In recent years,autonomous driving technology has made good progress,but the noncooperative intelligence of vehicle for autonomous driving still has many technical bottlenecks when facing urban road autonomous driving...In recent years,autonomous driving technology has made good progress,but the noncooperative intelligence of vehicle for autonomous driving still has many technical bottlenecks when facing urban road autonomous driving challenges.V2I(Vehicle-to-Infrastructure)communication is a potential solution to enable cooperative intelligence of vehicles and roads.In this paper,the RGB-PVRCNN,an environment perception framework,is proposed to improve the environmental awareness of autonomous vehicles at intersections by leveraging V2I communication technology.This framework integrates vision feature based on PVRCNN.The normal distributions transform(NDT)point cloud registration algorithm is deployed both on onboard and roadside to obtain the position of the autonomous vehicles and to build the local map objects detected by roadside multi-sensor system are sent back to autonomous vehicles to enhance the perception ability of autonomous vehicles for benefiting path planning and traffic efficiency at the intersection.The field-testing results show that our method can effectively extend the environmental perception ability and range of autonomous vehicles at the intersection and outperform the PointPillar algorithm and the VoxelRCNN algorithm in detection accuracy.展开更多
In vehicle ad-hoc networks(VANETs),the proliferation of wireless communication will give rise to the heterogeneous access environment where network selection becomes significant.Motivated by the self-adaptive paradigm...In vehicle ad-hoc networks(VANETs),the proliferation of wireless communication will give rise to the heterogeneous access environment where network selection becomes significant.Motivated by the self-adaptive paradigm of cellular attractors,this paper regards an individual communication as a cell,so that we can apply the revised attractor selection model to induce each connected vehicle.Aiming at improving the Quality of Service(QoS),we presented the bio-inspired handover decision-making mechanism.In addition,we employ the Technique for Order Preference by Similarity to an Ideal Solution(TOPSIS)for any vehicle to choose an access network.This paper proposes a novel framework where the bio-inspired mechanism is combined with TOPSIS.In a dynamic and random mobility environment,our method achieves the coordination of performance of heterogeneous networks by guaranteeing the efficient utilization and fair distribution of network resources in a global sense.The experimental results confirm that the proposed method performs better when compared with conventional schemes.展开更多
Dynamic channel assignment(DCA)is significant for extending vehicular ad hoc network(VANET)capacity and mitigating congestion.However,the un-known global state information and the lack of centralized control make chan...Dynamic channel assignment(DCA)is significant for extending vehicular ad hoc network(VANET)capacity and mitigating congestion.However,the un-known global state information and the lack of centralized control make channel assignment performances a challenging task in a distributed vehicular direct communication scenario.In our preliminary field test for communication under V2X scenario,we find that the existing DCA technology cannot fully meet the communication performance requirements of VANET.In order to improve the communication performance,we firstly demonstrate the feasibility and potential of reinforcement learning(RL)method in joint channel selection decision and access fallback adaptation design in this paper.Besides,a dual reinforcement learning(DRL)-based cooperative DCA(DRL-CDCA)mechanism is proposed.Specifically,DRL-CDCA jointly optimizes the decision-making behaviors of both the channel selection and back-off adaptation based on a multi-agent dual reinforcement learning framework.Besides,nodes locally share and incorporate their individual rewards after each communication to achieve regional consistency optimization.Simulation results show that the proposed DRL-CDCA can better reduce the one-hop packet delay,improve the packet delivery ratio on average when compared with two other existing mechanisms.展开更多
Rule-based autonomous driving systems may suffer from increased complexity with large-scale intercoupled rules,so many researchers are exploring learning-based approaches.Reinforcement learning(RL)has been applied in ...Rule-based autonomous driving systems may suffer from increased complexity with large-scale intercoupled rules,so many researchers are exploring learning-based approaches.Reinforcement learning(RL)has been applied in designing autonomous driving systems because of its outstanding performance on a wide variety of sequential control problems.However,poor initial performance is a major challenge to the practical implementation of an RL-based autonomous driving system.RL training requires extensive training data before the model achieves reasonable performance,making an RL-based model inapplicable in a real-world setting,particularly when data are expensive.We propose an asynchronous supervised learning(ASL)method for the RL-based end-to-end autonomous driving model to address the problem of poor initial performance before training this RL-based model in real-world settings.Specifically,prior knowledge is introduced in the ASL pre-training stage by asynchronously executing multiple supervised learning processes in parallel,on multiple driving demonstration data sets.After pre-training,the model is deployed on a real vehicle to be further trained by RL to adapt to the real environment and continuously break the performance limit.The presented pre-training method is evaluated on the race car simulator,TORCS(The Open Racing Car Simulator),to verify that it can be sufficiently reliable in improving the initial performance and convergence speed of an end-to-end autonomous driving model in the RL training stage.In addition,a real-vehicle verification system is built to verify the feasibility of the proposed pre-training method in a real-vehicle deployment.Simulations results show that using some demonstrations during a supervised pre-training stage allows significant improvements in initial performance and convergence speed in the RL training stage.展开更多
The formation control of unmanned aerial vehicle(UAV)swarms is of significant importance in various fields such as transportation,emergency management,and environmental monitoring.However,the complex dynamics,nonlinea...The formation control of unmanned aerial vehicle(UAV)swarms is of significant importance in various fields such as transportation,emergency management,and environmental monitoring.However,the complex dynamics,nonlinearity,uncertainty,and interaction among agents make it a challenging problem.In this paper,we propose a distributed robust control strategy that uses only local information of UAVs to improve the stability and robustness of the formation system in uncertain environments.We establish a nominal control strategy based on position relations and a semi-definite programming model to obtain control gains.Additionally,we propose a robust control strategy under the rotation setΩto address the noise and disturbance in the system,ensuring that even when the rotation angles of the UAVs change,they still form a stable formation.Finally,we extend the proposed strategy to a quadrotor UAV system with high-order kinematic models and conduct simulation experiments to validate its effectiveness in resisting uncertain disturbances and achieving formation control.展开更多
Self-driving vehicles require a number of tests to prevent fatal accidents and ensure their appropriate operation in the physical world.However,conducting vehicle tests on the road is difficult because such tests are ...Self-driving vehicles require a number of tests to prevent fatal accidents and ensure their appropriate operation in the physical world.However,conducting vehicle tests on the road is difficult because such tests are expensive and labor intensive.In this study,we used an autonomous-driving simulator,and investigated the three-dimensional environmental perception problem of the simulated system.Using the open-source CARLA simulator,we generated a CarlaSim from unreal traffic scenarios,comprising 15000 camera-LiDAR(Light Detection and Ranging)samples with annotations and calibration files.Then,we developed Multi-Sensor Fusion Perception(MSFP)model for consuming two-modal data and detecting objects in the scenes.Furthermore,we conducted experiments on the KITTI and CarlaSim datasets;the results demonstrated the effectiveness of our proposed methods in terms of perception accuracy,inference efficiency,and generalization performance.The results of this study will faciliate the future development of autonomous-driving simulated tests.展开更多
Purpose–This paper aims to introduce vehicular network platform,routing and broadcasting methods and vehicular positioning enhancement technology,which are three aspects of the applications of intelligent computing i...Purpose–This paper aims to introduce vehicular network platform,routing and broadcasting methods and vehicular positioning enhancement technology,which are three aspects of the applications of intelligent computing in vehicular networks.From this paper,the role of intelligent algorithm in thefield of transportation and the vehicular networks can be understood.Design/methodology/approach–In this paper,the authors introduce three different methods in three layers of vehicle networking,which are data cleaning based on machine learning,routing algorithm based on epidemic model and cooperative localization algorithm based on the connect vehicles.Findings–In Section 2,a novel classification-based framework is proposed to efficiently assess the data quality and screen out the abnormal vehicles in database.In Section 3,the authors canfind when traffic conditions varied from freeflow to congestion,the number of message copies increased dramatically and the reachability also improved.The error of vehicle positioning is reduced by 35.39%based on the CV-IMM-EKF in Section 4.Finally,it can be concluded that the intelligent computing in the vehicle network system is effective,and it will improve the development of the car networking system.Originality/value–This paper reviews the research of intelligent algorithms in three related areas of vehicle networking.In thefield of vehicle networking,these research results are conducive to promoting data processing and algorithm optimization,and it may lay the foundation for the new methods.展开更多
基金This research was supported by the National Key Research and Development Program of China under Grant No.2017YFB0102502the Beijing Municipal Natural Science Foundation No.L191001+2 种基金the National Natural Science Foundation of China under Grant No.61672082 and 61822101the Newton Advanced Fellowship under Grant No.62061130221the Young Elite Scientists Sponsorship Program by Hunan Provincial Department of Education under Grant No.18B142.
文摘In recent years,autonomous driving technology has made good progress,but the noncooperative intelligence of vehicle for autonomous driving still has many technical bottlenecks when facing urban road autonomous driving challenges.V2I(Vehicle-to-Infrastructure)communication is a potential solution to enable cooperative intelligence of vehicles and roads.In this paper,the RGB-PVRCNN,an environment perception framework,is proposed to improve the environmental awareness of autonomous vehicles at intersections by leveraging V2I communication technology.This framework integrates vision feature based on PVRCNN.The normal distributions transform(NDT)point cloud registration algorithm is deployed both on onboard and roadside to obtain the position of the autonomous vehicles and to build the local map objects detected by roadside multi-sensor system are sent back to autonomous vehicles to enhance the perception ability of autonomous vehicles for benefiting path planning and traffic efficiency at the intersection.The field-testing results show that our method can effectively extend the environmental perception ability and range of autonomous vehicles at the intersection and outperform the PointPillar algorithm and the VoxelRCNN algorithm in detection accuracy.
基金This research was supported in part by the National Natural Science Foundation of China under Grant Nos.61672082 and 61822101Beijing Municipal Natural Science Foundation Nos.4181002Beihang University Innovation&Practice Fund for Graduate(YCSJ-02-2018-05).
文摘In vehicle ad-hoc networks(VANETs),the proliferation of wireless communication will give rise to the heterogeneous access environment where network selection becomes significant.Motivated by the self-adaptive paradigm of cellular attractors,this paper regards an individual communication as a cell,so that we can apply the revised attractor selection model to induce each connected vehicle.Aiming at improving the Quality of Service(QoS),we presented the bio-inspired handover decision-making mechanism.In addition,we employ the Technique for Order Preference by Similarity to an Ideal Solution(TOPSIS)for any vehicle to choose an access network.This paper proposes a novel framework where the bio-inspired mechanism is combined with TOPSIS.In a dynamic and random mobility environment,our method achieves the coordination of performance of heterogeneous networks by guaranteeing the efficient utilization and fair distribution of network resources in a global sense.The experimental results confirm that the proposed method performs better when compared with conventional schemes.
基金Beijing Municipal Natural Science Foundation Nos.L191001 and 4181002the National Natural Science Foundation of China under Grant Nos.61672082 and 61822101the Newton Advanced Fellow-ship under Grant No.62061130221.
文摘Dynamic channel assignment(DCA)is significant for extending vehicular ad hoc network(VANET)capacity and mitigating congestion.However,the un-known global state information and the lack of centralized control make channel assignment performances a challenging task in a distributed vehicular direct communication scenario.In our preliminary field test for communication under V2X scenario,we find that the existing DCA technology cannot fully meet the communication performance requirements of VANET.In order to improve the communication performance,we firstly demonstrate the feasibility and potential of reinforcement learning(RL)method in joint channel selection decision and access fallback adaptation design in this paper.Besides,a dual reinforcement learning(DRL)-based cooperative DCA(DRL-CDCA)mechanism is proposed.Specifically,DRL-CDCA jointly optimizes the decision-making behaviors of both the channel selection and back-off adaptation based on a multi-agent dual reinforcement learning framework.Besides,nodes locally share and incorporate their individual rewards after each communication to achieve regional consistency optimization.Simulation results show that the proposed DRL-CDCA can better reduce the one-hop packet delay,improve the packet delivery ratio on average when compared with two other existing mechanisms.
基金Project supported by the National Natural Science Foundation of China(Nos.61672082 and 61822101)the Beijing Municipal Natural Science Foundation,China(No.4181002)the Beihang University Innovation and Practice Fund for Graduate,China(No.YCSJ-02-2018-05)。
文摘Rule-based autonomous driving systems may suffer from increased complexity with large-scale intercoupled rules,so many researchers are exploring learning-based approaches.Reinforcement learning(RL)has been applied in designing autonomous driving systems because of its outstanding performance on a wide variety of sequential control problems.However,poor initial performance is a major challenge to the practical implementation of an RL-based autonomous driving system.RL training requires extensive training data before the model achieves reasonable performance,making an RL-based model inapplicable in a real-world setting,particularly when data are expensive.We propose an asynchronous supervised learning(ASL)method for the RL-based end-to-end autonomous driving model to address the problem of poor initial performance before training this RL-based model in real-world settings.Specifically,prior knowledge is introduced in the ASL pre-training stage by asynchronously executing multiple supervised learning processes in parallel,on multiple driving demonstration data sets.After pre-training,the model is deployed on a real vehicle to be further trained by RL to adapt to the real environment and continuously break the performance limit.The presented pre-training method is evaluated on the race car simulator,TORCS(The Open Racing Car Simulator),to verify that it can be sufficiently reliable in improving the initial performance and convergence speed of an end-to-end autonomous driving model in the RL training stage.In addition,a real-vehicle verification system is built to verify the feasibility of the proposed pre-training method in a real-vehicle deployment.Simulations results show that using some demonstrations during a supervised pre-training stage allows significant improvements in initial performance and convergence speed in the RL training stage.
基金supported by the National Natural Science Foundation of China(Nos.52202391,U20A20155,and 52302397)the China Postdoctoral Science Foundation(No.2023M730173).
文摘The formation control of unmanned aerial vehicle(UAV)swarms is of significant importance in various fields such as transportation,emergency management,and environmental monitoring.However,the complex dynamics,nonlinearity,uncertainty,and interaction among agents make it a challenging problem.In this paper,we propose a distributed robust control strategy that uses only local information of UAVs to improve the stability and robustness of the formation system in uncertain environments.We establish a nominal control strategy based on position relations and a semi-definite programming model to obtain control gains.Additionally,we propose a robust control strategy under the rotation setΩto address the noise and disturbance in the system,ensuring that even when the rotation angles of the UAVs change,they still form a stable formation.Finally,we extend the proposed strategy to a quadrotor UAV system with high-order kinematic models and conduct simulation experiments to validate its effectiveness in resisting uncertain disturbances and achieving formation control.
基金supported by the National Natural Science Foundation of China(Nos.61822101 and 62061130221)the Beijing Municipal Key Research and Development Program(No.Z181100004618006)+1 种基金the Beijing Municipal Natural Science Foundation(No.L191001),the Zhuoyue Program of Beihang University(Postdoctoral Fellowship)(No.262716)the China Postdoctoral Science Foundation(No.2020M680299)。
文摘Self-driving vehicles require a number of tests to prevent fatal accidents and ensure their appropriate operation in the physical world.However,conducting vehicle tests on the road is difficult because such tests are expensive and labor intensive.In this study,we used an autonomous-driving simulator,and investigated the three-dimensional environmental perception problem of the simulated system.Using the open-source CARLA simulator,we generated a CarlaSim from unreal traffic scenarios,comprising 15000 camera-LiDAR(Light Detection and Ranging)samples with annotations and calibration files.Then,we developed Multi-Sensor Fusion Perception(MSFP)model for consuming two-modal data and detecting objects in the scenes.Furthermore,we conducted experiments on the KITTI and CarlaSim datasets;the results demonstrated the effectiveness of our proposed methods in terms of perception accuracy,inference efficiency,and generalization performance.The results of this study will faciliate the future development of autonomous-driving simulated tests.
基金supported in part by the National Natural Science Foundation of China under Grant Nos.61672082,U1564212.
文摘Purpose–This paper aims to introduce vehicular network platform,routing and broadcasting methods and vehicular positioning enhancement technology,which are three aspects of the applications of intelligent computing in vehicular networks.From this paper,the role of intelligent algorithm in thefield of transportation and the vehicular networks can be understood.Design/methodology/approach–In this paper,the authors introduce three different methods in three layers of vehicle networking,which are data cleaning based on machine learning,routing algorithm based on epidemic model and cooperative localization algorithm based on the connect vehicles.Findings–In Section 2,a novel classification-based framework is proposed to efficiently assess the data quality and screen out the abnormal vehicles in database.In Section 3,the authors canfind when traffic conditions varied from freeflow to congestion,the number of message copies increased dramatically and the reachability also improved.The error of vehicle positioning is reduced by 35.39%based on the CV-IMM-EKF in Section 4.Finally,it can be concluded that the intelligent computing in the vehicle network system is effective,and it will improve the development of the car networking system.Originality/value–This paper reviews the research of intelligent algorithms in three related areas of vehicle networking.In thefield of vehicle networking,these research results are conducive to promoting data processing and algorithm optimization,and it may lay the foundation for the new methods.
