Qingran Su
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Qingran Su has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Yongfeng Yin
School of Software, Beihang University, Beijing 100191, China
10 shared publications - Ruinan Qiu
School of Software, Beihang University, Haidian, Beijing 100191, China
3 shared publications Yongfeng Yin
School of Software Beihang University Beijing China
3 shared publicationsBing Fu
School of Software Beihang University Beijing China
2 shared publications- Bing Fu
School of Software, Beihang University, No. 37 Xueyuan Road, Haidian District, Beijing 100191, China
2 shared publications Chunming Hu
School of Software Beihang University Beijing China
2 shared publications
Latest Publications
- 1 September 2026
- Published by Zhejiang University Press in ENGINEERING Information Technology & Electronic Engineering
- Vol. 27 (9) , 1-16
Dynamic Trend Perception for Fine-Grained Unmanned Aerial Vehicle Anomaly Detection in Flight Data
- 1 September 2026
- Published by Zhejiang University Press in ENGINEERING Information Technology & Electronic Engineering
- Vol. 27 (9) , 1-16
Dynamic Trend Perception for Fine-Grained Unmanned Aerial Vehicle Anomaly Detection in Flight Data
- Open Access
- 21 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2420
MSGMamba: A Multi-Scale Dynamic Graph State-Space Model for Satellite Telemetry Anomaly Detection
Satellites are critical components of modern space information systems. During long-term on-orbit operation, satellite telemetry often exhibits multi-scale temporal dynamics, heterogeneous channel behavior, and time-varying inter-variable dependencies, which pose substantial challenges to anomaly detection. Existing methods remain limited in adaptively representing anomaly patterns across temporal scales, jointly modeling temporal evolution and dynamic asymmetric channel dependencies, and preventing over-generalized reconstruction of anomalous inputs. To address these limitations, this paper proposes MSGMamba, a multi-scale graph state space model for satellite telemetry anomaly detection. First, a multi-scale temporal patch decomposition and gated fusion mechanism partitions telemetry sequences into patches of different granularities and adaptively integrates their representations at each temporal position, enabling the joint modeling of short-term transients and relatively slow-varying patterns. Second, a graph–sequence alternating propagation mechanism couples selective state space updates with dynamic graph interaction. At each temporal patch, a directed and asymmetric dependency graph with self-connection priors is generated from the temporally encoded features, allowing temporal evolution and time-varying cross-channel dependencies to be modeled within a unified framework. Third, an orthogonal memory-augmented anomaly discrimination mechanism introduces an orthogonality-constrained memory bank to reduce redundancy among nominal prototypes and constrain the reconstruction space. A dual-pathway anomaly score further combines signal-space reconstruction error with encoder–memory discrepancy to improve the separability of nominal and anomalous samples. Experiments on the SMAP, MSL, and EIRSAT-1 datasets show that MSGMamba outperforms representative baseline methods in terms of average PA-F1 and AFF-F1.
- 21 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2420
Open AccessMSGMamba: A Multi-Scale Dynamic Graph State-Space Model for Satellite Telemetry Anomaly Detection
Satellites are critical components of modern space information systems. During long-term on-orbit operation, satellite telemetry often exhibits multi-scale temporal dynamics, heterogeneous channel behavior, and time-varying inter-variable dependencies, which pose substantial challenges to anomaly detection. Existing methods remain limited in adaptively representing anomaly patterns across temporal scales, jointly modeling temporal evolution and dynamic asymmetric channel dependencies, and preventing over-generalized reconstruction of anomalous inputs. To address these limitations, this paper proposes MSGMamba, a multi-scale graph state space model for satellite telemetry anomaly detection. First, a multi-scale temporal patch decomposition and gated fusion mechanism partitions telemetry sequences into patches of different granularities and adaptively integrates their representations at each temporal position, enabling the joint modeling of short-term transients and relatively slow-varying patterns. Second, a graph–sequence alternating propagation mechanism couples selective state space updates with dynamic graph interaction. At each temporal patch, a directed and asymmetric dependency graph with self-connection priors is generated from the temporally encoded features, allowing temporal evolution and time-varying cross-channel dependencies to be modeled within a unified framework. Third, an orthogonal memory-augmented anomaly discrimination mechanism introduces an orthogonality-constrained memory bank to reduce redundancy among nominal prototypes and constrain the reconstruction space. A dual-pathway anomaly score further combines signal-space reconstruction error with encoder–memory discrepancy to improve the separability of nominal and anomalous samples. Experiments on the SMAP, MSL, and EIRSAT-1 datasets show that MSGMamba outperforms representative baseline methods in terms of average PA-F1 and AFF-F1.
- Top Cited PapersOpen Access
- 16 February 2026
- Published by Wiley in Computational Intelligence
- Vol. 42 (1)
LeF‐MTP: Prioritizing GNN Test Cases by Fusing Model Uncertainty and Feature‐Space Confusability
Graph Neural Networks (GNNs) are widely applied in 3D computer vision tasks. However, validating their reliability is limited by the high cost of labeling large‐scale test data. Existing test case prioritization techniques for deep learning often rely on single‐dimensional heuristics or mutation‐based analysis (e.g., GraphPrior, PCPrior). However, these methods often limit fault diversity or incur prohibitive computational overhead. To address these issues, this paper proposes LeF‐MTP, a multi‐dimensional test prioritization framework. Unlike these state‐of‐the‐art methods that depend on computationally intensive external mutations, LeF‐MTP directly exploits the model's internal cognitive states. The method characterizes test cases through two dimensions: the predictive uncertainty and feature‐space confusability. These cognitive features are fused via a Random Forest classifier to predict failure probabilities. The resulting priority list is then optimized by a diversity‐aware re‐ranking process based on feature clustering. We evaluated the framework on four 3D point cloud datasets (including ScanObjectNN and S3DIS) across four GNN architectures. Experimental results demonstrate that LeF‐MTP achieves higher fault detection rates than heuristic and mutation‐based baselines, with Average Percentage of Fault Detection (APFD) improvements ranging from 8.79% to 74.69%. Moreover, an efficiency analysis demonstrates that the proposed approach reduces feature generation time compared to mutation‐based methods, validating its practicality for GNN testing.
- 16 February 2026
- Published by Wiley in Computational Intelligence
- Vol. 42 (1)
Top Cited PapersOpen AccessLeF‐MTP: Prioritizing GNN Test Cases by Fusing Model Uncertainty and Feature‐Space Confusability
Graph Neural Networks (GNNs) are widely applied in 3D computer vision tasks. However, validating their reliability is limited by the high cost of labeling large‐scale test data. Existing test case prioritization techniques for deep learning often rely on single‐dimensional heuristics or mutation‐based analysis (e.g., GraphPrior, PCPrior). However, these methods often limit fault diversity or incur prohibitive computational overhead. To address these issues, this paper proposes LeF‐MTP, a multi‐dimensional test prioritization framework. Unlike these state‐of‐the‐art methods that depend on computationally intensive external mutations, LeF‐MTP directly exploits the model's internal cognitive states. The method characterizes test cases through two dimensions: the predictive uncertainty and feature‐space confusability. These cognitive features are fused via a Random Forest classifier to predict failure probabilities. The resulting priority list is then optimized by a diversity‐aware re‐ranking process based on feature clustering. We evaluated the framework on four 3D point cloud datasets (including ScanObjectNN and S3DIS) across four GNN architectures. Experimental results demonstrate that LeF‐MTP achieves higher fault detection rates than heuristic and mutation‐based baselines, with Average Percentage of Fault Detection (APFD) improvements ranging from 8.79% to 74.69%. Moreover, an efficiency analysis demonstrates that the proposed approach reduces feature generation time compared to mutation‐based methods, validating its practicality for GNN testing.
- Open Access
Thinking Like an Expert: Aligning LLM Thought Processes for Automated Safety Modeling of High-Altitude Solar Drones
As the application of high-altitude solar drones expands, ensuring their safety is paramount. Traditional safety modeling, which relies on manual expert analysis, struggles to keep pace with rapid development cycles. While Large Language Models (LLMs) offer a path to automation, state-of-the-art reasoning frameworks like Graph of Thoughts (GoT) are too generic, lacking the domain-specific knowledge required for effective application. To address this gap, we introduce K-EGoT, a framework that grounds LLM reasoning in a verifiable, domain-specific knowledge base. Our method introduces a “Safety Rationale”—a mandatory, auditable link between LLM-generated model extensions and expert-curated safety principles. We then train a specialized model using a novel “thought process alignment” strategy, applying Direct Preference Optimization (DPO) to the quality of these rationales to ensure the model’sreasoning aligns with expert logic. On a high-fidelity dataset for the flight control–energy coupling problem, our 7B K-EGoT model achieved a Safety Extension Score (SES) of 92.7, significantly outperforming the 84.7 score from standard GoT prompting. Our work delivers a reliable and auditable solution for automated safety modeling for this critical class of drones.
Open AccessThinking Like an Expert: Aligning LLM Thought Processes for Automated Safety Modeling of High-Altitude Solar Drones
As the application of high-altitude solar drones expands, ensuring their safety is paramount. Traditional safety modeling, which relies on manual expert analysis, struggles to keep pace with rapid development cycles. While Large Language Models (LLMs) offer a path to automation, state-of-the-art reasoning frameworks like Graph of Thoughts (GoT) are too generic, lacking the domain-specific knowledge required for effective application. To address this gap, we introduce K-EGoT, a framework that grounds LLM reasoning in a verifiable, domain-specific knowledge base. Our method introduces a “Safety Rationale”—a mandatory, auditable link between LLM-generated model extensions and expert-curated safety principles. We then train a specialized model using a novel “thought process alignment” strategy, applying Direct Preference Optimization (DPO) to the quality of these rationales to ensure the model’sreasoning aligns with expert logic. On a high-fidelity dataset for the flight control–energy coupling problem, our 7B K-EGoT model achieved a Safety Extension Score (SES) of 92.7, significantly outperforming the 84.7 score from standard GoT prompting. Our work delivers a reliable and auditable solution for automated safety modeling for this critical class of drones.
- Open Access
LeF-MTP: Prioritizing GNN Test Cases by Fusing Model Uncertainty and Feature-Space Confusability
- Open Access
- 19 February 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (4) , 2243
Attention Transfer Reinforcement Learning for Test Case Prioritization in Continuous Integration
Continuous integration (CI), a crucial technology for accelerating software delivery, employs prioritization methods to optimize testing efficiency. Reinforcement-learning-based prioritization techniques of test cases can dynamically adjust the prioritization strategy based on test feedback. However, existing reinforcement learning models either use feature information from individual test cases or the complete feature of all test cases as input. The model’s input is fixed and does not vary based on the characteristics of the test case set. This leads to the model’s inability to handle various ranking scenarios in CI. In this paper, we propose a novel Attention Transfer Reinforcement Learning for Test Case Prioritization (ATRL-TCP) method. ATRL-TCP first leverages the richness of historical information and error rates to assess the test case set. It then determines whether the model’s input should be the feature information of a single test case or the relative order between them. This method allows for the shifting of focus based on the dataset characteristics. We validate the method on four industrial datasets and examine the impact of historical information length and error rate on its performance. The results demonstrate that ATRL-TCP has good generalization ability and achieves good ranking performance within an acceptable time.
- 19 February 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (4) , 2243
Open AccessAttention Transfer Reinforcement Learning for Test Case Prioritization in Continuous Integration
Continuous integration (CI), a crucial technology for accelerating software delivery, employs prioritization methods to optimize testing efficiency. Reinforcement-learning-based prioritization techniques of test cases can dynamically adjust the prioritization strategy based on test feedback. However, existing reinforcement learning models either use feature information from individual test cases or the complete feature of all test cases as input. The model’s input is fixed and does not vary based on the characteristics of the test case set. This leads to the model’s inability to handle various ranking scenarios in CI. In this paper, we propose a novel Attention Transfer Reinforcement Learning for Test Case Prioritization (ATRL-TCP) method. ATRL-TCP first leverages the richness of historical information and error rates to assess the test case set. It then determines whether the model’s input should be the feature information of a single test case or the relative order between them. This method allows for the shifting of focus based on the dataset characteristics. We validate the method on four industrial datasets and examine the impact of historical information length and error rate on its performance. The results demonstrate that ATRL-TCP has good generalization ability and achieves good ranking performance within an acceptable time.
- Open Access
- 17 January 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (2) , 905
Func-Bagging: An Ensemble Learning Strategy for Improving the Performance of Heterogeneous Anomaly Detection Models
In the field of ensemble learning, bagging and stacking are two widely used ensemble strategies. Bagging enhances model robustness through repeated sampling and weighted averaging of homogeneous classifiers, while stacking improves classification performance by integrating multiple models using meta-learning strategies, taking advantage of the diversity of heterogeneous classifiers. However, the fixed weight distribution strategy in traditional bagging methods often has limitations when handling complex or imbalanced datasets. This paper combines the concept of heterogeneous classifier integration in stacking with the weighted averaging strategy of bagging, proposing a new adaptive weight distribution approach to enhance bagging’s performance in heterogeneous ensemble settings. Specifically, we propose three weight generation functions with “high at both ends, low in the middle” curve shapes and demonstrate the superiority of this strategy over fixed weight methods on two datasets. Additionally, we design a specialized neural network, and by training it adequately, validate the rationality of the proposed adaptive weight distribution strategy, further improving the model’s robustness. The above methods are collectively called func-bagging. Experimental results show that func-bagging has an average 1.810% improvement in extreme performance compared to the base classifier, and is superior to stacking and bagging methods. It also has better dataset adaptability and interpretability than stacking and bagging. Therefore, func-bagging is particularly effective in scenarios with class imbalance and is applicable to classification tasks with imbalanced classes, such as anomaly detection.
- 17 January 2025
- Published by MDPI AG in Applied Sciences
- Vol. 15 (2) , 905
Open AccessFunc-Bagging: An Ensemble Learning Strategy for Improving the Performance of Heterogeneous Anomaly Detection Models
In the field of ensemble learning, bagging and stacking are two widely used ensemble strategies. Bagging enhances model robustness through repeated sampling and weighted averaging of homogeneous classifiers, while stacking improves classification performance by integrating multiple models using meta-learning strategies, taking advantage of the diversity of heterogeneous classifiers. However, the fixed weight distribution strategy in traditional bagging methods often has limitations when handling complex or imbalanced datasets. This paper combines the concept of heterogeneous classifier integration in stacking with the weighted averaging strategy of bagging, proposing a new adaptive weight distribution approach to enhance bagging’s performance in heterogeneous ensemble settings. Specifically, we propose three weight generation functions with “high at both ends, low in the middle” curve shapes and demonstrate the superiority of this strategy over fixed weight methods on two datasets. Additionally, we design a specialized neural network, and by training it adequately, validate the rationality of the proposed adaptive weight distribution strategy, further improving the model’s robustness. The above methods are collectively called func-bagging. Experimental results show that func-bagging has an average 1.810% improvement in extreme performance compared to the base classifier, and is superior to stacking and bagging methods. It also has better dataset adaptability and interpretability than stacking and bagging. Therefore, func-bagging is particularly effective in scenarios with class imbalance and is applicable to classification tasks with imbalanced classes, such as anomaly detection.
- Open Access
- 25 December 2024
- Published by MDPI AG in Preprints.org
Func-bagging: An Ensemble Learning Strategy for Improving the Performance of Heterogeneous Anomaly Detection Models
In the field of ensemble learning, bagging and stacking are two widely used ensemble strategies. Bagging enhances model robustness through repeated sampling and weighted averaging of homogeneous classifiers, while stacking improves classification performance by integrating multiple models using meta-learning strategies, taking advantage of the diversity of heterogeneous classifiers. However, the fixed weight distribution strategy in traditional bagging methods often has limitations when handling complex or imbalanced datasets. This paper combines the concept of heterogeneous classifier integration in stacking with the weighted averaging strategy of bagging, proposing a new adaptive weight distribution approach to enhance bagging's performance in heterogeneous ensemble settings. Specifically, we propose three weight generation functions with "high at both ends, low in the middle" curve shapes, and demonstrate the superiority of this strategy over fixed weight methods on two datasets. Additionally, we design a specialized neural network, and by training it adequately, validate the rationality of the proposed adaptive weight distribution strategy, further improving the model's robustness. Experimental results show that this method is particularly effective in scenarios with class imbalance and is applicable to classification tasks with imbalanced classes, such as anomaly detection.
- 25 December 2024
- Published by MDPI AG in Preprints.org
Open AccessFunc-bagging: An Ensemble Learning Strategy for Improving the Performance of Heterogeneous Anomaly Detection Models
In the field of ensemble learning, bagging and stacking are two widely used ensemble strategies. Bagging enhances model robustness through repeated sampling and weighted averaging of homogeneous classifiers, while stacking improves classification performance by integrating multiple models using meta-learning strategies, taking advantage of the diversity of heterogeneous classifiers. However, the fixed weight distribution strategy in traditional bagging methods often has limitations when handling complex or imbalanced datasets. This paper combines the concept of heterogeneous classifier integration in stacking with the weighted averaging strategy of bagging, proposing a new adaptive weight distribution approach to enhance bagging's performance in heterogeneous ensemble settings. Specifically, we propose three weight generation functions with "high at both ends, low in the middle" curve shapes, and demonstrate the superiority of this strategy over fixed weight methods on two datasets. Additionally, we design a specialized neural network, and by training it adequately, validate the rationality of the proposed adaptive weight distribution strategy, further improving the model's robustness. Experimental results show that this method is particularly effective in scenarios with class imbalance and is applicable to classification tasks with imbalanced classes, such as anomaly detection.
- Top Cited PapersOpen Access
- 21 June 2024
- Published by MDPI AG in Electronics
- Vol. 13 (13) , 2432
Autonomous UAV Navigation with Adaptive Control Based on Deep Reinforcement Learning
Unmanned aerial vehicle (UAV) navigation plays a crucial role in its ability to perform autonomous missions in complex environments. Most of the existing reinforcement learning methods to solve the UAV navigation problem fix the flight altitude and velocity, which largely reduces the difficulty of the algorithm. But the methods without adaptive control are not suitable in low-altitude environments with complex situations, generally suffering from weak obstacle avoidance. Some UAV navigation studies with adaptive flight only have weak obstacle avoidance capabilities. To address the problem of UAV navigation in low-altitude environments, we construct autonomous UAV navigation in 3D environments with adaptive control as a Markov decision process and propose a deep reinforcement learning algorithm. To solve the problem of weak obstacle avoidance, we creatively propose the guide attention method to make a UAV’s decision focus shift between the navigation task and obstacle avoidance task according to changes in the obstacle. We raise a novel velocity-constrained loss function and add it to the original actor loss to improve the UAV’s velocity control capability. Simulation experiment results demonstrate that our algorithm outperforms some of the state-of-the-art deep reinforcement learning algorithms performing UAV navigation tasks in a 3D environment and has outstanding performance in algorithm effectiveness, with the average reward increasing by 9.35, the success rate of navigation tasks increasing by 14%, and the collision rate decreasing by 14%.
- 21 June 2024
- Published by MDPI AG in Electronics
- Vol. 13 (13) , 2432
Top Cited PapersOpen AccessAutonomous UAV Navigation with Adaptive Control Based on Deep Reinforcement Learning
Unmanned aerial vehicle (UAV) navigation plays a crucial role in its ability to perform autonomous missions in complex environments. Most of the existing reinforcement learning methods to solve the UAV navigation problem fix the flight altitude and velocity, which largely reduces the difficulty of the algorithm. But the methods without adaptive control are not suitable in low-altitude environments with complex situations, generally suffering from weak obstacle avoidance. Some UAV navigation studies with adaptive flight only have weak obstacle avoidance capabilities. To address the problem of UAV navigation in low-altitude environments, we construct autonomous UAV navigation in 3D environments with adaptive control as a Markov decision process and propose a deep reinforcement learning algorithm. To solve the problem of weak obstacle avoidance, we creatively propose the guide attention method to make a UAV’s decision focus shift between the navigation task and obstacle avoidance task according to changes in the obstacle. We raise a novel velocity-constrained loss function and add it to the original actor loss to improve the UAV’s velocity control capability. Simulation experiment results demonstrate that our algorithm outperforms some of the state-of-the-art deep reinforcement learning algorithms performing UAV navigation tasks in a 3D environment and has outstanding performance in algorithm effectiveness, with the average reward increasing by 9.35, the success rate of navigation tasks increasing by 14%, and the collision rate decreasing by 14%.
- Open Access
Research on Drone Fault Detection Based on Failure Mode Databases
Drones are widely used in a number of key fields and are having a profound impact on all walks of life. Working out how to improve drone safety through fault detection is key to ensuring the smooth execution of tasks. At present, most research focuses on fault detection at the component level as it is not possible to locate faults quickly from the global system state of a UAV. Moreover, most methods are offline detection methods, which cannot achieve real-time monitoring of UAV faults. To remedy this, this paper proposes a fault detection method based on a fault mode database and runtime verification. Firstly, a large body of historical fault information is analyzed to generate a summary of fault modes, including fault modes at the system level. The key safety properties of UAVs during operation are further studied in terms of system-level fault modes. Next, a monitor generation algorithm and code instrumentation framework are designed to monitor whether a certain safety attribute is violated during the operation of a UAV in real time. The experimental results show that the fault detection method proposed in this paper can detect abnormal situations in a timely and accurate manner.
Open AccessResearch on Drone Fault Detection Based on Failure Mode Databases
Drones are widely used in a number of key fields and are having a profound impact on all walks of life. Working out how to improve drone safety through fault detection is key to ensuring the smooth execution of tasks. At present, most research focuses on fault detection at the component level as it is not possible to locate faults quickly from the global system state of a UAV. Moreover, most methods are offline detection methods, which cannot achieve real-time monitoring of UAV faults. To remedy this, this paper proposes a fault detection method based on a fault mode database and runtime verification. Firstly, a large body of historical fault information is analyzed to generate a summary of fault modes, including fault modes at the system level. The key safety properties of UAVs during operation are further studied in terms of system-level fault modes. Next, a monitor generation algorithm and code instrumentation framework are designed to monitor whether a certain safety attribute is violated during the operation of a UAV in real time. The experimental results show that the fault detection method proposed in this paper can detect abnormal situations in a timely and accurate manner.