A traffic state estimator design method and device for multi-agent systems

Abstract

The application discloses a traffic state estimator design method and device for a multi-agent system, comprising: dividing a highway network into a plurality of cells according to preset positions, taking traffic flow density as a traffic state variable, establishing a traffic flow model of the highway network, and determining an output matrix of a traffic sensor; according to observability or detectability of the highway network system composed of a highway network system matrix and an output matrix of each type of traffic sensor, at least one state observer and at least one Kalman filter are constructed; all vehicle density estimation values obtained based on the state observer and the Kalman filter are respectively subjected to weighted summation fusion calculation to obtain first and second total vehicle density estimation values; and the first and second total vehicle density estimation values are subjected to weighted summation fusion calculation to obtain a final vehicle density estimation value. The application improves the state estimation precision of mixed traffic flow and improves the utilization efficiency of different types of sensors.

Description

Technical Field

[0001] This invention relates to the fields of intelligent transportation and multi-source information fusion technology, specifically to a design method and apparatus for a traffic state estimator for a multi-agent system. Background Technology

[0002] As autonomous driving technology continues to develop, scenarios inevitably arise where traditional human-driven vehicles and autonomous vehicles of different levels coexist. The driving rules of autonomous vehicles differ from those of human drivers, leading to changes in mixed traffic flow compared to traditional traffic flow. In particular, changes in the mixing rate of autonomous vehicles render traditional traffic flow state estimation methods inapplicable. Therefore, it is necessary to propose a traffic state estimator design method based on mixed traffic flow for estimating mixed states. Human-driven vehicles and autonomous vehicles can be viewed as two different types of intelligent agents: ordinary intelligent agents and evolved intelligent agents (autonomous vehicles with connectivity). The resulting mixed traffic flow system can be considered a multi-agent system. Evolved intelligent agents possess information collection and transmission capabilities, enabling information transmission with other evolved intelligent agents and roadside control units; they can be considered as mobile sensors.

[0003] Traditional traffic state estimation typically employs a single type of state estimator, such as a state observer or a Kalman filter.

[0004] State observers transform the state estimation problem into a stability problem of the error system between the estimated value and the model-derived value. The optimal criterion for estimation is that the difference approaches zero, and it is commonly used in traffic state estimation. The basic idea is as follows: First, the design of the state observer is completed by using the difference between the observer's measured output and the actual measured output as feedback. Then, at each prediction / estimation step, the difference between the observer's estimated value and the model-derived value is calculated to obtain the error system. Finally, the observer's gain matrix is ​​obtained by solving for the feasible solution of the error system's stability, and the state estimate is then calculated. Because the estimation problem is transformed into a system stability problem, the estimation is performed through the system's asymptotic stability. While the estimation accuracy needs improvement, it is still much higher than that of directly using model derivation.

[0005] The Kalman filter uses minimum mean square error as the optimal estimation criterion to seek a recursive estimation algorithm. Its basic idea is to use a state-space model containing Gaussian white noise, updating the estimates of state variables using the previous time step's estimate and the current time step's observation, to obtain the current time step's estimate. Essentially, the Kalman filter reconstructs the system's state vector from measurements. It recursively follows a sequence of "prediction / pre-estimation—measurement—correction," eliminating random disturbances and reproducing the system's state based on the system's measurements. In the "prediction / pre-estimation" stage, the previous time step's estimate is used as the initial value, and the predicted / estimated value for the current time step is obtained through system model derivation.

[0006] Both types of estimators require design based on a system model and need actual measurements to correct their estimates. This is problematic considering the limited accuracy of a single sensor and its inability to function if it malfunctions. Furthermore, estimating and predicting the system state using a system model is equivalent to open-loop control; without feedback error correction, the estimation and prediction errors are relatively large and fail to reflect the true system state.

[0007] With the development of sensing technology, the devices capable of collecting traffic information in highway networks are becoming increasingly diverse, such as ETC data collection systems, video collection systems, microwave collection systems, and connected autonomous vehicles. Therefore, multi-source information fusion methods can be considered to obtain more accurate sensor measurement information for the correction of hybrid estimators. Considering the advantages and disadvantages of the two estimators mentioned above, this paper proposes a novel hybrid state estimator for multi-agent systems by fusing their estimation results to solve the problem of highway mixed traffic state estimation and prediction. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a method and apparatus for designing a traffic state estimator for a multi-agent system.

[0009] This invention discloses a method for designing a traffic state estimator for a multi-agent system, comprising:

[0010] The highway network is divided into several cells according to preset locations; wherein, the preset locations include the locations of entrance and exit ramps, ETC gantries, and lane number changes.

[0011] Based on the partitioned cells, a traffic flow model for the highway network is established using traffic flow density as the traffic state variable.

[0012] The vehicle density of the cell containing the traffic sensor is calculated based on the information collected by different types of traffic sensors, and the output matrix of different types of traffic sensors in the highway network is determined based on the vehicle density.

[0013] Based on the observability or detectability of the highway network system, which consists of the highway network system matrix and the output matrix of each type of traffic sensor, construct at least one state observer and at least one Kalman filter.

[0014] The traffic flow density estimate of the highway network is calculated based on the state observer. The first total vehicle density estimate is obtained by weighted summation of all vehicle density estimates.

[0015] Traffic flow density estimates for the highway network are calculated based on Kalman filters. A second total vehicle density estimate is obtained by weighted summation of all vehicle density estimates.

[0016] The first total vehicle density estimate and the second total vehicle density estimate are weighted and summed to obtain the final vehicle density estimate.

[0017] The traffic conditions of the highway network are estimated based on the final vehicle density estimate.

[0018] As a further improvement to the present invention, the traffic flow model for the highway network is as follows:

[0019]

[0020] Where, x∈R n Let u represent the traffic flow density vector, and n represent the number of cells in the partition; u∈R p Represents the system's control input, y∈R m denoted by , m represents the number of road segments where traffic sensors are deployed; A is the system matrix; B is the system input matrix; C is the output matrix related to the traffic sensors deployed in the road network; F is a constant matrix; σ(k) represents the mixing rate of different autonomous vehicles; ω(k) is the system noise, ω(k) ~ N(0,Q); v(k) is the measurement noise, v(k) ~ N(0,R), both system noise and system noise are Gaussian white noise; k is the time.

[0021] As a further improvement to the present invention, the output matrices corresponding to various types of traffic sensors include:

[0022]

[0023] Where l represents the type of traffic sensor, C l It is the output matrix corresponding to the respective sensor;

[0024]

[0025] As a further improvement of the present invention, based on the observability or detectability of the highway network system, which consists of a highway network system matrix and the output matrix of each type of traffic sensor, at least one state observer and at least one Kalman filter are constructed; including:

[0026] The determination is based on the highway network system matrix A and the output matrix C corresponding to each type of traffic sensor. i The high-speed road network system (A, C) i Is it observable or detectable?

[0027] If there are m observable or detectable high-speed road network systems, then construct m state observers and Kalman filters, with each high-speed road network system corresponding to one state observer and one Kalman filter; recombine the output matrices of the remaining nm unobservable or undetectable high-speed road network systems to obtain several new high-speed road network systems.

[0028] From the aforementioned new highway network systems, select the target highway network system with the fewest traffic sensors and the most uniform distribution of traffic sensors, and construct a corresponding state observer and Kalman filter based on the target highway network system.

[0029] As a further improvement to the present invention, the state observer i is represented as:

[0030]

[0031] in, f represents the road network density estimated by the i-th proportional-integral state observer; i L represents the integral term of the observer; σ(k)-i and G σ(k)-i Let X represent the proportional gain matrix and integral gain matrix of the i-th observer.

[0032] As a further improvement to the present invention, the Kalman filter i is represented as:

[0033]

[0034] in, This represents the estimated value of the highway network density by the i-th Kalman filter.

[0035] As a further improvement of the present invention, based on the state observer, the estimated traffic flow density of the highway network is calculated, and a weighted summation of all vehicle density estimates is performed to obtain a first total vehicle density estimate; including:

[0036] m+1 traffic flow density estimates are calculated based on m+1 state observers;

[0037] For each traffic flow density estimate, a corresponding weight is determined. The weight is determined by performing iterative calculations within each sampling period and finding the m+1 weight combination pair corresponding to the minimum standard deviation as the final weight.

[0038] The first total vehicle density estimate is obtained by merging the m+1 traffic flow density estimates with weighted summation according to the determined weights.

[0039] As a further improvement of the present invention, the traffic flow density estimate of the highway network is calculated based on the Kalman filter, and a weighted summation of all vehicle density estimates is performed to obtain a second total vehicle density estimate; including:

[0040] m+1 traffic flow density estimates are calculated based on m+1 Kalman filters;

[0041] For each traffic flow density estimate, a corresponding weight is determined. The weight is determined by performing iterative calculations within each sampling period and finding the m+1 weight combination pair corresponding to the minimum standard deviation as the final weight.

[0042] The m+1 traffic flow density estimates are weighted and summed according to the determined weights to obtain the second total vehicle density estimate.

[0043] As a further improvement of the present invention, a weighted summation calculation is performed on the first total vehicle density estimate and the second total vehicle density estimate to obtain the final vehicle density estimate; including:

[0044] For each traffic flow density estimate, a corresponding weight is determined. The weight is determined by iteratively calculating within each sampling period and finding the two weight pairs corresponding to the minimum standard deviation as the final weights.

[0045] The first total vehicle density estimate and the second total vehicle density estimate are weighted and summed according to the determined weights to obtain the final vehicle density estimate.

[0046] The present invention also discloses a design apparatus for a traffic state estimator for a multi-agent system, which is used to implement the above-described design method for a traffic state estimator for a multi-agent system.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] This invention integrates the advantages of state observers and Kalman filters, and combines the characteristics of multiple types of traffic sensors to design a state observer and a Kalman filter for multi-information fusion respectively. Then, the results of the two types of estimators are fused and calculated to obtain more accurate results. It improves the state estimation accuracy of mixed traffic flow and increases the utilization efficiency of different types of sensors such as connected autonomous vehicles. Attached Figure Description

[0049] Figure 1 This is a flowchart of a traffic state estimator design method for a multi-agent system disclosed in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of highway cell partitioning disclosed in one embodiment of the present invention;

[0051] Figure 3 This is a schematic diagram of the deployment of various sensors on a highway according to one embodiment of the present invention;

[0052] Figure 4 This is a flowchart illustrating the design of a multi-source fusion state observer according to one embodiment of the present invention.

[0053] Figure 5 This is a flowchart of a multi-source fusion Kalman filter design according to an embodiment of the present invention;

[0054] Figure 6 This is a logic diagram of a hybrid state estimator disclosed in one embodiment of the present invention.

[0055] In the picture:

[0056] 1. ETC gantry; 2. Video sensing device; 3. Roadside unit; 4. Roadside computing facility; 5. Autonomous vehicle. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] The present invention will now be described in further detail with reference to the accompanying drawings:

[0059] like Figure 1 , 6 As shown, this invention provides a method for designing a traffic state estimator for a multi-agent system, comprising:

[0060] Step 1: Divide the highway network into several cells according to preset locations; wherein, the preset locations include, but are not limited to, one or more of the following: entrance and exit ramp locations, ETC gantry locations, lane number change locations, and curvature radius change locations;

[0061] Specifically, it includes:

[0062] The highway is divided into several segments based on the locations of entrance and exit ramps, changes in curvature radius, ETC gantry locations, and changes in lane number. Each segment is called a cell, and these cells are sequentially numbered to facilitate the marking of traffic sensor deployment locations. Figure 2 As shown; where there are entrance and exit ramps, the boundary between two adjacent cells is used; the ETC gantry is also used as the boundary between two adjacent cells; and the location where the number of lanes changes is also used as the boundary between two adjacent cells.

[0063] Step 2: Based on the partitioned cells, a traffic flow model of the highway network is established using traffic flow density as the traffic state variable;

[0064] Specifically:

[0065] The traffic flow model for the highway network is as follows:

[0066]

[0067] Where, x∈R n This represents the traffic flow density vector, where n represents the number of cells in the partition; u∈R p This represents the system's control input, such as the traffic flow entering the main road through the entrance ramp; y∈R m denoted by , m represents the number of road segments where traffic sensors are deployed; A is the system matrix, which is related to the system's state; B is the system's input matrix; C is the output matrix, which is related to the traffic sensors deployed in the road network; F is a constant matrix; σ(k) represents the mixing rate of different autonomous vehicles; ω(k) is the system noise, ω(k) ~ N(0,Q); v(k) is the measurement noise, v(k) ~ N(0,R), and both system noise and system noise are Gaussian white noise; k is the time interval.

[0068] Step 3: Calculate the vehicle density of the cell where the traffic sensor is located based on the information collected by different types of traffic sensors, and determine the output matrix of different types of traffic sensors in the highway network based on the vehicle density.

[0069] Specifically:

[0070] like Figure 3As shown, the ETC gantry 1 is equipped with a video sensor 2 and a roadside unit 3. The video sensor 2 is used to collect traffic flow data on the road and automatically convert and calculate vehicle density. The roadside unit 3 is a device in the ETC system that communicates with the on-board unit to realize vehicle identification. Through this device, information such as vehicle speed and traffic flow can be obtained. The roadside computing facility 4 is mainly used to store information collected by different types of sensors, perform observability / detectability judgment and construct Kalman filters for multi-source information fusion, and transmit the final estimation results to the central control center. 5 is an autonomous vehicle with network connectivity, used to collect vehicle speed data.

[0071] For traffic sensors that simultaneously include multiple types such as video acquisition devices, motion sensors, and ETC gantry systems, the corresponding output equation is:

[0072]

[0073] Where l represents the type of traffic sensor, C l It is the output matrix corresponding to the respective sensor.

[0074] Since different types of sensors collect different traffic information, they need to be uniformly converted into traffic flow density through corresponding conversion relationships. For example, ETC can collect the instantaneous speed of vehicles and traffic flow, microwave detectors can collect traffic flow, and connected autonomous vehicles can calculate the speed of their cells. The traffic flow and speed collected by the above sensors can all be converted into road segment density as the measurement output result. Therefore, the output matrix corresponding to different types of sensors can adopt a unified representation method, as follows:

[0075]

[0076] That is, if a sensor is placed in the i-th cell, the corresponding element in its output matrix is ​​1; otherwise, it is 0.

[0077] The calculation and conversion between measured values ​​and state variables specifically include:

[0078] 1. For cells with ETC gantries at both the entrance and exit, the vehicle density of the cell can be calculated from the traffic flow collected by the ETC at the entrance and exit within one sampling period, i.e.:

[0079]

[0080] Where, ρ i Let L represent the density of cell i, T represent the period (in seconds), and L represent the density of cell i. i q represents the length of cell i (in meters). i-in A traffic flow q entering cell i within a period. i-outA vehicle flow rate using outflow cell i within a cycle;

[0081] The vehicle density calculated by the above formula can be regarded as the actual measured value of cell i. Therefore, the corresponding element in the output matrix C is 1, otherwise it is 0.

[0082] 2. For autonomous vehicles with intelligent connectivity within a cell, the average speed of that cell can be calculated from the speed of each vehicle, i.e.

[0083]

[0084] Where n is the number of connected autonomous vehicles in cell i, and V j Let be the speed of the j-th connected autonomous vehicle.

[0085] Then, based on the relationship between speed and density, the vehicle density of the cell is further calculated.

[0086] Step 4, Proportional-Integral State Observer Design: Based on the observability or detectability of the highway network system, which consists of the highway network system matrix and the output matrix of each type of traffic sensor, construct at least one state observer; calculate the traffic flow density estimate of the highway network based on the state observer, and perform a weighted summation fusion calculation on all vehicle density estimates to obtain the first total vehicle density estimate.

[0087] like Figure 4 As shown, it specifically includes:

[0088] The determination is based on the highway network system matrix A and the output matrix C corresponding to each type of traffic sensor. i The high-speed road network system (A, C) i Is it observable or detectable? If there are m observable or detectable highway network systems, then construct m state observers; state observer i is represented as:

[0089]

[0090] in, f represents the road network density estimated by the i-th proportional-integral state observer; i L represents the integral term of the observer; σ(k)-i and G σ(k)-i Represent the proportional gain matrix and integral gain matrix of the i-th observer;

[0091] The designed observer is used to estimate / predict the corresponding highway network traffic flow density. The output matrices of the remaining unobservable / undetectable systems at the nm level are recombine to obtain several new systems. The system (A) is then evaluated. σ(k) Ci If the traffic sensors (i = 1, ..., q) are observable / detectable, and several new combinations of systems satisfy the observability / detectability criteria, then the system corresponding to the traffic sensor with the fewest number of sensors and the most evenly distributed sensor configuration (A) is selected. σ(k) C j Design the observer and estimate / predict the corresponding traffic flow density of the highway network. Thus far, a total of m+1 estimated / predicted traffic flow densities have been obtained.

[0092] To obtain a more accurate estimated / predicted density, this invention assigns a weight to each traffic flow density estimate. The weights are determined as follows: a cyclical calculation is performed within each sampling period to find the m+1 weight pairs corresponding to the minimum standard deviation, which are then used as the final weights. The m+1 traffic flow density estimates are then weighted and summed according to the determined weights to obtain the first total vehicle density estimate. Specifically, this includes:

[0093] 1. Assume that each observer's estimated / predicted value has a weight λ. S-i Since i = 1, ..., m+1, we can calculate m+1 weight combinations {λ}. S-1 ,λ S-2 ,…,λ S-m+1 The weighted summation of traffic flow densities is then:

[0094]

[0095] 2. Calculate the estimation / prediction results for each of the m+1 observers. and The difference Δ S-i i = 1, ..., m+1, that is

[0096]

[0097] 3. Perform iterative calculations within each sampling period to find the m+1 weight combination pair {λ1,λ2,…,λ} corresponding to the minimum standard deviation. m+1} as the final weight:

[0098]

[0099] 4. Based on the obtained weight λ i Calculate the final estimated / predicted traffic flow:

[0100]

[0101] Step 5: Kalman filter design: Based on the observability or detectability of the highway network system, which consists of the highway network system matrix and the output matrix of each type of traffic sensor, construct at least one Kalman filter; calculate the traffic flow density estimate of the highway network based on the Kalman filter, and perform a weighted summation fusion calculation on all vehicle density estimates to obtain the second total vehicle density estimate;

[0102] like Figure 5 As shown, it specifically includes:

[0103] The determination is based on the highway network system matrix A and the output matrix C corresponding to each type of traffic sensor. i The high-speed road network system (A, C) i Whether it is observable or detectable; if there are m observable or detectable high-speed road network systems, then construct m Kalman filters;

[0104] The Kalman filter i is shown below:

[0105] 1. Based on the highway network model, perform a preliminary estimation of the Kalman filter, i.e., the prior state estimate of the Kalman filter:

[0106]

[0107] in, This represents the pre-estimated value of the highway network density by the i-th Kalman filter, also known as the prior state estimate.

[0108] 2. Calculate the prior error covariance matrix of the Kalman filter.

[0109]

[0110] 3. Calculate the gain matrix G of the Kalman filter. K-i :

[0111]

[0112] 4. Based on the pre-estimated values ​​of the Kalman filter Prior error covariance matrix and gain matrix G K-i Update the posterior state estimate to obtain the final estimate of Kalman filter i.

[0113] 5. Update posterior error covariance

[0114]

[0115] Finally, by repeating the above steps in a loop, the estimated value of the Kalman filter for each type of sensor can be calculated.

[0116] The corresponding highway network traffic flow density is estimated using the constructed Kalman filter. The output matrices of the remaining unobservable / undetectable systems at the nm level are recombine to obtain several new systems. The system (A) is then evaluated. σ(k) C i If the traffic sensors (i = 1, ..., q) are observable / detectable, and several new combinations of systems satisfy the observability / detectability criteria, then the system corresponding to the traffic sensor with the fewest number of sensors and the most evenly distributed sensor configuration (A) is selected. σ(k) C j The construction of Kalman filters and estimation of corresponding traffic flow densities in highway networks. Thus far, a total of m+1 traffic flow density estimates have been obtained.

[0117] To obtain a more accurate estimated / predicted density, this invention assigns a weight to each traffic flow density estimate. The weights are determined as follows: a cyclical calculation is performed within each sampling period to find the m+1 weight pairs corresponding to the minimum standard deviation, which are then used as the final weights. The m+1 traffic flow density estimates are then weighted and summed according to the determined weights to obtain a second total vehicle density estimate. Specifically, this includes:

[0118] 1) Assume that each Kalman filter estimate has a weight λ. K-i Since i = 1, ..., m+1, we can calculate m+1 weight combinations {λ}. K-1 ,λ K-2 ,…,λ K-m+1 The weighted summation of traffic flow densities is then:

[0119]

[0120] 2) Calculate the estimation results for m+1 Kalman filters respectively. and The difference Δ K-i i = 1, ..., m+1, that is

[0121]

[0122] 3) Perform iterative calculations within each sampling period to find the m+1 weight combination pair {λ} corresponding to the minimum standard deviation. K-1 ,λ K-2 ,…,λ K-m+1} as the final weight:

[0123]

[0124] 4) Based on the obtained weight λ K-i Calculate the final estimate of traffic flow:

[0125]

[0126] Step 6, Hybrid Estimator Design: The first total vehicle density estimate and the second total vehicle density estimate are weighted and summed to obtain the final vehicle density estimate. The traffic conditions of the highway network are estimated based on the final vehicle density estimate.

[0127] Specifically, it includes:

[0128] 1. Assume the weights of the state observer and the Kalman filter estimates are η and η, respectively. S and η K Therefore, two weight combinations {η} can be calculated. S ,η K The weighted summation of traffic flow densities is then:

[0129]

[0130] 2. Calculate the estimation results for the state observer and the Kalman filter respectively. and The difference Δ S and Δ K ,Right now

[0131]

[0132]

[0133] 3. Perform iterative calculations within each sampling period to find the two weight pairs {η} corresponding to the minimum standard deviation. S ,η K} as the final weight:

[0134]

[0135] 4. Based on the obtained weight η S and η K The final estimated value of traffic flow density is calculated as follows:

[0136]

[0137] This invention provides a design apparatus for a traffic state estimator for a multi-agent system, comprising:

[0138] The road network segmentation unit is used to implement step 1 in the traffic state estimator design method;

[0139] The road network modeling unit is used to implement steps 2 and 3 in the traffic state estimator design method.

[0140] The state observer design unit is used to implement step 4 in the traffic state estimator design method.

[0141] Kalman filter design unit, used to implement step 5 in the traffic state estimator design method;

[0142] The hybrid estimator design unit is used to implement step 6 in the traffic state estimator design method.

[0143] The advantages of this invention are:

[0144] This invention integrates the advantages of state observers and Kalman filters, and combines the characteristics of multiple types of traffic sensors to design a state observer and a Kalman filter for multi-information fusion respectively. Then, the results of the two types of estimators are fused and calculated to obtain more accurate results. It improves the state estimation accuracy of mixed traffic flow and increases the utilization efficiency of different types of sensors such as connected autonomous vehicles.

[0145] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for designing a traffic state estimator for a multi-agent system, characterized in that, include: The highway network is divided into several cells according to preset locations; wherein, the preset locations include the locations of entrance and exit ramps, ETC gantries, and lane number changes. Based on the partitioned cells, a traffic flow model for the highway network is established using traffic flow density as the traffic state variable; the traffic flow model for the highway network is as follows: in, Represents the traffic flow density vector. Indicates the number of cells in the division; Indicates the system's control input, This indicates traffic parameters that can be directly acquired by traffic sensors. This indicates the number of road sections where traffic sensors are deployed; It is a system matrix. It is the system's input matrix. It is the output matrix related to traffic sensors deployed in the road network. It is a constant matrix. This indicates the mixing rate of different autonomous vehicles. It's system noise. ; It measures noise. Both system noise and system noise are Gaussian white noise. k For a specific moment; The vehicle density of the cell containing the traffic sensor is calculated based on the information collected by different types of traffic sensors, and the output matrix of different types of traffic sensors in the highway network is determined based on the vehicle density. Based on the observability or detectability of the highway network system, which consists of the highway network system matrix and the output matrices of each type of traffic sensor, at least one state observer and at least one Kalman filter are constructed; specifically, this includes: determining the observability or detectability of the highway network system matrix. A and the output matrix corresponding to each type of traffic sensor C i The high-speed road network system ( A , C i Is it observable or detectable? If so... m A high-speed road network system that is observable or detectable will be constructed. m Each highway network system has one state observer and one Kalman filter; the remaining... n-m The output matrices of an unobservable or undetectable high-speed road network system are recombined to obtain several new high-speed road network systems; from these new high-speed road network systems, a target high-speed road network system with the fewest traffic sensors and uniform distribution of traffic sensors is selected, and a corresponding state observer and Kalman filter are constructed based on the target high-speed road network system. Traffic flow density estimates for the highway network are calculated based on the state observer. The first total vehicle density estimate is obtained by weighted summation of all vehicle density estimates. Traffic flow density estimates for the highway network are calculated based on Kalman filters. A second total vehicle density estimate is obtained by weighted summation of all vehicle density estimates. The first total vehicle density estimate and the second total vehicle density estimate are weighted and summed to obtain the final vehicle density estimate. The traffic conditions of the highway network are estimated based on the final vehicle density estimate.

2. The traffic state estimator design method for multi-agent systems as described in claim 1, characterized in that, The output matrices for each type of traffic sensor include: in, Indicates the types of traffic sensors. It is the output matrix corresponding to the respective sensor; 。 3. The method for designing a traffic state estimator for a multi-agent system as described in claim 2, characterized in that, State observer i Represented as: in, Indicates the first i The road network density estimated by a proportional-integral state observer; Represents the integral term of the observer; and Indicates the first i The proportional gain matrix and integral gain matrix of each observer.

4. The method for designing a traffic state estimator for a multi-agent system as described in claim 2, characterized in that, Kalman filter i Represented as: in, Indicates the first i The estimated value of highway network density by a Kalman filter.

5. The method for designing a traffic state estimator for a multi-agent system as described in claim 2, characterized in that, Traffic flow density estimates for the highway network are calculated based on the state observer. The first total vehicle density estimate is obtained by weighted summation of all vehicle density estimates. include: based on The state observers calculated the following: One traffic flow density estimate; A weight is assigned to each traffic flow density estimate; the weight is determined by iterative calculation within each sampling period to find the value corresponding to the minimum standard deviation. m +1 weight combination is used as the final weight; The traffic flow density estimates are weighted and summed according to the determined weights to obtain the first total vehicle density estimate.

6. The method for designing a traffic state estimator for a multi-agent system as described in claim 2, characterized in that, Traffic flow density estimates for the highway network are calculated based on Kalman filters. A second total vehicle density estimate is obtained by weighted summation of all vehicle density estimates. include: based on The Kalman filter is calculated to obtain One traffic flow density estimate; A weight is assigned to each traffic flow density estimate; the weight is determined by iterative calculation within each sampling period to find the value corresponding to the minimum standard deviation. m +1 weight combination is used as the final weight; The traffic flow density estimates are weighted and summed according to the determined weights to obtain the second total vehicle density estimate.

7. The method for designing a traffic state estimator for a multi-agent system as described in claim 1, characterized in that, The final vehicle density estimate is obtained by weighted summation of the first and second total vehicle density estimates; including: For each traffic flow density estimate, a corresponding weight is determined. The weight is determined by iteratively calculating within each sampling period and finding the two weight pairs corresponding to the minimum standard deviation as the final weights. The first total vehicle density estimate and the second total vehicle density estimate are weighted and summed according to the determined weights to obtain the final vehicle density estimate.

8. A design device for a traffic state estimator for a multi-agent system, characterized in that, This is a method for designing a traffic state estimator for a multi-agent system as described in any one of claims 1 to 7.

Images