The uplift resistance of the soil overlying shield tunnels significantly impacts their anti-floating stability.However,research on uplift resistance concerning special-shaped shield tunnels is limited.This study combi...The uplift resistance of the soil overlying shield tunnels significantly impacts their anti-floating stability.However,research on uplift resistance concerning special-shaped shield tunnels is limited.This study combines numerical simulation with machine learning techniques to explore this issue.It presents a summary of special-shaped tunnel geometries and introduces a shape coefficient.Through the finite element software,Plaxis3D,the study simulates six key parameters—shape coefficient,burial depth ratio,tunnel’s longest horizontal length,internal friction angle,cohesion,and soil submerged bulk density—that impact uplift resistance across different conditions.Employing XGBoost and ANN methods,the feature importance of each parameter was analyzed based on the numerical simulation results.The findings demonstrate that a tunnel shape more closely resembling a circle leads to reduced uplift resistance in the overlying soil,whereas other parameters exhibit the contrary effects.Furthermore,the study reveals a diminishing trend in the feature importance of buried depth ratio,internal friction angle,tunnel longest horizontal length,cohesion,soil submerged bulk density,and shape coefficient in influencing uplift resistance.展开更多
Cyclohexene is an important raw material in the production of nylon.Selective hydrogenation of benzene is a key method for preparing cyclohexene.However,the Ru catalysts used in current industrial processes still face...Cyclohexene is an important raw material in the production of nylon.Selective hydrogenation of benzene is a key method for preparing cyclohexene.However,the Ru catalysts used in current industrial processes still face challenges,including high metal usage,high process costs,and low cyclohexene yield.This study utilizes existing literature data combined with machine learning methods to analyze the factors influencing benzene conversion,cyclohexene selectivity,and yield in the benzene hydrogenation to cyclohexene reaction.It constructs predictive models based on XGBoost and Random Forest algorithms.After analysis,it was found that reaction time,Ru content,and space velocity are key factors influencing cyclohexene yield,selectivity,and benzene conversion.Shapley Additive Explanations(SHAP)analysis and feature importance analysis further revealed the contribution of each variable to the reaction outcomes.Additionally,we randomly generated one million variable combinations using the Dirichlet distribution to attempt to predict high-yield catalyst formulations.This paper provides new insights into the application of machine learning in heterogeneous catalysis and offers some reference for further research.展开更多
The electrocatalytic reduction of nitric oxide for ammonia synthesis(NORR)is a key green energy conversion technology.Its efficiency relies on high-performance electrocatalysts to enhance both ammonia yield(Y_(NH3))an...The electrocatalytic reduction of nitric oxide for ammonia synthesis(NORR)is a key green energy conversion technology.Its efficiency relies on high-performance electrocatalysts to enhance both ammonia yield(Y_(NH3))and Faradaic efficiency(F_(NH3)).However,conventional experimental methods for screening high-activity NORR catalysts often entail high resource consumption and time costs.Machine learning combined with SHAP feature analysis was employed to establish a stacked ensemble model that integrates multiple algorithms,to allow for a systematic investigation of the key descriptors governing NORR performance based on an experimental dataset.Evaluation of eight model algorithms revealed that the Stacked-SVR model achieved an R^(2)of 0.9223 and an RMSE of 0.0608 for predicting on the test set,whereas the Stacked-RF model achieved an R^(2)of 0.9042 and an RMSE of 0.0900 for predicting.The stacked ensemble model integrates the strengths of individual algorithms and demonstrates strong NORR prediction performance while avoiding overfitting.SHAP feature analysis results revealed that the Cu content in the catalyst composition has the most significant impact on catalytic performance.Moreover,the combination of the wet chemical reduction synthesis,a carbon fiber(CF)conductive substrate,and HCl electrolyte is more favorable for enhancing catalytic activity.Additionally,moderately lowering the working potential,controlling the electrolyte volume at low to medium levels,reducing catalyst loading,and increasing electrolyte concentration were found to synergistically enhance both and.展开更多
Nonlinear classification models are widely used in various fields due to their excellent performance in handling complex problems.This paper investigates the learning performance of nonlinear classification models bas...Nonlinear classification models are widely used in various fields due to their excellent performance in handling complex problems.This paper investigates the learning performance of nonlinear classification models based on Markov sampling,which builds upon the traditional framework using i.i.d.samples.Subsequently,we introduce a ueMC-NL algorithm,tailored specifically for nonlinear classification models,facilitating the production of ueMC samples from a finite dataset.Numerical investigations on the random forest and the MLP model reveal that nonlinear classification models utilizing ueMC samples yield lower misclassification rates compared to i.i.d.samples.展开更多
The application and promotion of waste glass powder concrete(WGPC)cansignificantly alleviate the pressure of concrete material scarcity and environmental pollution.Compressive strength(CS)is a critical parameter for e...The application and promotion of waste glass powder concrete(WGPC)cansignificantly alleviate the pressure of concrete material scarcity and environmental pollution.Compressive strength(CS)is a critical parameter for evaluating the efficacy of WGPC.Unlike conventional testing methods,machine learning techniques offer precise and reliable predictions of concrete’s compressive strength,especially in its long-term mechanical properties.In this work,four models,namely Multiple Linear Regression(MLR),Back Propagation Neural Network(BPNN),Support Vector Regression(SVR),and Random Forest Regression(RFR)were employed.Furthermore,particle swarm optimization(PSO)algorithm and cross-validation techniques were applied to fine-tune the model parameters,striving for peak prediction performance.The results indicated that optimized models generally exhibit enhanced predictive accuracy compared to their basic counterparts.Notably,the PSO-RFR model excels among all evaluated models,showcasing superior performance on the testing dataset.It achieves a coefficient of determination(R^(2))of 0.9231,a mean absolute error(MAE)of 2.1073,and a root mean square error(RMSE)of 3.6903.When compared to experimental results,the PSO-RFR and PSO-BPNN models demonstrate exceptional predictive accuracy.Notably,the PSO-BPNN model exhibits the closest R^(2)values between its training and test sets.This close alignment of R^(2)values between the training and testing sets reflects the PSO-BPNN model’s superior generalization ability for unseen data.The findings present an efficient method for predicting concrete’s compressive strength,contributing to the sustainable development of concrete materials,and providing theoretical support for their research and application.展开更多
Accurate prediction of rockburst intensity levels is crucial for ensuring the safety of deep hard rock engineering construction.This paper introduced an expert system for rockburst intensity level prediction that empl...Accurate prediction of rockburst intensity levels is crucial for ensuring the safety of deep hard rock engineering construction.This paper introduced an expert system for rockburst intensity level prediction that employs machine learning algorithms as the basis for its inference rules.The system comprises four modules:a database,a repository,an inference engine,and an interpreter.A database containing 1114 rockburst cases was used to construct 357 datasets that serve as the repository for the expert system.Additionally,19 types of machine learning algorithms were used to establish 6783 micro-models to construct cognitive rules within the inference engine.By integrating probability theory and marginal analysis,a fuzzy scoring method based on the SoftMax function was developed and applied to the interpreter for rockburst intensity level prediction,effectively restoring the continuity of rockburst characteristics.The research results indicate that ensemble algorithms based on decision trees are more effective in capturing the characteristics of rockburst.Key factors for accurate prediction of rockburst intensity include uniaxial compressive strength,elastic energy index,the maximum principal stress,tangential stress,and their composite indicators.The accuracy of the proposed rockburst intensity level prediction expert system was verified using 20 engineering rockburst cases,with predictions aligning closely with the actual rockburst intensity levels.展开更多
Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-onl...Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-only approaches.To address this issue,this paper proposes a framework named“a lightweight pure visual BEV perception method based on dual distillation of spatial-temporal knowledge”.This framework innovatively designs a lightweight vision-only student model based on Res Net,which leverages a dual distillation mechanism to learn from a powerful teacher model that integrates temporal information from both image and light detection and ranging(LiDAR)modalities.Specifically,we distill efficient multi-modal feature extraction and spatial fusion capabilities from the BEVFusion model,and distill advanced temporal information fusion and spatiotemporal attention mechanisms from the BEVFormer model.This dual distillation strategy enables the student model to achieve perception performance close to that of multi-modal models without relying on Li DAR.Experimental results on the nu Scenes dataset demonstrate that the proposed model significantly outperforms classical vision-only algorithms,achieves comparable performance to current state-of-the-art vision-only methods on the nu Scenes detection leaderboard in terms of both mean average precision(mAP)and the nu Scenes detection score(NDS)metrics,and exhibits notable advantages in inference computational efficiency.Although the proposed dual-teacher paradigm incurs higher offline training costs compared to single-model approaches,it yields a streamlined and highly efficient student model suitable for resource-constrained real-time deployment.This provides an effective pathway toward low-cost,high-performance autonomous driving perception systems.展开更多
As joint operations have become a key trend in modern military development,unmanned aerial vehicles(UAVs)play an increasingly important role in enhancing the intelligence and responsiveness of combat systems.However,t...As joint operations have become a key trend in modern military development,unmanned aerial vehicles(UAVs)play an increasingly important role in enhancing the intelligence and responsiveness of combat systems.However,the heterogeneity of aircraft,partial observability,and dynamic uncertainty in operational airspace pose significant challenges to autonomous collision avoidance using traditional methods.To address these issues,this paper proposes an adaptive collision avoidance approach for UAVs based on deep reinforcement learning.First,a unified uncertainty model incorporating dynamic wind fields is constructed to capture the complexity of joint operational environments.Then,to effectively handle the heterogeneity between manned and unmanned aircraft and the limitations of dynamic observations,a sector-based partial observation mechanism is designed.A Dynamic Threat Prioritization Assessment algorithm is also proposed to evaluate potential collision threats from multiple dimensions,including time to closest approach,minimum separation distance,and aircraft type.Furthermore,a Hierarchical Prioritized Experience Replay(HPER)mechanism is introduced,which classifies experience samples into high,medium,and low priority levels to preferentially sample critical experiences,thereby improving learning efficiency and accelerating policy convergence.Simulation results show that the proposed HPER-D3QN algorithm outperforms existing methods in terms of learning speed,environmental adaptability,and robustness,significantly enhancing collision avoidance performance and convergence rate.Finally,transfer experiments on a high-fidelity battlefield airspace simulation platform validate the proposed method's deployment potential and practical applicability in complex,real-world joint operational scenarios.展开更多
Investigating the detonation reaction zone structures of high explosives is significant for understanding detonation reaction mechanism.This study employed an integrated approach combining machine learning prediction,...Investigating the detonation reaction zone structures of high explosives is significant for understanding detonation reaction mechanism.This study employed an integrated approach combining machine learning prediction,theoretical calculation,and experimental characterization to determine the detonation reaction zone width of CL-20-based aluminized explosive.In this study,the detonation reaction zone refers to the reaction zone between the von Neumann(VN)peak and sonic point,which usually means the so-called detonation driving zone(DDZ).For the machine learning prediction,an ensemble model integrating Random Forest and Support Vector Regression was developed to predict the reaction zone width using a dataset of 19 publicly available samples.For the theoretical calculation,the Wood-Kirkwood(W-K)detonation theory model was utilized to implement numerical calculation of the reaction zone structures,incorporating chemical reaction kinetics to describe the detonation reaction progress.In experimental characterization,the Photon Doppler Velocimetry(PDV)was applied with LiF as the optical window to measure the particle velocity profile of detonation products and derive the reaction zone width.The results indicate that the reaction zone width values are 0.25 mm,0.28 mm,and 0.26 mm obtained from machine learning prediction,theoretical calculation,and experimental characterization,respectively.The corresponding velocities at the Chapman-Jouguet(CJ)point are 1,938 m/s,2,047 m/s,and 1,982 m/s,respectively.The maximum relative deviation in reaction zone width among three methods is approximately 7.7%,while that for CJ particle velocity is approximately 3.3%.These results from all three methods agree well within engineering error.This validates the effectiveness of integrating machine learning prediction,theoretical calculation and advanced experimental techniques for studying the detonation reaction zone structures of high explosives.This research provides insights into the detonation reaction mechanism and reaction zone characteristics of CL-20-based aluminized explosive.展开更多
In this paper, the reinforcement learning method for cooperative multi-agent systems(MAS) with incremental number of agents is studied. The existing multi-agent reinforcement learning approaches deal with the MAS with...In this paper, the reinforcement learning method for cooperative multi-agent systems(MAS) with incremental number of agents is studied. The existing multi-agent reinforcement learning approaches deal with the MAS with a specific number of agents, and can learn well-performed policies. However, if there is an increasing number of agents, the previously learned in may not perform well in the current scenario. The new agents need to learn from scratch to find optimal policies with others,which may slow down the learning speed of the whole team. To solve that problem, in this paper, we propose a new algorithm to take full advantage of the historical knowledge which was learned before, and transfer it from the previous agents to the new agents. Since the previous agents have been trained well in the source environment, they are treated as teacher agents in the target environment. Correspondingly, the new agents are called student agents. To enable the student agents to learn from the teacher agents, we first modify the input nodes of the networks for teacher agents to adapt to the current environment. Then, the teacher agents take the observations of the student agents as input, and output the advised actions and values as supervising information. Finally, the student agents combine the reward from the environment and the supervising information from the teacher agents, and learn the optimal policies with modified loss functions. By taking full advantage of the knowledge of teacher agents, the search space for the student agents will be reduced significantly, which can accelerate the learning speed of the holistic system. The proposed algorithm is verified in some multi-agent simulation environments, and its efficiency has been demonstrated by the experiment results.展开更多
Tracking and analyzing data from research projects is critical for understanding research trends and supporting the development of science and technology strategies.However,the data from these projects is often comple...Tracking and analyzing data from research projects is critical for understanding research trends and supporting the development of science and technology strategies.However,the data from these projects is often complex and inadequate,making it challenging for researchers to conduct in-depth data mining to improve policies or management.To address this problem,this paper adopts a top-down approach to construct a knowledge graph(KG)for research projects.Firstly,we construct an integrated ontology by referring to the metamodel of various architectures,which is called the meta-model integration conceptual reference model.Subsequently,we use the dependency parsing method to extract knowledge from unstructured textual data and use the entity alignment method based on weakly supervised learning to classify the extracted entities,completing the construction of the KG for the research projects.In addition,a knowledge inference model based on representation learning is employed to achieve knowledge completion and improve the KG.Finally,experiments are conducted on the KG for research projects and the results demonstrate the effectiveness of the proposed method in enriching incomplete data within the KG.展开更多
Recently, high-precision trajectory prediction of ballistic missiles in the boost phase has become a research hotspot. This paper proposes a trajectory prediction algorithm driven by data and knowledge(DKTP) to solve ...Recently, high-precision trajectory prediction of ballistic missiles in the boost phase has become a research hotspot. This paper proposes a trajectory prediction algorithm driven by data and knowledge(DKTP) to solve this problem. Firstly, the complex dynamics characteristics of ballistic missile in the boost phase are analyzed in detail. Secondly, combining the missile dynamics model with the target gravity turning model, a knowledge-driven target three-dimensional turning(T3) model is derived. Then, the BP neural network is used to train the boost phase trajectory database in typical scenarios to obtain a datadriven state parameter mapping(SPM) model. On this basis, an online trajectory prediction framework driven by data and knowledge is established. Based on the SPM model, the three-dimensional turning coefficients of the target are predicted by using the current state of the target, and the state of the target at the next moment is obtained by combining the T3 model. Finally, simulation verification is carried out under various conditions. The simulation results show that the DKTP algorithm combines the advantages of data-driven and knowledge-driven, improves the interpretability of the algorithm, reduces the uncertainty, which can achieve high-precision trajectory prediction of ballistic missile in the boost phase.展开更多
This work proposes the application of an iterative learning model predictive control(ILMPC)approach based on an adaptive fault observer(FOBILMPC)for fault-tolerant control and trajectory tracking in air-breathing hype...This work proposes the application of an iterative learning model predictive control(ILMPC)approach based on an adaptive fault observer(FOBILMPC)for fault-tolerant control and trajectory tracking in air-breathing hypersonic vehicles.In order to increase the control amount,this online control legislation makes use of model predictive control(MPC)that is based on the concept of iterative learning control(ILC).By using offline data to decrease the linearized model’s faults,the strategy may effectively increase the robustness of the control system and guarantee that disturbances can be suppressed.An adaptive fault observer is created based on the suggested ILMPC approach in order to enhance overall fault tolerance by estimating and compensating for actuator disturbance and fault degree.During the derivation process,a linearized model of longitudinal dynamics is established.The suggested ILMPC approach is likely to be used in the design of hypersonic vehicle control systems since numerical simulations have demonstrated that it can decrease tracking error and speed up convergence when compared to the offline controller.展开更多
基金Guangzhou Metro Scientific Research Project(No.JT204-100111-23001)Chongqing Municipal Special Project for Technological Innovation and Application Development(No.CSTB2022TIAD-KPX0101)Science and Technology Research and Development Program of China State Railway Group Co.,Ltd.(No.N2023G045)。
文摘The uplift resistance of the soil overlying shield tunnels significantly impacts their anti-floating stability.However,research on uplift resistance concerning special-shaped shield tunnels is limited.This study combines numerical simulation with machine learning techniques to explore this issue.It presents a summary of special-shaped tunnel geometries and introduces a shape coefficient.Through the finite element software,Plaxis3D,the study simulates six key parameters—shape coefficient,burial depth ratio,tunnel’s longest horizontal length,internal friction angle,cohesion,and soil submerged bulk density—that impact uplift resistance across different conditions.Employing XGBoost and ANN methods,the feature importance of each parameter was analyzed based on the numerical simulation results.The findings demonstrate that a tunnel shape more closely resembling a circle leads to reduced uplift resistance in the overlying soil,whereas other parameters exhibit the contrary effects.Furthermore,the study reveals a diminishing trend in the feature importance of buried depth ratio,internal friction angle,tunnel longest horizontal length,cohesion,soil submerged bulk density,and shape coefficient in influencing uplift resistance.
基金Supported by CAS Basic and Interdisciplinary Frontier Scientific Research Pilot Project(XDB1190300,XDB1190302)Youth Innovation Promotion Association CAS(Y2021056)+1 种基金Joint Fund of the Yulin University and the Dalian National Laboratory for Clean Energy(YLU-DNL Fund 2022007)The special fund for Science and Technology Innovation Teams of Shanxi Province(202304051001007)。
文摘Cyclohexene is an important raw material in the production of nylon.Selective hydrogenation of benzene is a key method for preparing cyclohexene.However,the Ru catalysts used in current industrial processes still face challenges,including high metal usage,high process costs,and low cyclohexene yield.This study utilizes existing literature data combined with machine learning methods to analyze the factors influencing benzene conversion,cyclohexene selectivity,and yield in the benzene hydrogenation to cyclohexene reaction.It constructs predictive models based on XGBoost and Random Forest algorithms.After analysis,it was found that reaction time,Ru content,and space velocity are key factors influencing cyclohexene yield,selectivity,and benzene conversion.Shapley Additive Explanations(SHAP)analysis and feature importance analysis further revealed the contribution of each variable to the reaction outcomes.Additionally,we randomly generated one million variable combinations using the Dirichlet distribution to attempt to predict high-yield catalyst formulations.This paper provides new insights into the application of machine learning in heterogeneous catalysis and offers some reference for further research.
文摘The electrocatalytic reduction of nitric oxide for ammonia synthesis(NORR)is a key green energy conversion technology.Its efficiency relies on high-performance electrocatalysts to enhance both ammonia yield(Y_(NH3))and Faradaic efficiency(F_(NH3)).However,conventional experimental methods for screening high-activity NORR catalysts often entail high resource consumption and time costs.Machine learning combined with SHAP feature analysis was employed to establish a stacked ensemble model that integrates multiple algorithms,to allow for a systematic investigation of the key descriptors governing NORR performance based on an experimental dataset.Evaluation of eight model algorithms revealed that the Stacked-SVR model achieved an R^(2)of 0.9223 and an RMSE of 0.0608 for predicting on the test set,whereas the Stacked-RF model achieved an R^(2)of 0.9042 and an RMSE of 0.0900 for predicting.The stacked ensemble model integrates the strengths of individual algorithms and demonstrates strong NORR prediction performance while avoiding overfitting.SHAP feature analysis results revealed that the Cu content in the catalyst composition has the most significant impact on catalytic performance.Moreover,the combination of the wet chemical reduction synthesis,a carbon fiber(CF)conductive substrate,and HCl electrolyte is more favorable for enhancing catalytic activity.Additionally,moderately lowering the working potential,controlling the electrolyte volume at low to medium levels,reducing catalyst loading,and increasing electrolyte concentration were found to synergistically enhance both and.
文摘Nonlinear classification models are widely used in various fields due to their excellent performance in handling complex problems.This paper investigates the learning performance of nonlinear classification models based on Markov sampling,which builds upon the traditional framework using i.i.d.samples.Subsequently,we introduce a ueMC-NL algorithm,tailored specifically for nonlinear classification models,facilitating the production of ueMC samples from a finite dataset.Numerical investigations on the random forest and the MLP model reveal that nonlinear classification models utilizing ueMC samples yield lower misclassification rates compared to i.i.d.samples.
文摘The application and promotion of waste glass powder concrete(WGPC)cansignificantly alleviate the pressure of concrete material scarcity and environmental pollution.Compressive strength(CS)is a critical parameter for evaluating the efficacy of WGPC.Unlike conventional testing methods,machine learning techniques offer precise and reliable predictions of concrete’s compressive strength,especially in its long-term mechanical properties.In this work,four models,namely Multiple Linear Regression(MLR),Back Propagation Neural Network(BPNN),Support Vector Regression(SVR),and Random Forest Regression(RFR)were employed.Furthermore,particle swarm optimization(PSO)algorithm and cross-validation techniques were applied to fine-tune the model parameters,striving for peak prediction performance.The results indicated that optimized models generally exhibit enhanced predictive accuracy compared to their basic counterparts.Notably,the PSO-RFR model excels among all evaluated models,showcasing superior performance on the testing dataset.It achieves a coefficient of determination(R^(2))of 0.9231,a mean absolute error(MAE)of 2.1073,and a root mean square error(RMSE)of 3.6903.When compared to experimental results,the PSO-RFR and PSO-BPNN models demonstrate exceptional predictive accuracy.Notably,the PSO-BPNN model exhibits the closest R^(2)values between its training and test sets.This close alignment of R^(2)values between the training and testing sets reflects the PSO-BPNN model’s superior generalization ability for unseen data.The findings present an efficient method for predicting concrete’s compressive strength,contributing to the sustainable development of concrete materials,and providing theoretical support for their research and application.
基金Project(42077244)supported by the National Natural Science Foundation of ChinaProject(2020-05)supported by the Open Research Fund of Guangdong Provincial Key Laboratory of Deep Earth Sciences and Geothermal Energy Exploitation and Utilization,China。
文摘Accurate prediction of rockburst intensity levels is crucial for ensuring the safety of deep hard rock engineering construction.This paper introduced an expert system for rockburst intensity level prediction that employs machine learning algorithms as the basis for its inference rules.The system comprises four modules:a database,a repository,an inference engine,and an interpreter.A database containing 1114 rockburst cases was used to construct 357 datasets that serve as the repository for the expert system.Additionally,19 types of machine learning algorithms were used to establish 6783 micro-models to construct cognitive rules within the inference engine.By integrating probability theory and marginal analysis,a fuzzy scoring method based on the SoftMax function was developed and applied to the interpreter for rockburst intensity level prediction,effectively restoring the continuity of rockburst characteristics.The research results indicate that ensemble algorithms based on decision trees are more effective in capturing the characteristics of rockburst.Key factors for accurate prediction of rockburst intensity include uniaxial compressive strength,elastic energy index,the maximum principal stress,tangential stress,and their composite indicators.The accuracy of the proposed rockburst intensity level prediction expert system was verified using 20 engineering rockburst cases,with predictions aligning closely with the actual rockburst intensity levels.
基金supported by the National Natural Science Foundation of China(42476084,62203456,42276199)the Stable Support Project of National Key Laboratory(WDZC 20245250302)the National Key R&D Program of China(2024YFC2813502,2024YFC2813302)。
文摘Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-only approaches.To address this issue,this paper proposes a framework named“a lightweight pure visual BEV perception method based on dual distillation of spatial-temporal knowledge”.This framework innovatively designs a lightweight vision-only student model based on Res Net,which leverages a dual distillation mechanism to learn from a powerful teacher model that integrates temporal information from both image and light detection and ranging(LiDAR)modalities.Specifically,we distill efficient multi-modal feature extraction and spatial fusion capabilities from the BEVFusion model,and distill advanced temporal information fusion and spatiotemporal attention mechanisms from the BEVFormer model.This dual distillation strategy enables the student model to achieve perception performance close to that of multi-modal models without relying on Li DAR.Experimental results on the nu Scenes dataset demonstrate that the proposed model significantly outperforms classical vision-only algorithms,achieves comparable performance to current state-of-the-art vision-only methods on the nu Scenes detection leaderboard in terms of both mean average precision(mAP)and the nu Scenes detection score(NDS)metrics,and exhibits notable advantages in inference computational efficiency.Although the proposed dual-teacher paradigm incurs higher offline training costs compared to single-model approaches,it yields a streamlined and highly efficient student model suitable for resource-constrained real-time deployment.This provides an effective pathway toward low-cost,high-performance autonomous driving perception systems.
基金supported by the National Key Research and Development Program of China(No.2022YFB4300902).
文摘As joint operations have become a key trend in modern military development,unmanned aerial vehicles(UAVs)play an increasingly important role in enhancing the intelligence and responsiveness of combat systems.However,the heterogeneity of aircraft,partial observability,and dynamic uncertainty in operational airspace pose significant challenges to autonomous collision avoidance using traditional methods.To address these issues,this paper proposes an adaptive collision avoidance approach for UAVs based on deep reinforcement learning.First,a unified uncertainty model incorporating dynamic wind fields is constructed to capture the complexity of joint operational environments.Then,to effectively handle the heterogeneity between manned and unmanned aircraft and the limitations of dynamic observations,a sector-based partial observation mechanism is designed.A Dynamic Threat Prioritization Assessment algorithm is also proposed to evaluate potential collision threats from multiple dimensions,including time to closest approach,minimum separation distance,and aircraft type.Furthermore,a Hierarchical Prioritized Experience Replay(HPER)mechanism is introduced,which classifies experience samples into high,medium,and low priority levels to preferentially sample critical experiences,thereby improving learning efficiency and accelerating policy convergence.Simulation results show that the proposed HPER-D3QN algorithm outperforms existing methods in terms of learning speed,environmental adaptability,and robustness,significantly enhancing collision avoidance performance and convergence rate.Finally,transfer experiments on a high-fidelity battlefield airspace simulation platform validate the proposed method's deployment potential and practical applicability in complex,real-world joint operational scenarios.
文摘Investigating the detonation reaction zone structures of high explosives is significant for understanding detonation reaction mechanism.This study employed an integrated approach combining machine learning prediction,theoretical calculation,and experimental characterization to determine the detonation reaction zone width of CL-20-based aluminized explosive.In this study,the detonation reaction zone refers to the reaction zone between the von Neumann(VN)peak and sonic point,which usually means the so-called detonation driving zone(DDZ).For the machine learning prediction,an ensemble model integrating Random Forest and Support Vector Regression was developed to predict the reaction zone width using a dataset of 19 publicly available samples.For the theoretical calculation,the Wood-Kirkwood(W-K)detonation theory model was utilized to implement numerical calculation of the reaction zone structures,incorporating chemical reaction kinetics to describe the detonation reaction progress.In experimental characterization,the Photon Doppler Velocimetry(PDV)was applied with LiF as the optical window to measure the particle velocity profile of detonation products and derive the reaction zone width.The results indicate that the reaction zone width values are 0.25 mm,0.28 mm,and 0.26 mm obtained from machine learning prediction,theoretical calculation,and experimental characterization,respectively.The corresponding velocities at the Chapman-Jouguet(CJ)point are 1,938 m/s,2,047 m/s,and 1,982 m/s,respectively.The maximum relative deviation in reaction zone width among three methods is approximately 7.7%,while that for CJ particle velocity is approximately 3.3%.These results from all three methods agree well within engineering error.This validates the effectiveness of integrating machine learning prediction,theoretical calculation and advanced experimental techniques for studying the detonation reaction zone structures of high explosives.This research provides insights into the detonation reaction mechanism and reaction zone characteristics of CL-20-based aluminized explosive.
基金supported by the National Key R&D Program of China (2018AAA0101400)the National Natural Science Foundation of China (62173251+3 种基金61921004U1713209)the Natural Science Foundation of Jiangsu Province of China (BK20202006)the Guangdong Provincial Key Laboratory of Intelligent Decision and Cooperative Control。
文摘In this paper, the reinforcement learning method for cooperative multi-agent systems(MAS) with incremental number of agents is studied. The existing multi-agent reinforcement learning approaches deal with the MAS with a specific number of agents, and can learn well-performed policies. However, if there is an increasing number of agents, the previously learned in may not perform well in the current scenario. The new agents need to learn from scratch to find optimal policies with others,which may slow down the learning speed of the whole team. To solve that problem, in this paper, we propose a new algorithm to take full advantage of the historical knowledge which was learned before, and transfer it from the previous agents to the new agents. Since the previous agents have been trained well in the source environment, they are treated as teacher agents in the target environment. Correspondingly, the new agents are called student agents. To enable the student agents to learn from the teacher agents, we first modify the input nodes of the networks for teacher agents to adapt to the current environment. Then, the teacher agents take the observations of the student agents as input, and output the advised actions and values as supervising information. Finally, the student agents combine the reward from the environment and the supervising information from the teacher agents, and learn the optimal policies with modified loss functions. By taking full advantage of the knowledge of teacher agents, the search space for the student agents will be reduced significantly, which can accelerate the learning speed of the holistic system. The proposed algorithm is verified in some multi-agent simulation environments, and its efficiency has been demonstrated by the experiment results.
基金supported by the National Natural Science Foundation of China(72101263).
文摘Tracking and analyzing data from research projects is critical for understanding research trends and supporting the development of science and technology strategies.However,the data from these projects is often complex and inadequate,making it challenging for researchers to conduct in-depth data mining to improve policies or management.To address this problem,this paper adopts a top-down approach to construct a knowledge graph(KG)for research projects.Firstly,we construct an integrated ontology by referring to the metamodel of various architectures,which is called the meta-model integration conceptual reference model.Subsequently,we use the dependency parsing method to extract knowledge from unstructured textual data and use the entity alignment method based on weakly supervised learning to classify the extracted entities,completing the construction of the KG for the research projects.In addition,a knowledge inference model based on representation learning is employed to achieve knowledge completion and improve the KG.Finally,experiments are conducted on the KG for research projects and the results demonstrate the effectiveness of the proposed method in enriching incomplete data within the KG.
基金the National Natural Science Foundation of China (Grants No. 12072090 and No.12302056) to provide fund for conducting experiments。
文摘Recently, high-precision trajectory prediction of ballistic missiles in the boost phase has become a research hotspot. This paper proposes a trajectory prediction algorithm driven by data and knowledge(DKTP) to solve this problem. Firstly, the complex dynamics characteristics of ballistic missile in the boost phase are analyzed in detail. Secondly, combining the missile dynamics model with the target gravity turning model, a knowledge-driven target three-dimensional turning(T3) model is derived. Then, the BP neural network is used to train the boost phase trajectory database in typical scenarios to obtain a datadriven state parameter mapping(SPM) model. On this basis, an online trajectory prediction framework driven by data and knowledge is established. Based on the SPM model, the three-dimensional turning coefficients of the target are predicted by using the current state of the target, and the state of the target at the next moment is obtained by combining the T3 model. Finally, simulation verification is carried out under various conditions. The simulation results show that the DKTP algorithm combines the advantages of data-driven and knowledge-driven, improves the interpretability of the algorithm, reduces the uncertainty, which can achieve high-precision trajectory prediction of ballistic missile in the boost phase.
基金supported by the National Natural Science Foundation of China(12072090).
文摘This work proposes the application of an iterative learning model predictive control(ILMPC)approach based on an adaptive fault observer(FOBILMPC)for fault-tolerant control and trajectory tracking in air-breathing hypersonic vehicles.In order to increase the control amount,this online control legislation makes use of model predictive control(MPC)that is based on the concept of iterative learning control(ILC).By using offline data to decrease the linearized model’s faults,the strategy may effectively increase the robustness of the control system and guarantee that disturbances can be suppressed.An adaptive fault observer is created based on the suggested ILMPC approach in order to enhance overall fault tolerance by estimating and compensating for actuator disturbance and fault degree.During the derivation process,a linearized model of longitudinal dynamics is established.The suggested ILMPC approach is likely to be used in the design of hypersonic vehicle control systems since numerical simulations have demonstrated that it can decrease tracking error and speed up convergence when compared to the offline controller.
