robotics

Robot software integration is often time-consuming and inefficient due to monolithic architectures and tight coupling, hindering rapid adaptation to task changes, performance optimization, or cross-hardware deployment. To address this, we propose Coral—a unified abstraction layer for composable robotics—that enables plug-and-play integration of heterogeneous components (e.g., LiDAR-SLAM, multi-robot coordination modules) without modifying underlying code. Coral employs a semantics-constrained, high-level interface mapping mechanism to decouple component behavior from implementation details. It is fully compatible with mainstream robotics toolchains (e.g., ROS/ROS 2), supports modular architecture design, and enables runtime reconfiguration. Experimental evaluation across diverse complex scenarios demonstrates significant improvements in development efficiency and system adaptability. Coral reduces integration and porting overhead while preserving functional correctness and real-time constraints. The framework is open-sourced and already deployed in multiple real-world robotic applications.

Real-time multi-step planning and obstacle avoidance for autonomous robots in dynamic environments remain challenging, particularly under resource constraints and without prior map knowledge. Method: We propose a lightweight, closed-loop reactive planning framework that requires no pre-mapping or offline computation. Our approach integrates biologically inspired attention mechanisms with local LiDAR perception to construct transient control-chain plans. It introduces forward depth-first model checking—novel in real-time multi-step planning—combined with environment-aware 2D LiDAR discretization and closed-loop feedback control. Contribution/Results: The framework provides theoretical guarantees on safety and interpretability. Empirically, it generates safe, multi-step local trajectories within 100 ms on low-power embedded hardware. In complex scenarios—including dead ends and playgrounds—it significantly outperforms single-step reactive systems in obstacle avoidance success rate and response robustness.

To address the limited flexibility in skill composition and insufficient system generalizability in robot task-level control, this paper proposes a model-driven AI planning and execution framework. The framework integrates formal world and action models with PDDL-based modeling, HTN/STRIPS planners, ROSPlan middleware, real-time state monitoring, and closed-loop execution feedback to enable automated composition and robust execution of primitive skills. Its key contribution is the first systematic survey and unified analysis of integration challenges, evolutionary trajectories, and design trade-offs associated with mainstream model-based planning paradigms—particularly within modern robotic platforms such as ROS. The work advances the engineering deployment of general-purpose task planning systems and provides theoretical foundations, principled design guidelines, and practical implementation insights for developing autonomous robotic systems that are interpretable, verifiable, and reusable.

To address the challenges of multi-simulator interoperability and the inability of closed-source platforms (e.g., Parrot Sphinx) to enable real-time control of non-robotic entities or provide full-state feedback, this paper proposes the Sphinx-Gazebo bidirectional co-simulation architecture. It employs a mirrored UAV entity mechanism to achieve ROS-based real-time control and high-fidelity, full-state estimation. To our knowledge, this is the first work to deploy an end-to-end model predictive control (MPC) algorithm within this integrated environment for multi-agent dynamic target tracking. Experimental results demonstrate that the proposed MPC controller improves trajectory tracking accuracy by 37% and reduces control latency by 29% compared to the Anafi’s native PID controller. The framework supports seamless deployment on physical Anafi drones, and all source code is fully open-sourced.

To address challenges in the Internet of Robotic Things (IoRT)—including limited onboard perception, insufficient environmental semantic understanding, and poor real-time performance in dynamic obstacle avoidance—this paper proposes a ROS-based collaborative architecture integrating intelligent spaces with mobile robots. Methodologically, it introduces a distributed environmental perception network as an external sensing enhancement module within ROS, enabling closed-loop coordination among multi-source sensor fusion, semantic environmental modeling, and autonomous robot decision-making. Additionally, a lightweight real-time collision prediction algorithm is designed to reduce reliance on individual robot intelligence. Experimental results demonstrate significant improvements in complex dynamic environments: obstacle avoidance success rate increases by 23.6%, and system response latency decreases by 41%. These findings validate the effectiveness and scalability of the proposed IoRT architecture for cost-constrained robotic systems.

This study addresses the challenge of simultaneously achieving spatial abstraction, temporal predictability, and state continuity in ROS 2 middleware operating within dynamic, resource-constrained wireless environments. For the first time, it introduces a three-dimensional “space–time–state” analytical framework, integrating architectural analysis, formal modeling, and a comprehensive literature review to systematically investigate the underlying mechanisms and structural trade-offs of ROS 2 middleware—particularly DDS and Zenoh—in discovery protocols, data exchange, and state management. The work uncovers critical performance bottlenecks in existing approaches concerning modular deployment, real-time control, and disconnected operation recovery. These insights lay a theoretical foundation and offer concrete design guidelines for developing robust, scalable next-generation robotic middleware.

This study addresses the challenges of robust 6D pose estimation and multi-object tracking for industrial mobile robots operating in dynamic production environments, where reliance on real-world data, sensitivity to perceptual noise, and spatiotemporal inconsistencies often degrade performance. To overcome these limitations, the authors propose a ROS 2-based LiDAR perception framework that innovatively integrates a transform-equivariant 3D detection model trained on synthetic data with a center-point-based multi-object tracking algorithm. This approach significantly enhances system robustness and generalization without requiring extensive real-world annotations. Evaluated across 72 diverse scenarios, the method achieves an IoU of 62.6% for standalone pose estimation, which improves to 83.12% when combined with tracking, while attaining a high-order tracking accuracy of 91.12%.

Existing simulation platforms struggle to support heterogeneous multi-robot systems—such as unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs)—in performing collaborative SLAM, perception, and exploration tasks within large-scale, high-fidelity dynamic environments. This work proposes an open-source simulation framework built upon Unreal Engine 5 that introduces a unified navigation and control architecture specifically designed for UAV–UGV heterogeneous collaboration. The framework features physically realistic long-wave infrared and night-vision sensor models and incorporates dynamic environmental elements such as fire and flooding. Extending AirSim/Cosys-AirSim, it integrates ROS 2 interfaces, a lightweight API, and strict time synchronization to support both replayable trajectories and online planning. The project also provides a benchmark dataset for heterogeneous multi-robot SLAM across desert, forest, and urban scenarios, demonstrating the system’s efficacy in collaborative mapping and exploration; all code and data are publicly released.

This work addresses the challenge that traditional model-based testing is ill-suited for distributed robotic systems due to their high nondeterminism, dynamic reconfiguration, and inherent complexity. To overcome this limitation, the paper proposes the Scenario Specification Language (SCSL), which enables the construction of system-level tests by composing basic scenarios. The approach integrates runtime online test generation and execution with mechanisms for dynamic component joining/leaving and interface reconnection, thereby supporting automated testing and dynamic reconfiguration. The syntax and semantics of SCSL are validated through a robotic salvage mission case study, where automatically generated tests effectively demonstrate the feasibility and advantages of the proposed method.

This study addresses the challenge of collision-free coordination and planning for multiple mobile robots in large-scale warehouse environments without onboard perception or computation hardware. By leveraging a distributed network of CCTV cameras and edge computing, the system constructs a topological camera graph directly in uncalibrated image space. A hierarchical planner generates sequences of cameras and corresponding image-space trajectories for each robot, while a combined priority-based and joint coordination strategy efficiently manages shared resources in overlapping camera fields of view. This work presents the first real-world demonstration of multi-robot collaborative operation in a warehouse setting relying solely on an external vision system, eliminating the need for dedicated onboard navigation hardware. Experiments successfully coordinated four robots across a facility equipped with 30 cameras and six 27-meter aisles, validating the effectiveness of the proposed image-space topological modeling and resource-sharing mechanisms in terms of task completion time and coordination efficiency.