Non-signalized intersection intelligent early warning system based on digital twinning

Abstract

The invention discloses a non-signalized intersection intelligent early warning system based on digital twinning, and the system comprises the following steps: a traffic perception data collection module which is used for collecting data in an intersection region; the digital behavior twin modeling module is used for constructing a digital behavior twin model; the conflict risk optimization module is used for introducing a role switching mechanism into a lion group optimization algorithm to generate a behavior strategy candidate solution set; the role division and switching management module is used for dividing a population role structure; the behavior adjustment mapping module is used for mapping the optimal strategy vector into a micro-behavior adjustment instruction set; the simulation verification and scoring module is used for executing an interactive simulation process; the early warning generation and classification module is used for constructing an early warning information object set; the edge communication scheduling module is used for executing target matching and path selection; and the terminal pushing and feedback module is used for pushing the issuing instruction set to the terminal and recording an issuing response state. According to the invention, intelligent early warning control of the non-signalized intersection is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent traffic safety technology, and in particular to an intelligent early warning system for unsignaled intersections based on digital twins. Background Art

[0002] In current research and practice in intelligent transportation systems, traffic management at unsignalized intersections has long been a key focus for improving traffic safety and efficiency. Compared to signalized intersections, unsignalized intersections lack the unified command and dispatch of traffic lights, forcing traffic participants to rely on their own judgment to make decisions about yielding or passing. This situation is particularly complex in mixed traffic environments, where multiple types of traffic entities, including motor vehicles, non-motor vehicles, and pedestrians, converge. Furthermore, dynamic changes in environmental factors such as weather, lighting, and road conditions create a high risk of conflict and potential traffic accidents at intersections. To this end, existing research has attempted to utilize sensing devices such as radar, cameras, and lidar to collect traffic data, combining them with rule engines or fixed threshold strategies to implement basic risk warning functions. Some solutions have also incorporated edge computing nodes to improve response time. However, most methods are still based on static models or fixed rules, making them difficult to handle highly dynamic and complex traffic behavior patterns.

[0003] Some existing approaches explore early warning mechanisms based on multi-agent simulation and traffic behavior modeling. However, these approaches often simplify the behavioral patterns of traffic participants into sets of regular actions and lack fine-grained simulation of individual behavioral evolution. Furthermore, existing behavioral modeling typically fails to dynamically couple environmental states with behavioral paths, making it difficult to accurately reproduce the evolution of traffic conflicts in real-world scenarios. Regarding optimization and control, previous studies have attempted to improve traffic strategies using traditional swarm intelligence optimization methods such as genetic algorithms and particle swarm algorithms. However, these algorithms have fixed individual roles and a single search space update mechanism. When faced with complex behavioral strategy spaces, they are prone to falling into local optimality and lack convergence capabilities, limiting the diversity and effectiveness of strategy generation.

[0004] In particular, in the feedback and warning generation process of behavioral simulation results, existing methods lack a closed-loop verification mechanism that evolves from the individual behavior level to the group interaction level, making it impossible to achieve efficient linkage between strategy generation, behavior adjustment, conflict verification, and risk information distribution. In addition, some solutions only utilize simulation status results at the visual display or statistical evaluation level, and have not formed a micro-behavior regulation system for real-time feedback control. At the same time, most current centralized architecture designs have bottlenecks such as high network latency, insufficient coverage, and poor communication stability during the information distribution and response process, making it difficult to meet the high-reliability warning needs in rapidly changing traffic scenarios.

[0005] Therefore, how to provide an intelligent warning system for unsignaled intersections based on digital twins is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0006] One purpose of the present invention is to propose an intelligent warning system for unsignaled intersections based on digital twins. The present invention fully integrates traffic status perception, multi-agent behavior modeling, lion group optimization algorithm and edge communication technology, and describes in detail the entire process of constructing a digital behavior twin model in an unsignaled control scenario, performing multi-objective conflict risk optimization, and realizing strategy-driven micro-behavior regulation and warning issuance. It has the advantages of refined behavior modeling, high conflict identification accuracy, strong strategy adaptability and high real-time warning response.

[0007] An intelligent warning system for unsignalized intersections based on digital twins according to an embodiment of the present invention includes:

[0008] Traffic perception data collection module, used to collect data in the intersection area and generate a traffic status data set;

[0009] The digital behavior twin modeling module is used to build a digital behavior twin model and output a set of behavior simulation states;

[0010] The conflict risk optimization module is used to perform multi-objective traffic conflict risk optimization operations based on the lion group optimization algorithm and generate a set of candidate behavioral strategy solutions through role division and role switching mechanisms;

[0011] The role division and switching management module is used to divide the population role structure based on individual fitness scores and behavioral strategy contribution rates;

[0012] A behavior adjustment mapping module is used to map the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set;

[0013] The simulation verification and scoring module is used to input the micro-behavior adjustment instruction set into the digital behavior twin model, execute the interactive simulation process, and output the conflict warning score result set;

[0014] The warning generation and classification module is used to construct a warning information object set based on the conflict warning score result set and the risk threshold set;

[0015] The edge communication scheduling module is used to perform target matching and path selection on the warning information object set and generate a set of instructions for issuance;

[0016] The terminal push and feedback module is used to push the issued instruction set to the target traffic participant terminal in real time through the edge communication link and record the issued response status.

[0017] Optionally, modules can be connected using the following methods:

[0018] S1. Collect traffic perception data of the intersection area and generate a traffic status data set;

[0019] S2. Constructing a digital behavior twin model based on the traffic state data set, wherein the digital behavior twin model includes multiple traffic subject behavior units and environment interaction units, and outputs a behavior simulation state set;

[0020] S3. Introducing the lion group optimization algorithm, performing multi-objective traffic conflict risk optimization operations on the behavioral simulation state set, generating a set of behavioral strategy candidate solutions, and introducing a fitness-driven role switching mechanism to dynamically adjust individual behavioral roles and strategy combinations;

[0021] S4, performing behavior adjustment processing based on the behavior strategy candidate solution set, mapping the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set;

[0022] S5. Input the micro-behavior adjustment instruction set into the digital behavior twin model, perform interactive simulation verification, update the traffic behavior path evolution state, and output the conflict warning score result set;

[0023] S6. Generate warning information based on the conflict warning score result set, and send the warning information to the target traffic participant terminal through the edge communication link.

[0024] Optionally, the traffic status data set includes vehicle position coordinates, vehicle speed, driving direction angle, acceleration, and pedestrian trajectory information, light intensity, road slipperiness coefficient and visibility index associated with the intersection area, which are used to characterize the joint characteristics of traffic behavior status and environmental perception status.

[0025] Optionally, the S2 specifically includes:

[0026] S21. Construct each set of state parameters in the traffic state data set into a multidimensional state vector, and construct a traffic state time series set in chronological order;

[0027] S22. performing normalization and time synchronization on the traffic state time series set to generate a standardized state series set;

[0028] S23. Construct a digital behavior twin model based on a standardized state sequence set, wherein the digital behavior twin model includes a traffic subject behavior unit, an environment interaction unit, a behavior propagation structure, and a state synchronization interface, wherein the traffic subject behavior unit is used to model the state transition behavior of vehicles and pedestrians respectively; the environment interaction unit is used to couple environmental parameters with behavior states; the behavior propagation structure is used to construct a dynamic interaction graph between traffic participants; and the state synchronization interface is used to align the actual traffic state with the state of the digital behavior twin model;

[0029] S24. Establish a behavior propagation mechanism in the digital behavior twin model. The propagation mechanism treats each traffic participant as a graph node, generates a propagation graph based on the adjacency relationship, and performs behavior state propagation calculation according to the state transition probability function:

[0030]

[0031] Among them, A t (i,j) represents the adjacency matrix element between node i and node j at time t, d ij is the spatial distance between the traffic participants corresponding to node i and node j, and δ is the spatial adjacency threshold for propagation judgment;

[0032] S25. Based on the behavior propagation mechanism, a semantic state vector is generated at each time step, and the semantic state vectors are combined into a behavior simulation state set output. The semantic state vector includes the intersection area occupancy status, the path expected area, the conflict probability score, the traffic priority code, the degree of behavior interaction influence, the visibility gap parameter, the yield state mark, the path smoothness index and the speed consistency score.

[0033] Optionally, the S3 specifically includes:

[0034] S31. Use the behavioral simulation state set as the initial input, construct an optimized population, and evaluate individual performance based on the behavioral fitness function:

[0035]

[0036] Among them, F i is the behavioral fitness score of the i-th individual, λ1 is the behavioral risk score weight, T is the total number of behavioral simulation time steps, t is the behavioral simulation time step, ω t is the timing attenuation factor, is the local behavior risk score of individual i at time t, λ2 is the conflict score weight, is the set of neighbor individuals that have path interactions with individual i, η i,j is the conflict importance weight between individual i and neighbor j, is the conflict score between individual i and neighbor j in the entire simulation path, λ3 is the behavior rationality penalty weight, κ1 is the path deviation penalty weight, is the maximum behavior path deviation distance of individual i, κ2 is the speed jump penalty weight, is the average speed jump degree of individual i, i is the optimized individual number, j is the number of the adjacent traffic participant, and the conflict score of the i-th individual in the t-th round of simulation is

[0037] S32. Based on the distribution of individual behavioral performance in the state space, the optimized population is divided into a set of territorial male lions, a set of female lions, and a set of wandering male lions;

[0038] S33, simulate the process of female lions hunting and male lions patrolling the territory, perform search updates in the individual neighborhood space and the global space respectively, and the position vector of the i-th individual in the t-th iteration is updated by superimposing the current optimal individual guidance term and the Gaussian perturbation term:

[0039]

[0040] in, is the behavioral strategy vector of the i-th individual in the t+1 round, is the behavioral strategy vector of the i-th individual in the t-th round, α is the optimal guidance weight coefficient, is the optimal behavior strategy vector of the population in the current round t, β is the local disturbance control coefficient, and randn() is a random vector obeying the standard normal distribution;

[0041] S34. Perform a crossover strategy between the lioness and the male lion, and generate a new individual through mating operation:

[0042] X child =γ·X male +(1-γ)·X female ;

[0043] Among them, X child is the offspring behavior strategy vector, γ is the strategy inheritance weight factor, X male is the lion’s individual behavior strategy vector, X female is the individual behavior strategy vector of the lioness;

[0044] S35. Execute the role switching mechanism based on the individual fitness change and the behavioral strategy contribution rate. When the conditions are met, update the role label set.

[0045] S36. Update the current lion group individual set and role structure information. When the number of iterations reaches the upper limit, the fitness improvement is lower than the threshold, or the behavioral strategy diversity converges, the optimization process is determined to be completed and the behavioral strategy candidate solution set is output.

[0046] Optionally, the S32 specifically includes:

[0047] S321, construct the optimized population individual set P = {X1, X2, ..., X i ,…,X N}, where X i represents the behavior strategy vector of the i-th behavior strategy individual;

[0048] S322, calculate the fitness score vector F = [F1, F2, ..., F N ], and calculate the average value of the fitness score vector and standard deviation σ F ;

[0049] S323, role division based on individual fitness scores and statistical indicators: Individuals are included in the territory of male lions set M t ,satisfy Individuals are classified into the lionesses set F t , the remaining individuals are classified into the wandering lion set D t ;

[0050] S324, for each individual X i Assign the role label role i ∈{male,female,nomad}, and initialize the behavior history vector set, strategy interaction counter and role stability flag.

[0051] Optionally, the S35 specifically includes:

[0052] S351, optimize the population individual set P = {X1, X2, ..., X i ,…,X N}, calculate the fitness change value in, is the individual X in the current round i The fitness score of Score the fitness of the previous round;

[0053] S352. Calculate the contribution rate of individual behavior strategies:

[0054]

[0055] Among them, ρ i is the behavioral strategy contribution rate of the i-th optimized population individual, T is the current cumulative number of behavioral simulation rounds, t is the iteration round number, Score the conflict of individual i in the tth round of simulation, is the basic conflict score when the optimization strategy is not used in the t-th round of simulation, N is the number of individuals in the optimization population, and j is the population traversal index;

[0056] S353. Setting the threshold value ρ of the strategy contribution rate thresh , the average fitness of the current round population F avg and the minimum fitness alert value F low ;

[0057] S354. Execute the role switching mechanism and update the individual role label when any of the following conditions is met: i >ρ thresh , then the current lioness will be promoted to the strategic commander, if ΔF i >0 and The current wandering lion will be promoted to a territory lion. If Then the current individual will be downgraded to a wandering lion;

[0058] S355. Update the individual role tag set, and synchronously update the role stability identifier and the strategy interaction counter.

[0059] Optionally, the S4 specifically includes:

[0060] S41. Based on the set of behavioral strategy candidate solutions, extract the speed adjustment factor, acceleration change factor, steering angle correction factor, and path priority factor contained in each strategy vector to construct a behavioral strategy parameter matrix;

[0061] S42. Constructing a family of micro-behavior regulation functions Corresponding to the speed, acceleration, direction and path priority adjustment dimensions respectively;

[0062] S43. For each strategy vector, input the corresponding parameter factor into the micro-behavior adjustment function family to calculate the micro-behavior adjustment instruction vector;

[0063] S44. Merge all micro-behavior adjustment instruction vectors to generate a micro-behavior adjustment instruction set, and pass the micro-behavior adjustment instruction set into the digital behavior twin model to enter the subsequent simulation verification process.

[0064] Optionally, the S5 specifically includes:

[0065] S51, inputting the micro-behavior adjustment instruction set into each traffic subject behavior unit in the digital behavior twin model to control its micro-behavior adjustment operation within the simulation time sequence;

[0066] S52. During the evolution of the digital behavioral twin model, record the minimum spatial distance between each agent, the predicted trajectory overlap area, the length of the horizontal and vertical time conflict window, and calculate the traffic conflict risk score matrix;

[0067] S53. Perform temporal aggregation on the risk score values ​​at all time steps to construct a global conflict score function:

[0068] S54, according to the conflict scoring function value S risk Output the conflict warning score result set and feed it back to the warning generation and communication distribution process.

[0069] Optionally, the S6 specifically includes:

[0070] S61. Receive a conflict warning score result set output by the digital behavior twin model;

[0071] S62. Set a set of conflict risk thresholds, and perform warning level classification based on the hierarchical judgment rules to construct a warning information object set. Each of w i Includes target subject identification, warning level, conflict type description and response time recommendation;

[0072] S63: Based on the edge node deployment structure within the intersection area, call the edge communication scheduling table, perform target matching and communication path selection on the warning information object set, and generate a set of instructions for issuance. Where each i k Including target terminal identification, issuing timestamp and instruction content summary;

[0073] S64. Push the issued instruction set to the corresponding target traffic participant terminal in real time through the edge communication link, and record the issued response status for subsequent feedback verification process.

[0074] The beneficial effects of the present invention are:

[0075] By constructing a digital behavioral twin model and introducing a lion group optimization algorithm, the present invention achieves a detailed simulation of the behavior of multiple traffic entities and efficient conflict risk identification in unsignaled intersection scenarios, thereby improving the early warning system's perception and response capabilities to complex dynamic traffic conditions. Compared with traditional early warning mechanisms based on static rules or fixed models, the present invention starts from traffic status data and constructs a digital behavioral twin structure with traffic participant behavior units and environmental interaction units, which can dynamically reflect the spatiotemporal interaction relationship between traffic entities. Through the behavior propagation structure and adjacency graph modeling method, the system can generate a set of semantic state vectors covering traffic paths, conflict probabilities, visibility, and yielding relationships to form a comprehensive behavior simulation state.

[0076] During the optimization process, a lion group optimization algorithm with a role switching mechanism was employed, introducing a fitness-driven role partitioning and dynamic evolution approach. This not only enhanced the directionality and diversity of population search but also significantly improved the coverage and overall quality of behavioral strategies. Specifically, by constructing an individual fitness function and combining it with a behavioral strategy contribution rate indicator, the active evolution of roles between individuals was achieved, enabling the system to automatically identify high-contributing individuals and strengthen and disseminate their behavioral strategies, thereby obtaining a more optimal micro-behavior adjustment solution. The behavioral strategy generates adjustment instructions through a function family mapping and is then input into the simulation model, directly driving the internal evolution of the twin traffic system, completing conflict verification and risk score calculation.

[0077] Ultimately, the system constructs warning information based on the simulation scoring results, distributes it via edge communication mechanisms, and integrates it with real-time response status to form a closed-loop feedback loop. The entire process embodies high-precision data-driven modeling capabilities, optimized and guided strategy evolution, and rapid response to scenario implementation, effectively improving the efficiency of warning decision-making and proactive safety at unsignalized intersections. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 This is a flowchart of a method for an intelligent warning system for unsignalized intersections based on digital twins proposed by the present invention;

[0080] Figure 2 This is a structural diagram of a digital behavior twin model of an intelligent warning system for unsignalized intersections based on digital twins proposed in the present invention;

[0081] Figure 3 This is a flowchart of the individual role division and role switching mechanism in the lion group optimization algorithm of the digital twin-based unsignaled intersection intelligent warning system proposed by the present invention;

[0082] Figure 4 This is a system functional module structure diagram of the digital twin-based intelligent warning system for unsignalized intersections proposed in this invention. DETAILED DESCRIPTION

[0083] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0084] refer to Figure 1-4 , an intelligent warning system for unsignalized intersections based on digital twins, including:

[0085] Traffic perception data collection module, used to collect data in the intersection area and generate a traffic status data set;

[0086] The digital behavior twin modeling module is used to build a digital behavior twin model and output a set of behavior simulation states;

[0087] The conflict risk optimization module is used to perform multi-objective traffic conflict risk optimization operations based on the lion group optimization algorithm and generate a set of candidate behavioral strategy solutions through role division and role switching mechanisms;

[0088] The role division and switching management module is used to divide the population role structure based on individual fitness scores and behavioral strategy contribution rates;

[0089] A behavior adjustment mapping module is used to map the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set;

[0090] The simulation verification and scoring module is used to input the micro-behavior adjustment instruction set into the digital behavior twin model, execute the interactive simulation process, and output the conflict warning score result set;

[0091] The warning generation and classification module is used to construct a warning information object set based on the conflict warning score result set and the risk threshold set;

[0092] The edge communication scheduling module is used to perform target matching and path selection on the warning information object set and generate a set of instructions for issuance;

[0093] The terminal push and feedback module is used to push the issued instruction set to the target traffic participant terminal in real time through the edge communication link and record the issued response status.

[0094] The present invention constructs an intelligent warning system for unsignaled intersections based on digital twins, and realizes a closed-loop processing flow from traffic status perception, behavioral strategy optimization, conflict simulation verification to the issuance of warning information through modular design. The system has high behavioral modeling accuracy and risk identification capabilities. It can dynamically evaluate behavioral conflict risks in a mixed environment with multiple traffic participants, and automatically generate and optimize micro-behavior adjustment strategies. The optimization search efficiency and diversity are improved through a role switching mechanism driven by role division and strategy contribution rate, ensuring the globality and stability of the strategy solution. In the real traffic scenario simulation test, the system showed the advantages of high warning accuracy, short response time and excellent edge communication efficiency, providing a feasible intelligent solution for improving the traffic safety of unsignaled intersections.

[0095] In this embodiment, the modules are connected through the following methods:

[0096] S1. Collect traffic perception data of the intersection area and generate a traffic status data set;

[0097] S2. Constructing a digital behavior twin model based on the traffic state data set, wherein the digital behavior twin model includes multiple traffic subject behavior units and environment interaction units, and outputs a behavior simulation state set;

[0098] S3. Introducing the lion group optimization algorithm, performing multi-objective traffic conflict risk optimization operations on the behavioral simulation state set, generating a set of behavioral strategy candidate solutions, and introducing a fitness-driven role switching mechanism to dynamically adjust individual behavioral roles and strategy combinations;

[0099] S4, performing behavior adjustment processing based on the behavior strategy candidate solution set, mapping the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set;

[0100] S5. Input the micro-behavior adjustment instruction set into the digital behavior twin model, perform interactive simulation verification, update the traffic behavior path evolution state, and output the conflict warning score result set;

[0101] S6. Generate warning information based on the conflict warning score result set, and send the warning information to the target traffic participant terminal through the edge communication link.

[0102] The present invention constructs an intelligent early warning method for unsignaled intersections based on a digital behavioral twin model and a lion group optimization algorithm, which can achieve accurate modeling and dynamic intervention of the behavioral evolution and conflict risks of multiple traffic entities. Through behavioral twin modeling driven by perception data, the system can simulate the interaction process between traffic participants in real time, and introduce an optimization algorithm to generate multi-objective strategy candidate solutions based on the behavioral simulation state. The fitness-driven role switching mechanism improves the global search capability and stability of strategy evolution. After the strategy is generated, the system performs micro-behavior adjustment mapping and simulation verification, which effectively reflects the actual impact of behavioral adjustments on conflict risks. Finally, the scoring results are used to realize the construction of hierarchical warning information and the distribution of edge communications, thereby improving the real-time response capability and traffic safety level of unsignaled intersections.

[0103] In this embodiment, the traffic status data set includes vehicle position coordinates, vehicle speed, driving direction angle, acceleration, as well as pedestrian trajectory information, light intensity, road slipperiness coefficient and visibility index associated with the intersection area, which are used to characterize the joint characteristics of traffic behavior status and environmental perception status.

[0104] The present invention constructs a traffic status data set by introducing multi-dimensional motion parameters of vehicles and pedestrians and environmental status indicators, realizes the joint modeling of traffic behavior and environmental perception status, and provides more comprehensive and accurate data support for subsequent behavior simulation and conflict risk assessment.

[0105] In this embodiment, S2 specifically includes:

[0106] S21. Construct each set of state parameters in the traffic state data set into a multidimensional state vector, and construct a traffic state time series set in chronological order;

[0107] S22. performing normalization and time synchronization on the traffic state time series set to generate a standardized state series set;

[0108] S23. Construct a digital behavior twin model based on a standardized state sequence set, wherein the digital behavior twin model includes a traffic subject behavior unit, an environment interaction unit, a behavior propagation structure, and a state synchronization interface, wherein the traffic subject behavior unit is used to model the state transition behavior of vehicles and pedestrians respectively; the environment interaction unit is used to couple environmental parameters with behavior states; the behavior propagation structure is used to construct a dynamic interaction graph between traffic participants; and the state synchronization interface is used to align the actual traffic state with the state of the digital behavior twin model;

[0109] S24. Establish a behavior propagation mechanism in the digital behavior twin model. The propagation mechanism treats each traffic participant as a graph node, generates a propagation graph based on the adjacency relationship, and performs behavior state propagation calculation according to the state transition probability function:

[0110]

[0111] Among them, A t (i,j) represents the adjacency matrix element between node i and node j at time t, d ij is the spatial distance between the traffic participants corresponding to node i and node j, and δ is the spatial adjacency threshold for propagation judgment;

[0112] S25. Based on the behavior propagation mechanism, a semantic state vector is generated at each time step, and the semantic state vectors are combined into a behavior simulation state set output. The semantic state vector includes the intersection area occupancy status, the path expected area, the conflict probability score, the traffic priority code, the degree of behavior interaction influence, the visibility gap parameter, the yield state mark, the path smoothness index and the speed consistency score.

[0113] The present invention achieves time-series modeling and propagation deduction of the behavioral states of traffic participants by constructing multidimensional traffic state vectors as a time series set and introducing normalization processing and behavior propagation mechanisms. In the propagation mechanism, a propagation graph is constructed based on the spatial adjacency relationship between traffic entities. The propagation relationship is determined by judging whether the distance between nodes in the propagation matrix is ​​less than a threshold, thereby realizing the dynamic transmission of traffic behavior states. The propagation matrix elements in the formula are used to determine whether any two traffic participants have an interactive influence at a certain moment, providing structural support for behavior prediction. The final generated semantic state vector contains conflict scores, priority codes, and path prediction results, providing an accurate behavioral input basis for subsequent strategy optimization and simulation verification.

[0114] In this embodiment, S3 specifically includes:

[0115] S31. Use the behavioral simulation state set as the initial input, construct an optimized population, and evaluate individual performance based on the behavioral fitness function:

[0116]

[0117] Among them, F i is the behavioral fitness score of the i-th individual, λ1 is the behavioral risk score weight, T is the total number of behavioral simulation time steps, t is the behavioral simulation time step, ω t is the timing attenuation factor, is the local behavior risk score of individual i at time t, λ2 is the conflict score weight, is the set of neighbor individuals that have path interactions with individual i, η i,j is the conflict importance weight between individual i and neighbor j, is the conflict score between individual i and neighbor j in the entire simulation path, λ3 is the behavior rationality penalty weight, κ1 is the path deviation penalty weight, is the maximum behavior path deviation distance of individual i, κ2 is the speed jump penalty weight, is the average speed jump degree of individual i, i is the optimized individual number, j is the number of the adjacent traffic participant, and the conflict score of the i-th individual in the t-th round of simulation is

[0118] S32. Based on the distribution of individual behavioral performance in the state space, the optimized population is divided into a set of territorial male lions, a set of female lions, and a set of wandering male lions;

[0119] S33, simulate the process of female lions hunting and male lions patrolling the territory, perform search updates in the individual neighborhood space and the global space respectively, and the position vector of the i-th individual in the t-th iteration is updated by superimposing the current optimal individual guidance term and the Gaussian perturbation term:

[0120]

[0121] in, is the behavioral strategy vector of the i-th individual in the t+1 round, is the behavioral strategy vector of the i-th individual in the t-th round, α is the optimal guidance weight coefficient, is the optimal behavior strategy vector of the population in the current round t, β is the local disturbance control coefficient, and randn() is a random vector obeying the standard normal distribution;

[0122] S34. Perform a crossover strategy between the lioness and the male lion, and generate a new individual through mating operation:

[0123] X child =γ·X male +(1-γ)·X female ;

[0124] Among them, X child is the offspring behavior strategy vector, γ is the strategy inheritance weight factor, X male is the lion’s individual behavior strategy vector, X female is the individual behavior strategy vector of the lioness;

[0125] S35. Execute the role switching mechanism based on the individual fitness change and the behavioral strategy contribution rate. When the conditions are met, update the role label set.

[0126] S36. Update the current lion group individual set and role structure information. When the number of iterations reaches the upper limit, the fitness improvement is lower than the threshold, or the behavioral strategy diversity converges, the optimization process is determined to be completed and the behavioral strategy candidate solution set is output.

[0127] The present invention can comprehensively evaluate the risk trend, path deviation and stability of individuals in the simulation process by constructing a multi-objective behavior fitness function that includes conflict scores, behavior rationality penalties and speed change factors. In the fitness function, the formula evaluates the risk of traffic conflict in the form of time step integration, and at the same time introduces the interaction intensity and conflict score between individuals and neighbors, thereby enhancing the sensitivity to local dense interactions. In addition, the penalty terms constructed by indicators such as individual maximum path deviation and speed mutation can effectively constrain the feasibility of the strategy solution. In the search stage, the lion group optimization algorithm completes the individual position update through the guidance term and Gaussian perturbation term, and generates the offspring strategy vector through crossover operation, thereby improving the evolutionary ability of the population and the solution space exploration ability. Finally, when the fitness no longer improves or the strategy diversity converges, the optimal strategy set is determined and output. The above mechanism realizes fine control and dynamic balance in the optimization process, effectively improving the stability, rationality and search convergence speed of the behavior strategy.

[0128] In this embodiment, the S32 specifically includes:

[0129] S321, construct the optimized population individual set P = {X1, X2, ..., X i ,…,X N}, where X i represents the behavior strategy vector of the i-th behavior strategy individual;

[0130] S322, calculate the fitness score vector F = [F1, F2, ..., F N ], and calculate the average value of the fitness score vector and standard deviation σ F ;

[0131] S323, role division based on individual fitness scores and statistical indicators: Individuals are included in the territory of male lions set M t ,satisfy Individuals are classified into the lionesses set F t , the remaining individuals are classified into the wandering lion set D t ;

[0132] S324, for each individual X i Assign the role label role i ∈{male,female,nomad}, and initialize the behavior history vector set, strategy interaction counter and role stability flag.

[0133] This method calculates the mean and standard deviation of the behavioral fitness scores of individuals in the optimized population and introduces a role partitioning mechanism based on score ranges. Individuals with scores above the mean plus the standard deviation are assigned to the territorial male lion set, those with scores between the upper and lower standard deviations are assigned to the female lion set, and the remainder are assigned to the wandering male lion set. This partitioning approach improves strategy diversity and role stability during the optimization process, helps form a differentiated search structure, and enhances both local and global search capabilities.

[0134] In this embodiment, the S35 specifically includes:

[0135] S351, optimize the population individual set P = {X1, X2, ..., X i ,…,X N}, calculate the fitness change value in, is the individual X in the current round i The fitness score of Score the fitness of the previous round;

[0136] S352. Calculate the contribution rate of individual behavior strategies:

[0137]

[0138] Among them, ρ i is the behavioral strategy contribution rate of the i-th optimized population individual, T is the current cumulative number of behavioral simulation rounds, t is the iteration round number, Score the conflict of individual i in the tth round of simulation, is the basic conflict score when the optimization strategy is not used in the t-th round of simulation, N is the number of individuals in the optimization population, and j is the population traversal index;

[0139] S353. Setting the threshold value ρ of the strategy contribution rate thresh , the average fitness of the current round population F avg and the minimum fitness alert value F low ;

[0140] S354. Execute the role switching mechanism and update the individual role label when any of the following conditions is met: i >ρ thresh , then the current lioness will be promoted to the strategic commander, if ΔF i >0 and The current wandering lion will be promoted to a territory lion. If Then the current individual will be downgraded to a wandering lion;

[0141] S355. Update the individual role tag set, and synchronously update the role stability identifier and the strategy interaction counter.

[0142] This invention introduces a role-switching mechanism driven by the combined contribution rate of behavioral strategies and fitness changes, enabling dynamic role adjustment for individual lions during evolution. The contribution rate formula measures the effectiveness of individual strategies by calculating their normalized contribution to conflict score improvement over multiple scoring rounds. By combining set contribution rate thresholds with upper and lower fitness limits, the system enables promotion and demotion of strategy leaders, enhancing strategy screening capabilities and evolutionary adaptability during the optimization process.

[0143] In this embodiment, the S4 specifically includes:

[0144] S41. Based on the set of behavioral strategy candidate solutions, extract the speed adjustment factor, acceleration change factor, steering angle correction factor, and path priority factor contained in each strategy vector to construct a behavioral strategy parameter matrix;

[0145] S42. Constructing a family of micro-behavior regulation functions Corresponding to the speed, acceleration, direction and path priority adjustment dimensions respectively;

[0146] S43. For each strategy vector, input the corresponding parameter factor into the micro-behavior adjustment function family to calculate the micro-behavior adjustment instruction vector;

[0147] S44. Merge all micro-behavior adjustment instruction vectors to generate a micro-behavior adjustment instruction set, and pass the micro-behavior adjustment instruction set into the digital behavior twin model to enter the subsequent simulation verification process.

[0148] This invention achieves fine-tuning of key behavioral factors, such as speed, acceleration, direction, and path priority, by constructing a family of micro-behavioral adjustment functions. These functions take behavioral strategy parameters as input and output corresponding micro-behavioral adjustment instructions, making candidate strategies executable and meaningful. This adjusted set of behavioral instructions directly drives the simulation units within the digital behavioral twin model, ensuring that the optimized results possess both traffic-physical feasibility and dynamic responsiveness.

[0149] In this embodiment, the S5 specifically includes:

[0150] S51, inputting the micro-behavior adjustment instruction set into each traffic subject behavior unit in the digital behavior twin model to control its micro-behavior adjustment operation within the simulation time sequence;

[0151] S52. During the evolution of the digital behavioral twin model, record the minimum spatial distance between each agent, the predicted trajectory overlap area, the length of the horizontal and vertical time conflict window, and calculate the traffic conflict risk score matrix;

[0152] S53. Perform temporal aggregation on the risk score values ​​at all time steps to construct a global conflict score function:

[0153] S54, according to the conflict scoring function value S risk Output the conflict warning score result set and feed it back to the warning generation and communication distribution process.

[0154] This invention precisely controls the behavioral evolution process by inputting micro-behavior adjustment instructions into each traffic agent's behavioral unit within the digital behavioral twin model. During the simulation, the system constructs a risk score matrix based on parameters such as spatial distance, trajectory overlap, and temporal conflict window. Using a time series aggregation function, it outputs a global conflict score, enabling a comprehensive assessment of traffic conflict trends. The score provides a quantitative basis for subsequent warning generation, improving the timeliness and accuracy of risk identification.

[0155] In this embodiment, S6 specifically includes:

[0156] S61. Receive a conflict warning score result set output by the digital behavior twin model;

[0157] S62. Set a set of conflict risk thresholds, and perform warning level classification based on the hierarchical judgment rules to construct a warning information object set. Each of wi Includes target subject identification, warning level, conflict type description and response time recommendation;

[0158] S63: Based on the edge node deployment structure within the intersection area, call the edge communication scheduling table, perform target matching and communication path selection on the warning information object set, and generate a set of instructions for issuance. Where each i k Including target terminal identification, issuing timestamp and instruction content summary;

[0159] S64. Push the issued instruction set to the corresponding target traffic participant terminal in real time through the edge communication link, and record the issued response status for subsequent feedback verification process.

[0160] This invention builds a collection of conflict warning information objects and implements hierarchical classification based on conflict scoring results and risk thresholds, enhancing the targeted nature of warning content and the ability to prioritize responses. Based on an edge node deployment structure, the system automatically matches target terminals and generates a set of dispatch instructions, enabling rapid transmission of warning information and target identification. Link feedback records terminal response status, creating a closed-loop control system that helps improve warning accuracy and transmission reliability in multi-agent environments.

[0161] Example 1:

[0162] To verify the feasibility of the present invention in practice, the present invention was applied to a typical urban traffic artery and branch road intersection area without signalization. During the morning and evening rush hours, this area has a mixed traffic situation of multiple types of traffic participants. Typical characteristics include motor vehicles, non-motor vehicles, and pedestrians entering the intersection at the same time, lack of signal control and clear markings, which easily lead to traffic flow conflicts. Traditional rule-based response systems rely on the preset threshold rules of edge devices to alert vehicles or pedestrians. They are difficult to deal with dynamic risk identification in scenarios with complex interactions between multi-source traffic status and behavior. In particular, in scenarios where the sight distance at the intersection is limited and the traffic speed fluctuates greatly, the false alarm rate and missed alarm rate are high.

[0163] In this embodiment, a traffic perception unit is deployed to obtain information such as vehicle position, speed, direction angle, acceleration, pedestrian trajectory, ambient lighting, road slip coefficient and visible distance around the intersection. The system first constructs the above data into a traffic status time series, and inputs it into the digital behavior twin model after normalization and synchronization. Within the model, various types of traffic subject behavior units and environmental interaction units work together to dynamically generate semantic state vectors and output a set of behavioral simulation states containing path occupancy, expected trajectory, conflict score and other content. The strategy search population is constructed through the lion group optimization algorithm. The model calculates the multi-objective fitness function for each individual, and dynamically divides and switches roles according to individual fitness changes and contribution rates to improve the diversity of strategy solutions and optimization directionality.

[0164] The experimental scenario simulation runs during a 15-minute rush hour, collecting and updating traffic status data every 5 seconds. During the simulation, the warning performance differences between the traditional fixed rule method and the method of the present invention were compared. The results showed that the system can effectively identify potential conflict trends and generate risk warnings 3.2 seconds in advance, which improves the response time by about 1.9 seconds compared with the traditional system. Among 30 warning events, the system of the present invention accurately identified 27 times, with 1 false alarm and 2 missed alarms, and the recognition accuracy rate reached 90%, which is better than the 73.3% of the traditional method. In scenarios where the number of concurrent vehicles exceeds 20 and the number of mixed participants exceeds 40, the system can still maintain an average warning response delay of less than 300 milliseconds, meeting the real-time traffic warning requirements. The edge communication module sends the warning information to the target vehicle and pedestrian terminals, and the acceptance confirmation response rate reaches more than 94%, and the communication stability is good.

[0165] The following table is a statistical summary of key warning data in this embodiment, showing that the present invention outperforms the prior art in multiple indicators:

[0166] Table 1 Performance comparison statistics of the system of the present invention and the traditional rule warning system

[0167]

[0168]

[0169] As can be seen from the above table, the present invention has significant improvements in multiple key performance indicators in the intelligent warning task of unsignaled intersections compared with the traditional rule response system. First, in terms of warning response time, the average warning lead time of the present invention reaches 3.2 seconds, which is significantly better than the 1.3 seconds of the traditional system, and the lead time is improved by about 1.9 seconds, reflecting the present invention's higher sensitivity to the evolution trend of traffic behavior and stronger forward-looking recognition ability, which helps traffic participants make decisions to slow down or avoid in advance. Secondly, in terms of warning accuracy, the present invention achieved a recognition accuracy rate of 90.0%, successfully identified 27 times in 30 typical conflict scenario tests, with only 1 false alarm and 2 missed alarms. Compared with the 73.3% accuracy rate, 3 false alarms and 5 missed alarms of the traditional method, the stability and robustness of the warning system in complex mixed traffic environments have been significantly improved.

[0170] At the same time, the system of the present invention performs well in terms of concurrent communication capabilities. In the scenario of concurrent traffic of multiple participants, the success rate of edge communication distribution of warning information reached 94.2%, which is 12.7 percentage points higher than the 81.5% of the traditional system, indicating that the present invention has stronger communication scheduling and target matching capabilities in high-density dynamic environments. In addition, in terms of warning response speed, the average response delay of this system is 284 milliseconds, which is significantly faster than the 617 milliseconds of the traditional system, and the overall delay is shortened by more than 50%, ensuring that the warning results can be quickly distributed within the critical time window under high-risk conditions and perceived and processed by traffic participants. The improvement in the above performance is mainly due to the fact that the digital behavior twin model constructed by the present invention can realize refined traffic behavior modeling and timing simulation, and complete the efficient search and dynamic role scheduling of multi-objective conflict risk control strategies through the lion group optimization algorithm.

[0171] In summary, the present invention not only significantly outperforms traditional systems in terms of traffic conflict warning accuracy, response time, and communication success rate, but also enhances the system's modeling and decision-making capabilities for complex traffic dynamics through adaptive optimization of behavioral strategies and digital twin deduction mechanisms, and has good engineering practicality and deployment and promotion value.

[0172] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An intelligent warning system for unsignalized intersections based on digital twins, characterized by: include: Traffic perception data collection module, used to collect data in the intersection area and generate a traffic status data set; The digital behavior twin modeling module is used to build a digital behavior twin model and output a set of behavior simulation states; The conflict risk optimization module is used to perform multi-objective traffic conflict risk optimization operations based on the lion group optimization algorithm and generate a set of candidate behavioral strategy solutions through role division and role switching mechanisms; The role division and switching management module is used to divide the population role structure based on individual fitness scores and behavioral strategy contribution rates; A behavior adjustment mapping module is used to map the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set; The simulation verification and scoring module is used to input the micro-behavior adjustment instruction set into the digital behavior twin model, execute the interactive simulation process, and output the conflict warning score result set; The warning generation and classification module is used to construct a warning information object set based on the conflict warning score result set and the risk threshold set; The edge communication scheduling module is used to perform target matching and path selection on the warning information object set and generate a set of instructions for issuance; The terminal push and feedback module is used to push the issued instruction set to the target traffic participant terminal in real time through the edge communication link and record the issued response status.

2. The intelligent warning system for unsignalized intersections based on digital twins according to claim 1 is characterized in that: The modules are implemented as follows: S1. Collect traffic perception data of the intersection area and generate a traffic status data set; S2. Constructing a digital behavior twin model based on the traffic state data set, wherein the digital behavior twin model includes multiple traffic subject behavior units and environment interaction units, and outputs a behavior simulation state set; S3. Introducing the lion group optimization algorithm, performing multi-objective traffic conflict risk optimization operations on the behavioral simulation state set, generating a set of behavioral strategy candidate solutions, and introducing a fitness-driven role switching mechanism to dynamically adjust individual behavioral roles and strategy combinations; S4, performing behavior adjustment processing based on the behavior strategy candidate solution set, mapping the optimal strategy vector in the behavior strategy candidate solution set into a micro-behavior adjustment instruction set; S5. Input the micro-behavior adjustment instruction set into the digital behavior twin model, perform interactive simulation verification, update the traffic behavior path evolution state, and output the conflict warning score result set; S6. Generate warning information based on the conflict warning score result set, and send the warning information to the target traffic participant terminal through the edge communication link.

3. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The traffic status data set includes vehicle position coordinates, vehicle speed, driving direction angle, acceleration, as well as pedestrian trajectory information, light intensity, road slipperiness coefficient and visibility index associated with the intersection area.

4. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The S2 specifically includes: S21. Construct each set of state parameters in the traffic state data set into a multidimensional state vector, and construct a traffic state time series set in chronological order; S22. performing normalization and time synchronization on the traffic state time series set to generate a standardized state series set; S23. Constructing a digital behavior twin model based on a standardized state sequence set, wherein the digital behavior twin model includes a traffic subject behavior unit, an environment interaction unit, a behavior propagation structure, and a state synchronization interface; S24. Establishing a behavior propagation mechanism in the digital behavior twin model, wherein the propagation mechanism treats each traffic participant as a graph node, generates a propagation graph based on adjacency relationships, and performs behavior state propagation calculations according to a state transition probability function; S25. Based on the behavior propagation mechanism, a semantic state vector is generated at each time step, and the semantic state vectors are combined into a behavior simulation state set output.

5. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The S3 specifically includes: S31. Use the behavioral simulation state set as the initial input, construct an optimized population, and evaluate individual performance based on the behavioral fitness function: Among them, F i is the behavioral fitness score of the i-th individual, λ1 is the behavioral risk score weight, T is the total number of behavioral simulation time steps, t is the behavioral simulation time step, ω t is the timing attenuation factor, is the local behavior risk score of individual i at time t, λ2 is the conflict score weight, is the set of neighbor individuals that have path interactions with individual i, η i,j is the conflict importance weight between individual i and neighbor j, is the conflict score between individual i and neighbor j in the entire simulation path, λ3 is the behavior rationality penalty weight, κ1 is the path deviation penalty weight, is the maximum behavior path deviation distance of individual i, κ2 is the speed jump penalty weight, is the average speed jump degree of individual i, i is the optimized individual number, and j is the number of the adjacent traffic participant; S32. Based on the distribution of individual behavioral performance in the state space, the optimized population is divided into a set of territorial male lions, a set of female lions, and a set of wandering male lions; S33, simulating the process of female lions hunting and male lions patrolling their territory, performing search updates in the individual neighborhood space and the global space respectively. The position vector of the i-th individual in the t-th iteration is updated by superimposing the current optimal individual guidance term and the Gaussian perturbation term; S34. Perform a crossover strategy between the lioness and the male lion, and generate a new individual through mating operation: X child =γ·X male +(1-c)·X female ; Among them, X child is the offspring behavior strategy vector, γ is the strategy inheritance weight factor, X male is the lion’s individual behavior strategy vector, X female is the individual behavior strategy vector of the lioness; S35. Execute the role switching mechanism based on the individual fitness change and the behavioral strategy contribution rate. When the conditions are met, update the role label set. S36. Update the current lion group individual set and role structure information. When the number of iterations reaches the upper limit, the fitness improvement is lower than the threshold, or the behavioral strategy diversity converges, the optimization process is determined to be completed and the behavioral strategy candidate solution set is output.

6. The intelligent warning system for unsignalized intersections based on digital twins according to claim 5 is characterized in that: The S32 specifically includes: S321, construct the optimized population individual set P = {X1, X2, ..., X i ,…,X N }, where X i represents the behavior strategy vector of the i-th behavior strategy individual; S322, calculate the fitness score vector F = [F1, F2, ..., F N ], and calculate the average value of the fitness score vector and standard deviation σ F ; S323, role division based on individual fitness scores and statistical indicators: Individuals are included in the territory of male lions set M t ,satisfy Individuals are classified into the lionesses set F t The remaining individuals are classified into the wandering lion set D t ; S324, for each individual X i Assign the role label role i ∈{male,female,nomad}, and initialize the behavior history vector set, strategy interaction counter and role stability flag.

7. The intelligent warning system for unsignalized intersections based on digital twins according to claim 5 is characterized in that: The S35 specifically includes: S351, optimize the population individual set P = {X1, X2, ..., X i ,…,X N }, calculate the fitness change value in, is the individual X in the current round i The fitness score of Score the fitness of the previous round; S352. Calculate the contribution rate of individual behavior strategies: Among them, ρ i is the behavioral strategy contribution rate of the i-th optimized population individual, T is the current cumulative number of behavioral simulation rounds, t is the iteration round number, Score the conflict of individual i in the tth round of simulation, is the basic conflict score when the optimization strategy is not used in the t-th round of simulation, N is the number of individuals in the optimization population, and j is the population traversal index; S353. Setting the threshold value ρ of the strategy contribution rate thresh , the average fitness of the current round population F avg and the minimum fitness alert value F low ; S354. Execute the role switching mechanism and update the individual role label when any of the following conditions is met: i >ρ thresh , then the current lioness will be promoted to the strategic commander, if ΔF i >0 and The current wandering lion will be promoted to a territory lion. If Then the current individual will be downgraded to a wandering lion; S355. Update the individual role tag set, and synchronously update the role stability identifier and the strategy interaction counter.

8. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The S4 specifically includes: S41. Based on the set of behavioral strategy candidate solutions, extract the speed adjustment factor, acceleration change factor, steering angle correction factor, and path priority factor contained in each strategy vector to construct a behavioral strategy parameter matrix; S42. Constructing a family of micro-behavior regulation functions Corresponding to the speed, acceleration, direction and path priority adjustment dimensions respectively; S43. For each strategy vector, input the corresponding parameter factor into the micro-behavior adjustment function family to calculate the micro-behavior adjustment instruction vector; S44. Merge all micro-behavior adjustment instruction vectors to generate a micro-behavior adjustment instruction set, and pass the micro-behavior adjustment instruction set into the digital behavior twin model.

9. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The S5 specifically includes: S51, inputting the micro-behavior adjustment instruction set into each traffic subject behavior unit in the digital behavior twin model to control its micro-behavior adjustment operation within the simulation time sequence; S52. During the evolution of the digital behavioral twin model, record the minimum spatial distance between each agent, the predicted trajectory overlap area, the length of the horizontal and vertical time conflict window, and calculate the traffic conflict risk score matrix; S53. Perform temporal aggregation on the risk score values ​​at all time steps to construct a global conflict score function: S54, according to the conflict scoring function value S risk Output the conflict warning score result set and feed it back to the warning generation and communication distribution process.

10. The intelligent warning system for unsignalized intersections based on digital twins according to claim 2 is characterized in that: The S6 specifically includes: S61. Receive a conflict warning score result set output by the digital behavior twin model; S62. Set a set of conflict risk thresholds, and perform warning level classification based on the hierarchical judgment rules to construct a warning information object set. Each of w i Includes target subject identification, warning level, conflict type description and response time recommendation; S63: Based on the edge node deployment structure within the intersection area, call the edge communication scheduling table, perform target matching and communication path selection on the warning information object set, and generate a set of instructions for issuance. Each i k Including target terminal identification, issuing timestamp and instruction content summary; S64. Push the issued instruction set to the corresponding target traffic participant terminal in real time through the edge communication link, and record the issued response status.

Images