An artificial intelligence-based second-level adaptive traffic signal timing optimization system

Abstract

The present invention provides an artificial intelligence-based second-level adaptive traffic signal timing optimization system, which relates to the field of electrical digital data processing, including a data acquisition module, an edge computing control module, a comprehensive model analysis module and a feedback optimization management module. The data acquisition module is used to collect traffic data at intersections, the edge computing control module obtains phase information based on traffic characteristic data processing and controls traffic lights, the comprehensive model analysis module is used to obtain traffic effect data of all intersections and optimize the control model. The feedback optimization management module is used to feed back the optimization results to the edge computing control module; the system can adjust the phase time in real time according to road conditions to improve the circulation efficiency of traffic intersections.

Description

Technical Field

[0001] The invention relates to the field of electrical digital data processing, and in particular to an artificial intelligence-based second-level adaptive traffic signal timing optimization system. Background Art

[0002] In the existing traffic system, the duration of traffic lights is pre-set and fixed. Although setting different times in different time periods can improve circulation efficiency, this method is too mechanical and cannot adjust the phase time according to actual conditions. It cannot solve the problem of traffic congestion when encountering special circumstances. Therefore, a signal light timing system that can adaptively adjust road conditions is needed to better improve the traffic efficiency of vehicles at intersections.

[0003] The foregoing discussion of the background art is intended only to facilitate an understanding of the present invention. This discussion does not acknowledge or admit that any of the material referred to is part of the common general knowledge.

[0004] Now many traffic control systems have been developed. After a lot of searching and reference, it is found that the existing traffic control system is such as the system disclosed in the publication number CN110766958B. These system methods generally include: a first sensor system at a first intersection, a cloud and a second signal light at a second intersection. The first sensor system monitors vehicles coming and going in a first direction at the first intersection, and transmits first vehicle data and the first allowed passage time of the first intersection in the first direction to the cloud; after processing the data, the cloud sends the second signal light at the second intersection a parameter for allowing passage in the first direction; the allowed passage parameter satisfies: T2 = T1 + tx1 + t12, T2 is the starting time point when the second signal light allows passage in the first direction, T1 is the starting time point when the first intersection allows passage, tx1 is the first passage time, and t12 is the driving time between the first intersection and the second intersection in the first direction. However, the adjustment method of the system with time is too simple and cannot cope with complex road condition information. Summary of the invention

[0005] The purpose of the present invention is to propose a second-level adaptive traffic signal timing optimization system based on artificial intelligence to address the existing deficiencies.

[0006] The present invention adopts the following technical solution:

[0007] An artificial intelligence-based second-level adaptive traffic signal timing optimization system, including a data acquisition module, an edge computing control module, a comprehensive model analysis module, and a feedback optimization management module;

[0008] The data acquisition module is used to collect traffic data at the intersection, the edge computing control module obtains phase information based on traffic feature data processing and controls the traffic lights, the comprehensive model analysis module is used to obtain traffic effect data of all intersections and optimize the control model, and the feedback optimization management module is used to feed back the optimization results to the edge computing control module;

[0009] The data acquisition module includes an intelligent sensing unit, a traffic feature analysis unit and a traffic effect analysis unit, wherein the intelligent sensing unit is used to sense vehicle information at the intersection, the traffic feature analysis unit is used to analyze the sensed information to obtain feature data, and the traffic effect analysis unit is used to analyze the sensed information to obtain effect data;

[0010] The edge computing control module includes a basic phase management unit, a feature data storage unit and a phase adjustment processing unit. The basic phase management unit is used to set the phase basic information and control the signal light based on the phase information. The feature data storage unit is used to store the feature data corresponding to each phase. The phase adjustment processing unit processes the feature data based on the control model to obtain the phase adjustment time and sends it to the basic phase management unit.

[0011] The comprehensive model analysis module includes an effect data storage unit, an intersection evaluation unit and a model optimization unit, wherein the effect data storage unit is used to receive and save effect data of each intersection, the intersection evaluation unit is used to evaluate the traffic status of each intersection, and the model optimization unit optimizes the parameters of the control model based on the evaluation result;

[0012] Furthermore, the basic phase management unit includes a phase mapping register, a pointer movement processor and a phase switching processor, wherein the phase mapping register is used to store the mapping relationship between each phase and the signal light and the basic duration of each phase, the pointer movement processor is used to indicate the current phase and the corresponding time point, and the phase switching processor is used to determine whether to enter the next phase and send a color change signal to the signal light when entering the next phase;

[0013] Further, the pointer movement processor is represented by a pointer vector (Np, T), wherein Np represents the phase sequence number, T is the duration of the pointer being in the phase, and T increases naturally in seconds through a timer;

[0014] The phase switching processor calculates the switching indication value Vs according to the following formula:

[0015] Vs=T 0 +ΔT-T;

[0016] Among them, T 0is the basic duration of the current phase, and ΔT is the phase adjustment value;

[0017] When Vs is less than or equal to 0, it indicates switching to the next phase, the pointer vector Np changes to the next phase number, and T is reset and then restarted;

[0018] Further, the phase adjustment processing unit includes a phase data matching processor, a model calculation processor and an adjustment output processor, wherein the phase data matching processor obtains characteristic data of the corresponding lane based on the current phase, the model calculation processor is used to run the control model to obtain the phase adjustment value, and the adjustment output processor is used to output the phase adjustment value to the basic phase management unit;

[0019] The model calculation processor calculates the phase adjustment value according to the following formula:

[0020]

[0021] Among them, X i represents the value of the i-th eigenvalue, α i represents the i-th single coefficient, β i represents the coefficient of the i-th quadratic term, δ ij represents the interaction coefficient between the i-th feature item and the j-th feature item, T' represents the offset time;

[0022] When the phase adjustment value exceeds the upper limit, the adjustment output processor directly outputs the upper limit value;

[0023] Further, the intersection evaluation unit includes a data sorting processor and an evaluation calculation processor, the data sorting processor is used to classify and sort the effect data according to time periods and intersections, and the evaluation calculation processor is used to calculate the efficiency index of the intersection in the corresponding time period;

[0024] The evaluation calculation processor calculates the efficiency index P according to the following formula:

[0025]

[0026] Among them, n 0 is the number of vehicles whose waiting time is 0, n a is the number of all vehicles in the corresponding period, T i is the i-th non-zero waiting time, and Y is the idle time ratio;

[0027] The efficiency index of each intersection in the corresponding time period is sent as a sample to the model optimization unit.

[0028] The beneficial effects achieved by the present invention are:

[0029] This system adjusts the duration of each phase by collecting and analyzing four types of feature data. It can make targeted adjustments as road conditions change to improve the flow efficiency at intersections. By collecting and analyzing the effect data of all intersections, the parameters in the control model are optimized and adjusted according to the analysis results, so that the adjustment effect of the phase duration can be continuously improved.

[0030] To further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only for reference and description and are not intended to limit the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic diagram of the overall structural framework of the present invention;

[0032] Figure 2 This is a schematic diagram of the data acquisition module of the present invention;

[0033] Figure 3 This is a schematic diagram of the edge computing control module of the present invention;

[0034] Figure 4 This is a schematic diagram of the composition of the comprehensive model analysis module of the present invention;

[0035] Figure 5 This is a schematic diagram of the feedback optimization management module of the present invention. DETAILED DESCRIPTION

[0036] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0037] Embodiment 1.

[0038] This embodiment provides a second-level adaptive traffic signal timing optimization system based on artificial intelligence, combined with Figure 1 , including data acquisition module, edge computing control module, comprehensive model analysis module and feedback optimization management module;

[0039] The data acquisition module is used to collect traffic data at the intersection, the edge computing control module obtains phase information based on traffic feature data processing and controls the traffic lights, the comprehensive model analysis module is used to obtain traffic effect data of all intersections and optimize the control model, and the feedback optimization management module is used to feed back the optimization results to the edge computing control module;

[0040] The data acquisition module includes an intelligent sensing unit, a traffic feature analysis unit and a traffic effect analysis unit, wherein the intelligent sensing unit is used to sense vehicle information at the intersection, the traffic feature analysis unit is used to analyze the sensed information to obtain feature data, and the traffic effect analysis unit is used to analyze the sensed information to obtain effect data;

[0041] The edge computing control module includes a basic phase management unit, a feature data storage unit and a phase adjustment processing unit. The basic phase management unit is used to set the phase basic information and control the signal light based on the phase information. The feature data storage unit is used to store the feature data corresponding to each phase. The phase adjustment processing unit processes the feature data based on the control model to obtain the phase adjustment time and sends it to the basic phase management unit.

[0042] The comprehensive model analysis module includes an effect data storage unit, an intersection evaluation unit and a model optimization unit, wherein the effect data storage unit is used to receive and save effect data of each intersection, the intersection evaluation unit is used to evaluate the traffic status of each intersection, and the model optimization unit optimizes the parameters of the control model based on the evaluation result;

[0043] The basic phase management unit includes a phase mapping register, a pointer movement processor and a phase switching processor. The phase mapping register is used to store the mapping relationship between each phase and the signal light and the basic duration of each phase. The pointer movement processor is used to indicate the current phase and the corresponding time point. The phase switching processor is used to determine whether to enter the next phase and send a color change signal to the signal light when entering the next phase.

[0044] The pointer movement processor is represented by a pointer vector (Np, T), wherein Np represents the phase number, T is the duration of the pointer being in the phase, and T increases naturally in seconds through a timer;

[0045] The phase switching processor calculates the switching indication value Vs according to the following formula:

[0046] Vs=T 0 +ΔT-T;

[0047] Among them, T 0 is the basic duration of the current phase, and ΔT is the phase adjustment value;

[0048] When Vs is less than or equal to 0, it indicates switching to the next phase, the pointer vector Np changes to the next phase number, and T is reset and then restarted;

[0049] The phase adjustment processing unit includes a phase data matching processor, a model calculation processor and an adjustment output processor, wherein the phase data matching processor obtains characteristic data of the corresponding lane based on the current phase, the model calculation processor is used to run the control model to obtain a phase adjustment value, and the adjustment output processor is used to output the phase adjustment value to the basic phase management unit;

[0050] The model calculation processor calculates the phase adjustment value according to the following formula:

[0051]

[0052] Among them, X i represents the value of the i-th eigenvalue, α i represents the i-th single coefficient, β i represents the coefficient of the i-th quadratic term, δ ij represents the interaction coefficient between the i-th feature item and the j-th feature item, T' represents the offset time;

[0053] When the phase adjustment value exceeds the upper limit, the adjustment output processor directly outputs the upper limit value;

[0054] The intersection evaluation unit includes a data sorting processor and an evaluation calculation processor, wherein the data sorting processor is used to classify and sort the effect data according to time periods and intersections, and the evaluation calculation processor is used to calculate the efficiency index of the intersection in the corresponding time period;

[0055] The evaluation calculation processor calculates the efficiency index P according to the following formula:

[0056]

[0057] Among them, n 0 is the number of vehicles whose waiting time is 0, n a is the number of all vehicles in the corresponding period, T i is the i-th non-zero waiting time, and Y is the idle time ratio;

[0058] The efficiency index of each intersection in the corresponding time period is sent as a sample to the model optimization unit.

[0059] Embodiment 2.

[0060] This embodiment includes all the contents of the first embodiment, and provides an artificial intelligence-based second-level adaptive traffic signal timing optimization system, including a data acquisition module, an edge computing control module, a comprehensive model analysis module, and a feedback optimization management module;

[0061] The data acquisition module is used to collect traffic data at the intersection, the edge computing control module obtains phase information based on traffic feature data processing and controls the traffic lights, the comprehensive model analysis module is used to obtain traffic effect data of all intersections and optimize the control model, and the feedback optimization management module is used to feed back the optimization results to the edge computing control module;

[0062] Combination Figure 2 The data acquisition module includes an intelligent sensing unit, a traffic feature analysis unit and a traffic effect analysis unit. The intelligent sensing unit is used to sense vehicle information at the intersection. The traffic feature analysis unit is used to analyze the sensed information to obtain feature data. The traffic effect analysis unit is used to analyze the sensed information to obtain effect data.

[0063] Combination Figure 3 The edge computing control module includes a basic phase management unit, a feature data storage unit and a phase adjustment processing unit. The basic phase management unit is used to set the basic phase information and control the signal light based on the phase information. The feature data storage unit is used to store the feature data corresponding to each phase. The phase adjustment processing unit processes the feature data based on the control model to obtain the phase adjustment time and sends it to the basic phase management unit.

[0064] Combination Figure 4 The comprehensive model analysis module includes an effect data storage unit, an intersection evaluation unit and a model optimization unit, wherein the effect data storage unit is used to receive and save effect data of each intersection, the intersection evaluation unit is used to evaluate the traffic status of each intersection, and the model optimization unit optimizes the parameters of the control model based on the evaluation results;

[0065] Combination Figure 5 , the feedback optimization management module includes a network management unit, an execution decision unit and a parameter feedback unit, the network management unit is used to manage the network connection information of the edge computing control module, the execution decision unit is used to determine whether to execute the optimized parameters, and the parameter feedback unit is used to send the optimized parameters to the edge computing control module;

[0066] Traffic lights change their status in a cyclical manner. Each cycle includes a fixed number of phases arranged in a fixed order. The phase refers to the time that the corresponding signal light maintains the green light state.

[0067] The intelligent sensing unit includes a radar sensing processor and an image sensing processor, wherein the radar sensing processor is used to obtain vehicle information at a short distance, and the image sensing processor is used to obtain vehicle information at a long distance;

[0068] Each road at the intersection is divided into a radar area and an image area. The radar area is a section of road near the intersection, which is the monitoring area of ​​the radar perception processor. The image area is a section of road at the intersection, which is the monitoring area of ​​the image perception processor.

[0069] The traffic feature analysis unit includes an occupancy analysis processor, a flow analysis processor, a speed analysis processor and a queue analysis processor, wherein the occupancy analysis processor is used to analyze and obtain the percentage of lane occupancy, the flow analysis processor is used to analyze and obtain the number of vehicles passing through the lane per unit time, the speed analysis processor is used to analyze and obtain the average speed of vehicles passing through the lane, and the queue analysis processor is used to analyze and obtain the length of the vehicle queue on the lane;

[0070] The traffic effect analysis unit includes a time analysis processor and an empty space analysis processor, wherein the time analysis processor is used to analyze and obtain the waiting time of the vehicle, and the empty space analysis processor is used to analyze and obtain the time proportion of the road without traffic flow during the phase time;

[0071] The basic phase management unit includes a phase mapping register, a pointer movement processor and a phase switching processor. The phase mapping register is used to store the mapping relationship between each phase and the signal light and the basic duration of each phase. The pointer movement processor is used to indicate the current phase and the corresponding time point. The phase switching processor is used to determine whether to enter the next phase and send a color change signal to the signal light when entering the next phase.

[0072] The pointer movement processor is represented by a pointer vector (Np, T), wherein Np represents the phase number, T is the duration of the pointer being in the phase, and T increases naturally in seconds through a timer;

[0073] The phase switching processor calculates the switching indication value Vs according to the following formula:

[0074] Vs=T 0 +ΔT-T;

[0075] Among them, T 0 is the basic duration of the current phase, and ΔT is the phase adjustment value;

[0076] When Vs is less than or equal to 0, it indicates switching to the next phase, the pointer vector Np changes to the next phase number, and T is reset and then restarted;

[0077] The phase adjustment processing unit includes a phase data matching processor, a model calculation processor and an adjustment output processor, wherein the phase data matching processor obtains characteristic data of the corresponding lane based on the current phase, the model calculation processor is used to run the control model to obtain a phase adjustment value, and the adjustment output processor is used to output the phase adjustment value to the basic phase management unit;

[0078] The model calculation processor calculates the phase adjustment value according to the following formula:

[0079]

[0080] Among them, X i represents the value of the i-th eigenvalue, α i represents the i-th single coefficient, β i represents the coefficient of the i-th quadratic term, δ ij represents the interaction coefficient between the i-th feature item and the j-th feature item, T' represents the offset time;

[0081] When the phase adjustment value exceeds the upper limit, the adjustment output processor directly outputs the upper limit value;

[0082] The intersection evaluation unit includes a data sorting processor and an evaluation calculation processor, wherein the data sorting processor is used to classify and sort the effect data according to time periods and intersections, and the evaluation calculation processor is used to calculate the efficiency index of the intersection in the corresponding time period;

[0083] The evaluation calculation processor calculates the efficiency index P according to the following formula:

[0084]

[0085] Among them, n 0 is the number of vehicles whose waiting time is 0, n a is the number of all vehicles in the corresponding period, T i is the i-th non-zero waiting time, and Y is the idle time ratio;

[0086] The efficiency index of each intersection in the corresponding time period is sent as a sample to the model optimization unit;

[0087] The model optimization unit includes an efficiency comparison processor and a parameter optimization processor, wherein the efficiency comparison processor is used to compare the efficiency indexes in different samples, and the parameter optimization processor optimizes the parameters in the control model according to the comparison results;

[0088] The efficiency comparison processor selects the samples with the maximum efficiency index and the minimum efficiency index in each time period, obtains the characteristic data of the two samples, and selects three parameter items that need to be optimized according to the differences;

[0089] The three parameter terms selected include a one-way coefficient, a quadratic term coefficient, and an interaction coefficient;

[0090] The efficiency comparison processor first calculates {Xi}, and {X i X j}Then the coefficient of the item with the largest absolute value of the mean difference is selected as the parameter item to be optimized. The three parameter items are denoted as α′, β′ and δ′, and the corresponding three absolute values ​​of the difference are denoted as Δ 1 , Δ 2 and Δ 3 ;

[0091] For example, X 1 The absolute value of the mean difference is 3, X 2 The absolute value of the mean difference is 4, X 3 The absolute value of the mean difference is 2, X 4 The absolute value of the mean difference is 1, then α′ is α 2 , Δ 1 is 4;

[0092] The parameter optimization processor calculates the parameter adjustment value according to the following formula:

[0093]

[0094] Among them, k 1 , k 2 and k 3 is the consistency coefficient, which takes the value of -1 or 1;

[0095] When the efficiency index is i}When the increase and decrease of the mean are consistent, k 1 The value is 1, otherwise -1;

[0096] When the efficiency index and When the increase and decrease of the mean are consistent, k 2 The value is 1, otherwise -1;

[0097] When the efficiency index is consistent with the increase and decrease of the mean value of {XiXj}, k 3 The value is 1, otherwise -1;

[0098] The parameter optimization processor performs similar accumulation processing on the parameter adjustment values ​​of all time periods to obtain the comprehensive adjustment values ​​of all parameter items;

[0099] The execution decision unit includes an adjustment data register and an execution judgment processor, wherein the adjustment data register is used to receive and save the comprehensive adjustment value of all parameter items, and the execution judgment processor is used to judge whether the comprehensive adjustment value needs to be updated to the control model;

[0100] The execution judgment processor calculates the execution index Q according to the following formula:

[0101]

[0102] Among them, Δα i , Δβ i and Δδ ij is the comprehensive adjustment value of the corresponding parameter item;

[0103] When the execution index is greater than the threshold, it is determined that the model parameters need to be updated;

[0104] The i and j appearing in the above text are both ordinal numbers used to indicate serial numbers.

[0105] The contents disclosed above are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated as technology develops.

Claims

1. A second-level adaptive traffic signal timing optimization system based on artificial intelligence, characterized in that: It includes data acquisition module, edge computing control module, comprehensive model analysis module and feedback optimization management module; The data acquisition module is used to collect traffic data at the intersection, the edge computing control module obtains phase information based on traffic feature data processing and controls the traffic lights, the comprehensive model analysis module is used to obtain traffic effect data of all intersections and optimize the control model, and the feedback optimization management module is used to feed back the optimization results to the edge computing control module; The data acquisition module includes an intelligent sensing unit, a traffic feature analysis unit and a traffic effect analysis unit, wherein the intelligent sensing unit is used to sense vehicle information at the intersection, the traffic feature analysis unit is used to analyze the sensed information to obtain feature data, and the traffic effect analysis unit is used to analyze the sensed information to obtain effect data; The edge computing control module includes a basic phase management unit, a feature data storage unit and a phase adjustment processing unit. The basic phase management unit is used to set the phase basic information and control the signal light based on the phase basic information. The feature data storage unit is used to store the feature data corresponding to each phase. The phase adjustment processing unit processes the feature data based on the control model to obtain the phase adjustment time and sends it to the basic phase management unit. The comprehensive model analysis module includes an effect data storage unit, an intersection evaluation unit and a model optimization unit, wherein the effect data storage unit is used to receive and save effect data of each intersection, the intersection evaluation unit is used to evaluate the traffic status of each intersection, and the model optimization unit optimizes the parameters of the control model based on the evaluation result; The traffic feature analysis unit includes an occupancy analysis processor, a flow analysis processor, a speed analysis processor and a queue analysis processor, wherein the occupancy analysis processor is used to analyze and obtain the percentage of lane occupancy, the flow analysis processor is used to analyze and obtain the number of vehicles passing through the lane per unit time, the speed analysis processor is used to analyze and obtain the average speed of vehicles passing through the lane, and the queue analysis processor is used to analyze and obtain the length of the vehicle queue on the lane; The traffic effect analysis unit includes a time analysis processor and an empty space analysis processor, wherein the time analysis processor is used to analyze and obtain the waiting time of the vehicle, and the empty space analysis processor is used to analyze and obtain the time proportion of the road without traffic flow during the phase time; The basic phase management unit includes a phase mapping register, a pointer movement processor and a phase switching processor. The phase mapping register is used to store the mapping relationship between each phase and the signal light and the basic duration of each phase. The pointer movement processor is used to indicate the current phase and the corresponding time point. The phase switching processor is used to determine whether to enter the next phase and send a color change signal to the signal light when entering the next phase. The pointer movement processor is represented by a pointer vector (Np, T), wherein Np represents the phase number, T is the duration of the pointer being in the phase, and T increases naturally in seconds through a timer; The phase switching processor calculates the switching indication value Vs according to the following formula: ； Among them, T0 is the basic duration of the current phase, is the phase adjustment value; When Vs is less than or equal to 0, it indicates switching to the next phase, the pointer vector Np changes to the next phase number, and T is reset and then restarted; The phase adjustment processing unit includes a phase data matching processor, a model calculation processor and an adjustment output processor, wherein the phase data matching processor obtains characteristic data of the corresponding lane based on the current phase, the model calculation processor is used to run the control model to obtain a phase adjustment value, and the adjustment output processor is used to output the phase adjustment value to the basic phase management unit; The model calculation processor calculates the phase adjustment value according to the following formula: ； in, represents the value of the i-th feature item, represents the i-th single coefficient, represents the coefficient of the i-th quadratic term, represents the interaction coefficient between the i-th feature item and the j-th feature item, T' represents the offset time; When the phase adjustment value exceeds the upper limit, the adjustment output processor directly outputs the upper limit value; The intersection evaluation unit includes a data sorting processor and an evaluation calculation processor, wherein the data sorting processor is used to classify and sort the effect data according to time periods and intersections, and the evaluation calculation processor is used to calculate the efficiency index of the intersection in the corresponding time period; The evaluation calculation processor calculates the efficiency index P according to the following formula: ； Among them, n0 is the number of vehicles with a waiting time of 0, n a is the number of all vehicles in the corresponding period, T i is the i-th non-zero waiting time, and Y is the idle time ratio; The efficiency index of each intersection in the corresponding time period is sent as a sample to the model optimization unit; The model optimization unit includes an efficiency comparison processor and a parameter optimization processor, wherein the efficiency comparison processor is used to compare the efficiency indexes in different samples, and the parameter optimization processor optimizes the parameters in the control model according to the comparison results; The efficiency comparison processor selects the samples with the maximum efficiency index and the minimum efficiency index in each time period, obtains the characteristic data of the two samples at the same time, and selects three parameter items that need to be optimized according to the differences; The three parameter terms selected include a single term coefficient, a quadratic term coefficient, and an interaction coefficient; The efficiency comparison processor first calculates the characteristic data , and The mean of the three types of items, and then the coefficient of the item with the largest absolute value of the corresponding mean difference is selected as the parameter item to be optimized. The three parameter items are recorded as , and , the corresponding three mean differences are recorded as , and ; The parameter optimization processor calculates the parameter adjustment value according to the following formula: ； ； ； Among them, k1, k2 and k3 are consistency coefficients, with values ​​of -1 or 1; When the efficiency index and When the increase and decrease of the mean are consistent, k1 takes the value of 1, otherwise it takes the value of -1; When the efficiency index and When the increase and decrease of the mean are consistent, k2 takes the value of 1, otherwise it is -1; When the efficiency index and When the increase and decrease of the mean are consistent, k3 takes the value of 1, otherwise it is -1; The parameter optimization processor performs similar accumulation processing on the parameter adjustment values ​​of all time periods to obtain the comprehensive adjustment values ​​of all parameter items; The feedback optimization management module includes a network management unit, an execution decision unit and a parameter feedback unit. The network management unit is used to manage the network connection information of the edge computing control module. The execution decision unit is used to determine whether to execute the optimized parameters. The parameter feedback unit is used to send the optimized parameters to the edge computing control module. The execution decision unit includes an adjustment data register and an execution judgment processor, wherein the adjustment data register is used to receive and save the comprehensive adjustment value of all parameter items, and the execution judgment processor is used to judge whether the comprehensive adjustment value needs to be updated to the control model; The execution judgment processor calculates the execution index Q according to the following formula: ； in, , and is the comprehensive adjustment value of the corresponding parameter item; When the execution index is greater than the threshold, it is determined that the model parameters need to be updated; Both i and j are ordinal numbers used to represent sequence numbers.

Images