This paper addresses the problem of visual object tracking for Unmanned Aerial Vehicles(UAVs).Most Siamese trackers are used to regard object tracking as classification and regression problems.However,it is difficult ...This paper addresses the problem of visual object tracking for Unmanned Aerial Vehicles(UAVs).Most Siamese trackers are used to regard object tracking as classification and regression problems.However,it is difficult for these trackers to accurately classify in the face of similar objects,background clutters and other common challenges in UAV scenes.So,a reliable classifier is the key to improving UAV tracking performance.In this paper,a simple yet efficient tracker following the basic architecture of the Siamese neural network is proposed,which improves the classification ability from three stages.First,the frequency channel attention module is introduced to enhance the target features via frequency domain learning.Second,a template-guided attention module is designed to promote information exchange between the template branch and the search branch,which can get reliable classification response maps.Third,adaptive cross-entropy loss is proposed to make the tracker focus on hard samples that contribute more to the training process,solving the data imbalance between positive and negative samples.To evaluate the performance of the proposed tracker,comprehensive experiments are conducted on two challenging aerial datasets,including UAV123 and UAVDT.Experimental results demonstrate that the proposed tracker achieves favorable tracking performances in aerial benchmarks beyond 41 frames/s.We conducted experiments in real UAV scenes to further verify the efficiency of our tracker in the real world.展开更多
The coordination of multiple Earth-observing satellites presents a significant scheduling challenge.This paper introduces a fully distributed autonomous scheduling solution that utilizes a learning-based mechanism thr...The coordination of multiple Earth-observing satellites presents a significant scheduling challenge.This paper introduces a fully distributed autonomous scheduling solution that utilizes a learning-based mechanism through an independent proximal policy optimization(IPPO)algorithm.Each satellite independently makes decisions regarding tasks,such as imaging,desaturation,and charging,while adapting to dynamic environmental changes to enhance its real-time constellation scheduling performance.The proposed fully distributed strategy enables individual satellites to update their policies based solely on their observations.The only requirement is the unidirectional broadcast of a completion flag upon target observation.This approach distinguishes itself from traditional centralized methods,thus enhancing the overall robustness and security of the system.In simulations,our strategy exhibited effective observational mission planning results for major cities worldwide.The results show that the proposed method addresses both autonomous scheduling and significantly improves constellation performance and reliability.展开更多
The propulsion systems of a multi-rotor unmanned aerial vehicle(UAV)is crucial,as it directly affects the UAV’s performance,efficiency,and safety.Since the components of the UAV propulsion system are highly interconn...The propulsion systems of a multi-rotor unmanned aerial vehicle(UAV)is crucial,as it directly affects the UAV’s performance,efficiency,and safety.Since the components of the UAV propulsion system are highly interconnectioned,we developed a fuzzy fault tree analysis method to analysis the varying reliability under different fault conditions.Combining the fuzzy fault tree analysis of the T-S model and the UAV propulsion system model,we constructed a fuzzy fault tree of the T-S type for the system and performed a reliability analysis.This fuzzy fault tree allows us to model the system from two perspectives:fuzzy failure rate and failure degree.Consequently,two methods can be used for failure analysis of UAV systems.The first method involves calculating the system’s fuzzy failure rate based on the component’s fuzzy failure rate.The second method calculates the fuzzy failure rate of the system based on the failure degree of the component.The computational results indicate that both methods are well-suited for fault diagnosis in UAV propulsion systems.Compared to traditional fault tree analysis,which does not subdivide fault degrees,the proposed methods provide more accurate fault rate assessments.展开更多
This paper proposes an intermittent measurement-based attitude tracking control strategy for spacecraft operating in the presence of sensor-actuator faults.Asampled-data(self-)learning observer is developed to estimat...This paper proposes an intermittent measurement-based attitude tracking control strategy for spacecraft operating in the presence of sensor-actuator faults.Asampled-data(self-)learning observer is developed to estimate both the spacecraft’s states and lumped disturbances,effectively mitigating the impact of faults.This observer acts as a virtual predictor,reconstructing states and actuator fault deviations using only intermittent measurement data,addressing the limitations imposed by sensor failures.The control scheme incorporates compensation based on the predictor’s estimates,ensuring robust attitude tracking despite the presence of faults.We provide the first proof of bounded stability for this learning observer utilizing intermittent information,expanding its applicability.Numerical simulations demonstrate the effectiveness of this innovative strategy,highlighting its potential for enhancing spacecraft autonomy and reliability in challenging operational scenarios.展开更多
This paper presents an image-feature-aware(IFA)planner for quadrotors,which integrates image feature tracking into its path-planning framework.The IFA-planner aims to improve the visual localization performance of qua...This paper presents an image-feature-aware(IFA)planner for quadrotors,which integrates image feature tracking into its path-planning framework.The IFA-planner aims to improve the visual localization performance of quadrotors in multifarious environments where feature points may be sparse or diverse.Unlike traditional methods that decouple visual localization and path planning,the IFA-planner adaptively identifies and tracks feature-rich spatial units,called anchors,along a feasible path.The anchors provide additional feature points to the visual localization module,especially in scenarios with sparse or uneven features,thus enhancing localization robustness.Via clustering-based method,the anchor selection can handle different feature point distributions without manual tuning.Moreover,a detachment prediction mechanism is incorporated to convert the selected anchors into yaw constraints and update them according to the quadrotor’s predicted state.This mechanism ensures the environmental adaptability of the anchors and avoids sudden feature changes.The effectiveness of the IFA-planner is demonstrated in simulation experiments.展开更多
基金This study was co-supported by the National Natural Science Foundation of China(Nos.61673017 and 61403398).
文摘This paper addresses the problem of visual object tracking for Unmanned Aerial Vehicles(UAVs).Most Siamese trackers are used to regard object tracking as classification and regression problems.However,it is difficult for these trackers to accurately classify in the face of similar objects,background clutters and other common challenges in UAV scenes.So,a reliable classifier is the key to improving UAV tracking performance.In this paper,a simple yet efficient tracker following the basic architecture of the Siamese neural network is proposed,which improves the classification ability from three stages.First,the frequency channel attention module is introduced to enhance the target features via frequency domain learning.Second,a template-guided attention module is designed to promote information exchange between the template branch and the search branch,which can get reliable classification response maps.Third,adaptive cross-entropy loss is proposed to make the tracker focus on hard samples that contribute more to the training process,solving the data imbalance between positive and negative samples.To evaluate the performance of the proposed tracker,comprehensive experiments are conducted on two challenging aerial datasets,including UAV123 and UAVDT.Experimental results demonstrate that the proposed tracker achieves favorable tracking performances in aerial benchmarks beyond 41 frames/s.We conducted experiments in real UAV scenes to further verify the efficiency of our tracker in the real world.
基金supported by the National Key R&D Program of China(2022YFB3902801)the project JSSCBS20230196Shenzhen Science and Technology Program(JCYJ20220818102207015).
文摘The coordination of multiple Earth-observing satellites presents a significant scheduling challenge.This paper introduces a fully distributed autonomous scheduling solution that utilizes a learning-based mechanism through an independent proximal policy optimization(IPPO)algorithm.Each satellite independently makes decisions regarding tasks,such as imaging,desaturation,and charging,while adapting to dynamic environmental changes to enhance its real-time constellation scheduling performance.The proposed fully distributed strategy enables individual satellites to update their policies based solely on their observations.The only requirement is the unidirectional broadcast of a completion flag upon target observation.This approach distinguishes itself from traditional centralized methods,thus enhancing the overall robustness and security of the system.In simulations,our strategy exhibited effective observational mission planning results for major cities worldwide.The results show that the proposed method addresses both autonomous scheduling and significantly improves constellation performance and reliability.
文摘The propulsion systems of a multi-rotor unmanned aerial vehicle(UAV)is crucial,as it directly affects the UAV’s performance,efficiency,and safety.Since the components of the UAV propulsion system are highly interconnectioned,we developed a fuzzy fault tree analysis method to analysis the varying reliability under different fault conditions.Combining the fuzzy fault tree analysis of the T-S model and the UAV propulsion system model,we constructed a fuzzy fault tree of the T-S type for the system and performed a reliability analysis.This fuzzy fault tree allows us to model the system from two perspectives:fuzzy failure rate and failure degree.Consequently,two methods can be used for failure analysis of UAV systems.The first method involves calculating the system’s fuzzy failure rate based on the component’s fuzzy failure rate.The second method calculates the fuzzy failure rate of the system based on the failure degree of the component.The computational results indicate that both methods are well-suited for fault diagnosis in UAV propulsion systems.Compared to traditional fault tree analysis,which does not subdivide fault degrees,the proposed methods provide more accurate fault rate assessments.
基金partially supported by the National Key R&D Program of China(2022YFB3902801)Fundamental Research Funds for the Central Universities(No.JUSRP123063).
文摘This paper proposes an intermittent measurement-based attitude tracking control strategy for spacecraft operating in the presence of sensor-actuator faults.Asampled-data(self-)learning observer is developed to estimate both the spacecraft’s states and lumped disturbances,effectively mitigating the impact of faults.This observer acts as a virtual predictor,reconstructing states and actuator fault deviations using only intermittent measurement data,addressing the limitations imposed by sensor failures.The control scheme incorporates compensation based on the predictor’s estimates,ensuring robust attitude tracking despite the presence of faults.We provide the first proof of bounded stability for this learning observer utilizing intermittent information,expanding its applicability.Numerical simulations demonstrate the effectiveness of this innovative strategy,highlighting its potential for enhancing spacecraft autonomy and reliability in challenging operational scenarios.
基金supported by the National Key R&D Program of China(2022YFB3902801)the Fundamental Research Funds for the Central Universities(No.JUSRP123063)111 Project(B23008).
文摘This paper presents an image-feature-aware(IFA)planner for quadrotors,which integrates image feature tracking into its path-planning framework.The IFA-planner aims to improve the visual localization performance of quadrotors in multifarious environments where feature points may be sparse or diverse.Unlike traditional methods that decouple visual localization and path planning,the IFA-planner adaptively identifies and tracks feature-rich spatial units,called anchors,along a feasible path.The anchors provide additional feature points to the visual localization module,especially in scenarios with sparse or uneven features,thus enhancing localization robustness.Via clustering-based method,the anchor selection can handle different feature point distributions without manual tuning.Moreover,a detachment prediction mechanism is incorporated to convert the selected anchors into yaw constraints and update them according to the quadrotor’s predicted state.This mechanism ensures the environmental adaptability of the anchors and avoids sudden feature changes.The effectiveness of the IFA-planner is demonstrated in simulation experiments.
