review identifies and connects key components enabling intuitive information exchange and skill transfer between humans and robots. We examine the full interaction pipeline: from the human-to-robot communication bridge translating multimodal inputs into robot-understandable representations, through adaptive planning and role allocation, to the control layer and feedback mechanisms to close the loop. Finally, we highlight trends and promising directions toward more adaptive, accessible HRC.
Machine learning has facilitated significant advancements across various robotics domains, including navigation, locomotion, and manipulation. Many such achievements have been driven by the extensive use of simulation as a critical tool for training and testing robotic systems prior to their deployment in real-world environments. However, simulations consist of abstractions and approximations that inevitably introduce discrepancies between simulated and real environments, known as the reality gap. These discrepancies significantly hinder the successful transfer of systems from simulation to the real world. Closing this gap remains one of the most pressing challenges in robotics. Recent advances in sim-to-real transfer have demonstrated promising results across various platforms, including locomotion, navigation, and manipulation. By leveraging techniques such as domain randomization, real-to-sim transfer, state and action abstractions, and sim–real cotraining, many works have overcome the reality gap. However, challenges persist, and a deeper understanding of the reality gap's root causes and solutions is necessary. In this survey, we present a comprehensive overview of the sim-to-real landscape, highlighting the causes, solutions, and evaluation metrics for the reality gap and sim-to-real transfer. Our goal is to provide a guide to identifying the challenges and opportunities for future advancements in sim-to-real transfer for robotic systems.
This review offers a comprehensive overview of current traffic modeling, estimation, and control methods, along with resulting field experiments. It highlights key developments and future directions in leveraging technological advancements to improve traffic management and safety. The focus is on macroscopic, microscopic, and micro-macro models, as well as state-of-the-art control techniques and estimation methods for deploying vehicles in traffic field experiments.
Environmental intelligence brings together synergies among environmental science, advanced sensor research, data science, robotics, and artificial intelligence to enable a better understanding of the natural environment and effective coordination of the responses to the associated challenges. Environmental intelligence is a particularly relevant application field for the new trend of ecorobotics research, in which robots are envisioned as environmentally responsible, energy-efficient, bioinspired machines capable of adapting and safely interacting within natural ecosystems. By incorporating an ecological approach to robot design, innovative fabrication technologies, biodegradable materials, distributed architectures for sensing and intelligence, new power sources, and energy-harvesting solutions, ecorobots offer practical solutions for in-depth analysis of natural processes. This increases environmental knowledge and supports sustainable strategies to safeguard the environment. This article reviews the field of environmental intelligence with a focus on the roles, features, and abilities of ecorobots and provides future perspectives toward environmental sustainability.
The Human-Robot Interaction (HRI) community often highlights the social context of an interaction as a key consideration when designing, implementing, and evaluating robot behavior. Unfortunately, researchers use the term "social context" in varied ways. This can lead to miscommunication, making it challenging to draw connections between related work on understanding and modeling the social contexts of human-robot interactions. To address this gap, we survey the HRI literature for existing definitions and uses of the term "social context". Then, we propose a conceptual model for describing the social context of a human-robot interaction. We apply this model to existing work, and we discuss a range of attributes of social contexts that can help researchers plan for interactions, develop behavior models for robots, and gain insights after interactions have taken place. We conclude with a discussion of open research questions in relation to understanding and modeling the social contexts of human-robot interactions.
The control of marine robots has long relied on model-based methods grounded in classical and modern control theory. However, the nonlinearity and uncertainties inherent in robot dynamics, coupled with the complexity of marine environments, have revealed the limitations of conventional control methods. The rapid evolution of machine learning has opened new avenues for incorporating data-driven intelligence into control strategies, prompting a paradigm shift in the control of marine robots. This paper provides a review of recent progress in marine robot control through the lens of this emerging paradigm. The review covers both individual and cooperative marine robotic systems, highlighting notable achievements in data-driven control of marine robots and summarizing open-source resources that support the development and validation of advanced control methods. Finally, several future perspectives are outlined to guide research toward achieving high-level autonomy for marine robots in real-world applications. This paper aims to serve as a roadmap toward the next-generation control framework of marine robots in the era of data-driven intelligence.
Machine learning is rapidly advancing nearly every field of science and engineering, and control theory is no exception. In particular, it has shown incredible promise for handling several of the main challenges facing modern dynamics and control, including complexity, unmodeled dynamics, strong nonlinearity, and hidden variables. However, machine learning models are often expensive to train and deploy, fail to generalize beyond the training data, and suffer from a lack of explainability, interpretability, and guarantees, all of which limit their use in real-world and safety-critical control applications. Sparse nonlinear modeling and control techniques are a powerful class of machine learning that promote parsimony through sparse optimization, providing data-efficient models that are more interpretable and generalizable and have proven effective for control. In this review, we explore the use of sparse optimization in the context of machine learning to develop compact models and controllers that are easy to train, require significantly less data, and make low-latency predictions. In particular, we focus on applications in model predictive control and reinforcement learning, two of the foundational algorithms in control theory.
