Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.667
M. Caccia , M. Bibuli
Virtual target based path-following is a consolidated methodology able to be easily extended to design cooperative guidance systems for unmanned marine vehicles (UMVs) at the kinematics level. Although the generation of reference yawrate as control output is compatible with advanced UMV guidance and control (GC) systems, specific formulations tailored to integrate the concept of virtual target path-following with conventional autopilots, providing heading control, and guidance systems providing line-following, are implemented and validated. To this aim suitable procedures to execute replicable experiments are defined as well as quantitative metrics to evaluate performance.
{"title":"Virtual target based path-following:integration with conventional NGC architectures and performance evaluation⁎","authors":"M. Caccia , M. Bibuli","doi":"10.1016/j.ifacol.2025.11.667","DOIUrl":"10.1016/j.ifacol.2025.11.667","url":null,"abstract":"<div><div>Virtual target based path-following is a consolidated methodology able to be easily extended to design cooperative guidance systems for unmanned marine vehicles (UMVs) at the kinematics level. Although the generation of reference yawrate as control output is compatible with advanced UMV guidance and control (GC) systems, specific formulations tailored to integrate the concept of virtual target path-following with conventional autopilots, providing heading control, and guidance systems providing line-following, are implemented and validated. To this aim suitable procedures to execute replicable experiments are defined as well as quantitative metrics to evaluate performance.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 405-410"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665319","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.640
Juraj Obradović , Matej Fabijanić , Josip Lovrić , Nadir Kapetanović , Ðula Nađ , Fausto Ferreira , Nikola Mišković
Accurate information about objects in the maritime environment is essential for the development of reliable navigation systems for autonomous surface vehicles. To provide such information, advanced deep learning models are typically employed, which depend heavily on high-quality training data. In this paper, we present the design of a sensor rack system that enables data acquisition from multiple boats, including small rental vessels commonly available along the coast. The sensor rack is constructed using modular profiles, allowing for easy integration and configuration of various sensors.
In addition to the design, we showcase a two-day data recording session conducted using the rack, which included LiDAR, two cameras, AIS, IMU, and two GNSS modules. Finally, we discuss potential improvements and future work aimed at enhancing the next version of the sensor rack.
{"title":"Design and Development of a Boat-Mountable Sensor Rack for Maritime Perception and Data Acquisition","authors":"Juraj Obradović , Matej Fabijanić , Josip Lovrić , Nadir Kapetanović , Ðula Nađ , Fausto Ferreira , Nikola Mišković","doi":"10.1016/j.ifacol.2025.11.640","DOIUrl":"10.1016/j.ifacol.2025.11.640","url":null,"abstract":"<div><div>Accurate information about objects in the maritime environment is essential for the development of reliable navigation systems for autonomous surface vehicles. To provide such information, advanced deep learning models are typically employed, which depend heavily on high-quality training data. In this paper, we present the design of a sensor rack system that enables data acquisition from multiple boats, including small rental vessels commonly available along the coast. The sensor rack is constructed using modular profiles, allowing for easy integration and configuration of various sensors.</div><div>In addition to the design, we showcase a two-day data recording session conducted using the rack, which included LiDAR, two cameras, AIS, IMU, and two GNSS modules. Finally, we discuss potential improvements and future work aimed at enhancing the next version of the sensor rack.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 248-253"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.606
Dingyi Wu , Shaolong Yang , Xinwei Cuan , Jinrong Zheng , Yan Wang
With the growing demand for the exploration and exploitation of marine resources, underwater vehicle manipulator systems have experienced rapid advancement. This paper presents an innovative design of a deformable quadruped robot for deep-sea benthic operation. The robot features a morphable system enabling multiple operational configurations. Its quadruped mechanism is adaptable to the undulating seabed environment, and its adjustable bottom-sitting posture improves operational stability under ocean current disturbances. The paper introduces the robot’s general layout, key system designs, and hydrodynamic performance analysis, laying a solid foundation for subsequent detailed design, manufacturing, and experimental validation.
{"title":"Preliminary Design of a Deformable Quadruped Underwater Robot for Deep-sea Benthic Operation","authors":"Dingyi Wu , Shaolong Yang , Xinwei Cuan , Jinrong Zheng , Yan Wang","doi":"10.1016/j.ifacol.2025.11.606","DOIUrl":"10.1016/j.ifacol.2025.11.606","url":null,"abstract":"<div><div>With the growing demand for the exploration and exploitation of marine resources, underwater vehicle manipulator systems have experienced rapid advancement. This paper presents an innovative design of a deformable quadruped robot for deep-sea benthic operation. The robot features a morphable system enabling multiple operational configurations. Its quadruped mechanism is adaptable to the undulating seabed environment, and its adjustable bottom-sitting posture improves operational stability under ocean current disturbances. The paper introduces the robot’s general layout, key system designs, and hydrodynamic performance analysis, laying a solid foundation for subsequent detailed design, manufacturing, and experimental validation.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 48-52"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.601
Lite Zhang , Yiwen Lu , Peng Tao , Zhenhua Wei
Structural health monitoring (SHM) is critical for detecting degradation in engineering structures, but traditional methods face challenges with sensor faults and computational costs. This paper proposes a deep learning-based approach using single-sensor vibration data and Transformer networks to achieve robust damage detection despite sensor failures. By integrating cosine similarity, Bayesian optimization, and a modified loss function, the method balances accuracy and efficiency, enabling cost-effective global SHM while handling noisy or faulty sensor signals.
{"title":"Deep Learning-Based Structural Health Monitoring Using Vibration Signals Under Sensor Fault Conditions","authors":"Lite Zhang , Yiwen Lu , Peng Tao , Zhenhua Wei","doi":"10.1016/j.ifacol.2025.11.601","DOIUrl":"10.1016/j.ifacol.2025.11.601","url":null,"abstract":"<div><div>Structural health monitoring (SHM) is critical for detecting degradation in engineering structures, but traditional methods face challenges with sensor faults and computational costs. This paper proposes a deep learning-based approach using single-sensor vibration data and Transformer networks to achieve robust damage detection despite sensor failures. By integrating cosine similarity, Bayesian optimization, and a modified loss function, the method balances accuracy and efficiency, enabling cost-effective global SHM while handling noisy or faulty sensor signals.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 18-23"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665676","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.637
Zeping Liu , Ziteng Huo , Yufu Gao , Yi Guo , Shenghai Wang , Guangdong Han
Establishing an efficient and accurate speed prediction model is crucial for the navigation management and state monitoring of the airfoil sail VLCC. This study proposes a GPR-BO-Physics speed prediction model based on Gaussian Process Regression, Bayesian Optimization, and a physical speed prediction model. By integrating route ocean data, Automatic Identification System data, and sail parameters, a dataset containing multiple features was constructed, followed by data preprocessing, correlation analysis, and the removal of redundant features. To improve model performance, the White Shark Optimizer was used to optimize the model weights, while Bayesian Optimization was applied for hyperparameter tuning of the Gaussian Process Regression model. Simulation results under varying wave heights indicate that the GPR-BO-Physics model effectively captures complex nonlinear relationships between inputs and speed, significantly enhancing prediction stability and accuracy.
建立高效、准确的速度预测模型对于翼型风帆VLCC的航行管理和状态监测至关重要。本文提出了一种基于高斯过程回归、贝叶斯优化和物理速度预测模型的gpr - bo -物理速度预测模型。通过整合航线海洋数据、自动识别系统数据和风帆参数，构建了包含多个特征的数据集，并对数据进行预处理、相关性分析和冗余特征去除。为提高模型性能，采用White Shark Optimizer优化模型权值，采用贝叶斯优化对高斯过程回归模型进行超参数整定。在不同波高条件下的模拟结果表明，GPR-BO-Physics模型有效地捕捉了输入与速度之间复杂的非线性关系，显著提高了预测的稳定性和精度。
{"title":"Application of Speed Prediction Based on Gaussian Process Regression to the Airfoil Sail VLCC","authors":"Zeping Liu , Ziteng Huo , Yufu Gao , Yi Guo , Shenghai Wang , Guangdong Han","doi":"10.1016/j.ifacol.2025.11.637","DOIUrl":"10.1016/j.ifacol.2025.11.637","url":null,"abstract":"<div><div>Establishing an efficient and accurate speed prediction model is crucial for the navigation management and state monitoring of the airfoil sail VLCC. This study proposes a GPR-BO-Physics speed prediction model based on Gaussian Process Regression, Bayesian Optimization, and a physical speed prediction model. By integrating route ocean data, Automatic Identification System data, and sail parameters, a dataset containing multiple features was constructed, followed by data preprocessing, correlation analysis, and the removal of redundant features. To improve model performance, the White Shark Optimizer was used to optimize the model weights, while Bayesian Optimization was applied for hyperparameter tuning of the Gaussian Process Regression model. Simulation results under varying wave heights indicate that the GPR-BO-Physics model effectively captures complex nonlinear relationships between inputs and speed, significantly enhancing prediction stability and accuracy.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 231-236"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665678","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.623
Shumin Fan , Ning Wang , Tianyu Song
Visible-light vision sensors have emerged as a cost-effective and easily deployable solution for enhancing marine vessel navigation safety. However, their effectiveness is substantially compromised in hazy marine environments, where captured images often exhibit severe contrast reduction, color distortion, and interference from sea surface waves and specular reflections. These degradation effects collectively impair visual perception accuracy and range, posing risks to navigational safety. To address these challenges, this paper proposes MDNet, a Maritime Dehazing Network designed to improve visual clarity for maritime vehicles in low-visibility hazy conditions. The MDNet framework integrates two core components: a Multi-receptive-field feature enhancement block (MFEB) and a multi-expert collaborative interaction block (MCIB). The MFEB leverages parallel multi-branch processing to perform multi-scale feature extraction, enabling the network to simultaneously capture global contextual information and fine-grained texture details. The MCIB further enhances adaptability by dynamically selecting optimal expert modules tailored to learn degradation features at different network depths, thereby effectively modulating multi-granularity contextual information in degraded images. Extensive experiments validate that MDNet achieves state-of-the-art performance, outperforming existing methods in both quantitative metrics and qualitative visual assessments. Additionally, the proposed method demonstrates practical utility by significantly boosting the accuracy and reliability of target detection in hazy marine scenarios, thereby advancing the safety and stability of autonomous navigation systems.
{"title":"Low Visibility Enhancement for Intelligent Marine Vehicles in Hazy Environments⁎","authors":"Shumin Fan , Ning Wang , Tianyu Song","doi":"10.1016/j.ifacol.2025.11.623","DOIUrl":"10.1016/j.ifacol.2025.11.623","url":null,"abstract":"<div><div>Visible-light vision sensors have emerged as a cost-effective and easily deployable solution for enhancing marine vessel navigation safety. However, their effectiveness is substantially compromised in hazy marine environments, where captured images often exhibit severe contrast reduction, color distortion, and interference from sea surface waves and specular reflections. These degradation effects collectively impair visual perception accuracy and range, posing risks to navigational safety. To address these challenges, this paper proposes MDNet, a Maritime Dehazing Network designed to improve visual clarity for maritime vehicles in low-visibility hazy conditions. The MDNet framework integrates two core components: a Multi-receptive-field feature enhancement block (MFEB) and a multi-expert collaborative interaction block (MCIB). The MFEB leverages parallel multi-branch processing to perform multi-scale feature extraction, enabling the network to simultaneously capture global contextual information and fine-grained texture details. The MCIB further enhances adaptability by dynamically selecting optimal expert modules tailored to learn degradation features at different network depths, thereby effectively modulating multi-granularity contextual information in degraded images. Extensive experiments validate that MDNet achieves state-of-the-art performance, outperforming existing methods in both quantitative metrics and qualitative visual assessments. Additionally, the proposed method demonstrates practical utility by significantly boosting the accuracy and reliability of target detection in hazy marine scenarios, thereby advancing the safety and stability of autonomous navigation systems.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 148-153"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665681","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.644
Ren Mengyao , Su Xin , Zhang Qi
In real-world engineering scenarios, deep-water jacket platforms are exposed to complex, varying loads— including tension and bending—resulting from the combined actions of waves, currents, and wind. These conditions can induce structural damage and potentially compromise the integrity of the platform. This study establishes a data foundation for further analysis by utilizing the dynamic response data collected from an actual platform and applies a multi-output regression modeling approach. Six representative environmental parameters—wind speed, wind direction, current speed, current direction, significant wave height, and peak period—are selected as input features. A K-Nearest Neighbors (KNN) algorithm is employed to construct the mapping relationship between environmental conditions and structural responses. Subsequently, a multi-output regression model is developed to predict the mechanical response of the platform based on marine environmental inputs. This method effectively addresses the high computational cost associated with traditional numerical analysis, overcomes the limitations of single-output prediction models, and enhances both prediction accuracy and efficiency, thereby contributing to the safe operation of offshore platforms.
{"title":"A Multi-Output Regression-Based Method for Predicting Structural Responses of Deepwater Jacket Platforms","authors":"Ren Mengyao , Su Xin , Zhang Qi","doi":"10.1016/j.ifacol.2025.11.644","DOIUrl":"10.1016/j.ifacol.2025.11.644","url":null,"abstract":"<div><div>In real-world engineering scenarios, deep-water jacket platforms are exposed to complex, varying loads— including tension and bending—resulting from the combined actions of waves, currents, and wind. These conditions can induce structural damage and potentially compromise the integrity of the platform. This study establishes a data foundation for further analysis by utilizing the dynamic response data collected from an actual platform and applies a multi-output regression modeling approach. Six representative environmental parameters—wind speed, wind direction, current speed, current direction, significant wave height, and peak period—are selected as input features. A K-Nearest Neighbors (KNN) algorithm is employed to construct the mapping relationship between environmental conditions and structural responses. Subsequently, a multi-output regression model is developed to predict the mechanical response of the platform based on marine environmental inputs. This method effectively addresses the high computational cost associated with traditional numerical analysis, overcomes the limitations of single-output prediction models, and enhances both prediction accuracy and efficiency, thereby contributing to the safe operation of offshore platforms.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 271-276"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.648
Chao Zhu, Deda Kong, Shouxu Zhang, Yang Zhao, Xiaoxu Liu
To address the motion control requirements of underwater vehicle-manipulator systems (UVMS) in complex environments, this study proposes an integrated ROS-based simulation and motion planning method. By combining the MoveIt motion planning framework with the Gazebo physics engine, a high-fidelity virtual control and dynamic simulation platform for UVMS is established. Key tasks include configuring the URDF model of the manipulator, optimizing kinematic solutions, collision detection, and implementing path planning algorithms. Additionally, multi-joint cooperative control is achieved using the ros_control framework. Experimental results demonstrate that the system effectively supports obstacle-avoidance trajectory generation and precise motion control in dynamic environments, providing a reliable simulation verification foundation for underwater operational tasks.
{"title":"Construction of Motion Planning and Control System for UVMS Manipulator Based on MoveIt","authors":"Chao Zhu, Deda Kong, Shouxu Zhang, Yang Zhao, Xiaoxu Liu","doi":"10.1016/j.ifacol.2025.11.648","DOIUrl":"10.1016/j.ifacol.2025.11.648","url":null,"abstract":"<div><div>To address the motion control requirements of underwater vehicle-manipulator systems (UVMS) in complex environments, this study proposes an integrated ROS-based simulation and motion planning method. By combining the MoveIt motion planning framework with the Gazebo physics engine, a high-fidelity virtual control and dynamic simulation platform for UVMS is established. Key tasks include configuring the URDF model of the manipulator, optimizing kinematic solutions, collision detection, and implementing path planning algorithms. Additionally, multi-joint cooperative control is achieved using the ros_control framework. Experimental results demonstrate that the system effectively supports obstacle-avoidance trajectory generation and precise motion control in dynamic environments, providing a reliable simulation verification foundation for underwater operational tasks.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 295-300"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665700","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01Epub Date: 2025-12-04DOI: 10.1016/j.ifacol.2025.11.651
Xinwei Cuan , Andong Wang , Heng Guo , Yuhao Hu , Shaolong Yang , Gang Wan , Xinyu Li
To address the inherent limitations of Underactuated underwater unmanned underwater vehicles (AUVs) in terms of rudder efficiency degradation at low speeds and high energy consumption of transverse thrusters, this paper proposes a speed-adaptive collaborative control framework that dynamically allocates steering torque between the rudder and thrusters based on the real-time speed and environmental disturbances. The core innovation lies in a control architecture that mixes selfdisturbance control (ADRC) with an enhanced Sigmoid-based allocation strategy. A dynamic speed threshold adjustment mechanism is designed to adaptively augment thruster actuation in the event of persistent heading errors and rudder inefficiency, ensuring fast response in critical situations. Finally, it is verified through the lake test experiment that the lateral error of AUV is always controlled less than 0.2m at 2 knots of navigation speed, and the average heading error is no more than 0.85% when the AUV is carried out at variable navigation speed. The test proves that this method effectively solves the problem of synergistic optimization of low-speed and high-speed maneuverability and energy efficiency of underdriven UUVs.
{"title":"Underactuated unmanned underwater vehicle speed adaptive cooperative allocation control","authors":"Xinwei Cuan , Andong Wang , Heng Guo , Yuhao Hu , Shaolong Yang , Gang Wan , Xinyu Li","doi":"10.1016/j.ifacol.2025.11.651","DOIUrl":"10.1016/j.ifacol.2025.11.651","url":null,"abstract":"<div><div>To address the inherent limitations of Underactuated underwater unmanned underwater vehicles (AUVs) in terms of rudder efficiency degradation at low speeds and high energy consumption of transverse thrusters, this paper proposes a speed-adaptive collaborative control framework that dynamically allocates steering torque between the rudder and thrusters based on the real-time speed and environmental disturbances. The core innovation lies in a control architecture that mixes selfdisturbance control (ADRC) with an enhanced Sigmoid-based allocation strategy. A dynamic speed threshold adjustment mechanism is designed to adaptively augment thruster actuation in the event of persistent heading errors and rudder inefficiency, ensuring fast response in critical situations. Finally, it is verified through the lake test experiment that the lateral error of AUV is always controlled less than 0.2m at 2 knots of navigation speed, and the average heading error is no more than 0.85% when the AUV is carried out at variable navigation speed. The test proves that this method effectively solves the problem of synergistic optimization of low-speed and high-speed maneuverability and energy efficiency of underdriven UUVs.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 22","pages":"Pages 313-317"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665703","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Autonomous Mobile Robot (AMR) navigation in dynamic environments that may be GPS denied, without a-priori maps, is an unsolved problem with potential to improve humanity’s capabilities. Conventional modular methods are computationally inefficient, and require explicit feature extraction and engineering that inhibit generalization and deployment at scale. We present an Out-of-Distribution (OOD) Deep Reinforcement Learning (DRL) approach that includes functionality in unstructured terrain and dynamic obstacle avoidance capabilities. We leverage accelerated simulation training in a racetrack with a transition probability to parameterize spatial reasoning with intrinsic exploratory behavior, in a compact, computationally efficient Artificial Neural Network (ANN), which we transfer zero-shot with a reward component to mitigate differences between simulation and real world physics. Our approach enables utility without a separate high-level planner or real-time cartography and utilizes a fraction of the computation resources of modular methods, enabling execution in a range of AMRs with different embedded computer payloads.
自主移动机器人（Autonomous Mobile Robot， AMR）在动态环境中导航，在GPS可能被拒绝的情况下，没有先验地图，是一个尚未解决的问题，有可能提高人类的能力。传统的模块化方法计算效率低下，并且需要明确的特征提取和工程，这抑制了大规模的泛化和部署。我们提出了一种分布外（OOD）深度强化学习（DRL）方法，包括非结构化地形的功能和动态避障能力。我们在一个紧凑的、计算效率高的人工神经网络（ANN）中利用具有转移概率的赛道加速模拟训练来参数化具有内在探索行为的空间推理，我们将零射击与奖励组件转移到模拟和现实世界物理之间的差异。我们的方法无需单独的高级规划器或实时制图即可实现实用，并且利用模块化方法的一小部分计算资源，可以在具有不同嵌入式计算机有效载荷的一系列amr中执行。
{"title":"Mobile Robot Exploration Without Maps via Out-of-Distribution Deep Reinforcement Learning","authors":"Shathushan Sivashangaran , Apoorva Khairnar , Azim Eskandarian","doi":"10.1016/j.ifacol.2025.12.292","DOIUrl":"10.1016/j.ifacol.2025.12.292","url":null,"abstract":"<div><div>Autonomous Mobile Robot (AMR) navigation in dynamic environments that may be GPS denied, without a-priori maps, is an unsolved problem with potential to improve humanity’s capabilities. Conventional modular methods are computationally inefficient, and require explicit feature extraction and engineering that inhibit generalization and deployment at scale. We present an Out-of-Distribution (OOD) Deep Reinforcement Learning (DRL) approach that includes functionality in unstructured terrain and dynamic obstacle avoidance capabilities. We leverage accelerated simulation training in a racetrack with a transition probability to parameterize spatial reasoning with intrinsic exploratory behavior, in a compact, computationally efficient Artificial Neural Network (ANN), which we transfer zero-shot with a reward component to mitigate differences between simulation and real world physics. Our approach enables utility without a separate high-level planner or real-time cartography and utilizes a fraction of the computation resources of modular methods, enabling execution in a range of AMRs with different embedded computer payloads.</div></div>","PeriodicalId":37894,"journal":{"name":"IFAC-PapersOnLine","volume":"59 30","pages":"Pages 533-538"},"PeriodicalIF":0.0,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145801840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号