Inspection time dynamic calculation method of unmanned aerial vehicle curb parking charge management system

Abstract

The invention discloses a polling time dynamic calculation method of an unmanned aerial vehicle curb parking charge management system. The polling time dynamic calculation method comprises the following steps: step 1, fitting a probability density function of vehicle arrival and parking duration in a curb parking area based on historical parking data; 2, calculating the parking duration of the vehicle based on the real-time vehicle information polled by the unmanned aerial vehicle; 3, calculating the change probability of all parking space states in the inspection time interval based on the Bayesian theorem; step 4, establishing a parking income maximization model based on an on-road charging rule; and 5, solving the parking income maximization model based on a dual simulated annealing algorithm, and obtaining the take-off time of the dynamic inspection of the unmanned aerial vehicle. The method can be used in cooperation with an existing on-road parking charge management system based on the autonomous inspection unmanned aerial vehicle, the inspection time of the unmanned aerial vehicle is effectively and dynamically calculated in real time, the economical efficiency and scientificity of the system are improved, and the system is helped to be applied in an actual commercialized scene and effectively implemented on the ground.

Description

Technical Field

[0001] The present invention belongs to the technical field of traffic engineering, and in particular relates to a method for dynamically calculating the inspection time of a drone in an on-street parking fee management system. Background Art

[0002] With the continuous growth of private car ownership in my country, the parking problem in urban residential areas has become more and more prominent, and the parking resource tension has become an important problem that needs to be solved in urban development. On-street parking areas have become one of the key strategies to alleviate parking problems due to their short construction period and low cost. As for the current status of on-street parking management, it mainly relies on management modes such as traffic police patrols, parking meter charges, and high-position video charges. These modes are not entirely suitable for the management of on-street parking in residential areas. Some researchers have begun to carry out relevant research on the application of automatic patrol drones in the field of parking management, and proposed an automatic patrol parking management system based on drones. With its advantage of being able to quickly and efficiently cover a large area, it shows broad application prospects. Compared with existing methods, this system has the advantages of fast, efficient, and low carbon. However, the frequency of drone inspections is closely related to the economy and scientificity of the system. At present, very little research has been carried out, which has led to the system facing difficulties in actual commercial application scenarios and is difficult to implement effectively. Therefore, it is particularly important to determine the take-off time of drones in real time, dynamically and effectively in the on-street parking management system, and to realize the dynamic calculation of the inspection time based on the drone on-street parking fee management system. Summary of the invention

[0003] In order to solve the problems existing in the prior art, the present invention proposes a method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system, comprising the following steps:

[0004] Step 1: Fit the probability density function of vehicle arrival and parking time in the on-street parking area based on historical parking data;

[0005] Step 2: Calculate the parking time of the vehicle based on the real-time vehicle information inspected by the drone;

[0006] Step 3: Calculate the probability of change of the status of all parking spaces within the inspection time interval based on Bayes' theorem;

[0007] Step 4: Establish a parking revenue maximization model based on on-street charging rules;

[0008] Step 5: Solve the parking revenue maximization model based on the double simulated annealing algorithm to obtain the take-off time of the UAV dynamic inspection.

[0009] Furthermore, in step one, the specific steps include:

[0010] Collect vehicle parking information data in the on-street parking area, including: license plate information, vehicle entry time, departure time, latitude and longitude of the parking location, and vehicle photos;

[0011] Analyze the quality of vehicle parking information data and perform data preprocessing to remove erroneous and unreasonable information;

[0012] The preprocessed vehicle parking information data is statistically analyzed to obtain the parking time pattern and vehicle arrival pattern of the vehicles in the area, and the probability density function of the vehicle parking time and the probability density function of the vehicle arrival are calculated.

[0013] Furthermore, in step 2, the real-time vehicle information detected by the drone inspection is used to calculate the vehicle entry time by detecting the same vehicle in two different states of "present" or "absent" through two consecutive drone inspections. and departure time The interval is the time between the two, so as to calculate the vehicle's stopped time Time.

[0014] Further, into time Departure time The calculation method of the stopped time is:

[0015] When a car T i The vehicle status is "absent" at the time of inspection, but T i+1 If the vehicle status at the time of inspection is "in", the expected value of the vehicle's entry time is calculated as:

[0016]

[0017] Where, T i is the time of the drone’s i-th inspection;

[0018] When a car T i The vehicle status is "in" at the time of inspection, but T i+1 If the vehicle status at the time of inspection is "absent", the expected value of the departure time of the departing vehicle is calculated as:

[0019]

[0020] The parking time of the vehicle is:

[0021]

[0022] For a certain drone inspection, if a vehicle is found whose status has not changed in the system, that is, the vehicle has been parked here, then the parking time of the vehicle is:

[0023]

[0024] Bayesian methods are well suited for small sample sizes. By using prior knowledge and prior distributions, Bayesian methods can provide more stable and reliable predictions when the amount of data is insufficient.

[0025] Furthermore, in step 3, after the drone completes a certain inspection, it will be in the drone inspection time interval [T min ,T max ] to make the next flight, and based on the vehicle’s parking time and the probability density function of the parking time, use the Bayesian method to calculate the probability value of the vehicle leaving the parking space in the next period of time:

[0026]

[0027] Where P leave Time ij is the probability that the jth vehicle stops for Time in the i-th inspection, x ij For the i-th inspection, the j-th vehicle stops for another x hours;

[0028] According to the probability density function of vehicle arrival, the probability value P of a vehicle arriving at any time in the future can be calculated. come .

[0029] Furthermore, in step 4, based on the parking management policy of the area, the parking fee rules are clarified, and the amount of on-street parking fees in the area is assumed to be a function of the parking duration, which is recorded as:

[0030] fee=Gt

[0031] Where Gt is the charging function for parking duration;

[0032] Combined with the probability values ​​of parking space status changes of all vehicles within the drone inspection time interval obtained in step 3, a parking revenue maximization model is established:

[0033]

[0034] Where j is the vehicle number of the i-th inspection, N is the number of parking lots, and λ is the vehicle arrival probability cost coefficient, yuan.

[0035] Due to the parking lot capacity limit, assuming the parking lot capacity is M, then:

[0036] N≤M

[0037] There are restrictions on the time intervals between drone inspections:

[0038] T min ≤x ij <T max .

[0039] Furthermore, in step 5, the parking revenue maximization model is solved based on the dual simulated annealing algorithm. Based on the classic simulated annealing algorithm, this algorithm introduces a second "simulated annealing" process to increase the depth and breadth of the search, thereby finding the optimal solution in a larger range. It can adapt to different types of optimization problems, including complex nonlinear and multi-peak problems, can reduce the impact of local optimality, and has a strong advantage in finding the global optimal solution.

[0040] The update of the algorithm solution needs to rely on the Metropolis sampling criterion. The change of the objective function from state i to state j can be described as:

[0041]

[0042] Beneficial effects:

[0043] The present invention mainly proposes a method for dynamically calculating the inspection time of unmanned aerial vehicles based on on-street parking management. The real-time information of on-street parked vehicles is collected by unmanned aerial vehicles, the entry and exit times of vehicles are extracted, the parking time is calculated, and then the charging situation corresponding to the parking time is calculated based on the probability value of all vehicles parking within the next unmanned aerial vehicle inspection time interval. A method for dynamically calculating the inspection time of unmanned aerial vehicles based on on-street parking management is constructed.

[0044] The dynamic method of drone inspection time proposed in the present invention can effectively, in real time and dynamically calculate the inspection time of drones based on the on-street parking fee management system of drones, avoid the waste of resources caused by excessive take-off frequency of drones in the system, reduce the excessive error of vehicle charging caused by too low frequency, and contribute to the commercialization of the existing drone on-street parking fee management system. In addition, compared with the fixed-duration drone flight method (drone take-off time interval of 30 minutes), it can significantly improve the charging efficiency, and the charging accuracy rate reaches more than 95%, which solves the problem of lack of theoretical basis for determining the inspection cycle in the field of on-street parking automated inspection, and provides specific methods and ideas for future on-street low-carbon and energy-saving parking fees. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flow chart of the calculation of the dynamic inspection time of the UAV proposed by the present invention;

[0046] Figure 2 It is a probability density function diagram of the on-street parking time according to a specific embodiment of the present invention;

[0047] Figure 3 It is a probability density function diagram of on-road vehicle arrival according to a specific implementation manner of the present invention. DETAILED DESCRIPTION

[0048] Please refer to the attached Figure 1, the specific implementation method of the present invention is described in detail below.

[0049] The implementation target is a parking lot on a bicycle road with eight lanes in both directions in Harbin.

[0050] Step 1: Use manual recording methods to observe and count on-street parking lots, record parking data, including license plate information, vehicle entry time, departure time, latitude and longitude of the parking location, vehicle photos, etc., and record historical parking data multiple times.

[0051] The recorded data is then uploaded to the computer, the data quality is visualized and analyzed, and data preprocessing is performed. Through PyCharm software programming, vehicle data information with only entry time, only departure time, no license plate information, vehicle photos, erroneous information, and unreasonable information are eliminated.

[0052] Then, based on the processed data, the Origin software is used to fit and analyze the data to obtain the vehicle arrival distribution and parking time distribution of the on-street parking area, such as Figure 2 and Figure 3 shown.

[0053] Among them, the probability density function of vehicle arrival is:

[0054] Among them, the probability density function of vehicle arrival is:

[0055]

[0056] The probability density function of parking time is:

[0057]

[0058] Step 2: The drone will start flying at the beginning of the daily charging time interval (i.e. 8:00) and complete the last flight before the end (i.e. 20:00). Any flight of the drone after the first flight of the day has a "previous flight". During the first flight, the vehicle in the parking space will be identified and recorded, and the time when the vehicle is photographed is recorded as the entry time; at the end of the "previous flight" of the i-th flight, the photographed and identified vehicle information will be compared with the information database, and the newly appeared vehicle information will be added to the database of vehicles entering, the disappeared vehicle information will be added to the database of vehicles leaving, and the vehicles that still exist will be added to the database for parking information updates.

[0059] The real-time vehicle information detected by the drone inspection can be used to determine the time interval between the vehicle entering and leaving the area. This interval is the time between two different existence states of the same vehicle detected by the drone during two consecutive inspections. For example, for a certain vehicle, T iThe vehicle status is "in" at the time of inspection, but T i+1 The vehicle status at the time of inspection is "absent", otherwise it also meets the requirements. Then the entry time of the vehicle is calculated based on the probability density function of vehicle arrival:

[0060]

[0061] The departure time of the leaving vehicle can also be calculated based on the probability density function of the vehicle parking time:

[0062]

[0063] The parking time of the vehicle is:

[0064]

[0065] For vehicles whose status in the system has not changed, that is, the vehicle has been parked here, the entry time is calculated as above. The parking time of such vehicles is:

[0066]

[0067] Step 3: After the drone completes a flight, the updated vehicle parking data can be obtained from step 2, mainly including the license plate number, vehicle entry time, and parking time. The drone will min , T max ] for the next flight. min is the minimum time interval for the drone to take off, which is set to 6 minutes in this example; T max is the maximum time interval for drone takeoff, which is set to 30 minutes in this paper.

[0068] Based on the probability density function of the vehicle's parking time and the vehicle's parking time, the Bayesian method is used to calculate the probability value of the parked vehicle leaving in the next period of time. The Bayesian method is very suitable for small sample situations. By using prior knowledge and prior distribution, the Bayesian method can provide more stable and reliable predictions when the amount of data is insufficient.

[0069] According to the probability density function of the parking time of the vehicle, the probability value of a vehicle still parked at any time in the future can be calculated:

[0070]

[0071] And according to the probability density function of vehicle arrival, the probability value P of a vehicle arriving at any time in the future can be calculated. come .

[0072] Step 4: Refer to the on-street parking management regulations issued to determine the charging rules for the studied road section. The parking time of this road section is divided into free parking time and paid parking time. The free parking time is from 20:00 to 8:00 the next day, and the paid parking time is from 8:00 to 20:00. The charging rules of this road section are as follows: free parking within 30 minutes; 5 yuan for 30 minutes to 2 hours; if it is more than 2 hours, 1 yuan will be charged for each hour exceeding 2 hours; this charging rule is expressed by the function method, that is:

[0073]

[0074] The DJI Mavic 3 drone flies and records in the on-street parking fee area. Through the above steps, we can obtain the probability value of the parked vehicle continuing to park at any time in the future after the drone completes a certain inspection. At this time, the on-street fee management rules are determined, and then the parking revenue maximization model can be constructed:

[0075]

[0076] Step 5: After the model is established, the above parking revenue maximization model is solved based on the dual simulated annealing algorithm. Based on the classic simulated annealing algorithm, this algorithm introduces a second "simulated annealing" process to increase the depth and breadth of the search, thereby finding the optimal solution in a larger range. It can adapt to different types of optimization problems, including complex nonlinear and multi-peak problems, can reduce the impact of local optimality, and has a strong advantage in finding the global optimal solution.

[0077] The update of the algorithm solution needs to rely on the Metropolis sampling criterion. The change of the objective function from state i to state j can be described as:

[0078]

[0079] The process of intelligent calculation of the double simulated annealing algorithm is realized by using Python program. The initial temperature is 5230℃, the restart temperature ratio is 0.00002, the access parameter is 2.62, the acceptance probability parameter is -5, and the initial solution is randomly generated. The charging accuracy of this method is as high as 95%, and the optimization effect is very good, which shows that this model has high optimization performance.

Claims

1. A method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system, comprising the following steps: Step 1: Fit the probability density function of vehicle arrival and parking time in the on-street parking area based on historical parking data; Step 2: Calculate the parking time of the vehicle based on the real-time vehicle information inspected by the drone; Step 3: Calculate the probability of change of the status of all parking spaces within the inspection time interval based on Bayes' theorem; Step 4: Establish a parking revenue maximization model based on on-street charging rules; Step 5: Solve the parking revenue maximization model based on the double simulated annealing algorithm to obtain the take-off time of the UAV dynamic inspection.

2. The method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system according to claim 1 is characterized in that: In step one, the specific steps include: Collect vehicle parking information data in the on-street parking area, including: license plate information, vehicle entry time, departure time, latitude and longitude of the parking location, and vehicle photos; Analyze the quality of vehicle parking information data and perform data preprocessing to remove erroneous and unreasonable information; The preprocessed vehicle parking information data is statistically analyzed to obtain the parking time pattern and vehicle arrival pattern of the vehicles in the area, and the probability density function of the vehicle parking time and the probability density function of the vehicle arrival are calculated.

3. The method for dynamically calculating the inspection time of the unmanned aerial vehicle on-street parking fee management system according to claim 1 is characterized in that: In step 2, the real-time vehicle information detected by the drone inspection is used to calculate the vehicle entry time by detecting the same vehicle in two different states of "present" or "absent" through two consecutive drone inspections. and departure time The interval is the time between the two, so as to calculate the vehicle's stopped time Time.

4. The method for dynamically calculating the inspection time of the unmanned aerial vehicle on-street parking fee management system according to claim 3 is characterized in that: Enter time Departure time The calculation method of the stopped time is: When a car T i The vehicle status is "absent" at the time of inspection, but T i+1 If the vehicle status at the time of inspection is "in", the expected value of the vehicle's entry time is calculated as: Where, T i is the time of the drone’s i-th inspection; When a car T i The vehicle status is "in" at the time of inspection, but T i+1 If the vehicle status at the time of inspection is "absent", the expected value of the departure time of the departing vehicle is calculated as: The parking time of the vehicle is: For a certain drone inspection, if a vehicle is found whose status has not changed in the system, that is, the vehicle has been parked here, then the parking time of the vehicle is:

5. The method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system according to claim 1 is characterized in that: In step 3, after the drone completes a certain inspection, it will [T min ,T max ] to make the next flight, and based on the vehicle’s parking time and the probability density function of the parking time, use the Bayesian method to calculate the probability value of the vehicle leaving the parking space in the next period of time: Where P leave Time ij is the probability that the jth vehicle stops for Time in the i-th inspection, x ij For the i-th inspection, the j-th vehicle stops for another x hours; According to the probability density function of vehicle arrival, the probability value P of a vehicle arriving at any time in the future can be calculated. come .

6. The method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system according to claim 1 is characterized in that: In step 4, based on the parking management policy of the area, the parking fee rules are clarified, and the amount of on-street parking fees in the area is assumed to be a function of the parking duration, which is recorded as: fee=G t Where Gt is the charging function for parking duration; Combined with the probability values ​​of parking space status changes of all vehicles within the drone inspection time interval obtained in step 3, a parking revenue maximization model is established: Where j is the vehicle number of the i-th inspection, N is the number of parking lots, and λ is the vehicle arrival probability cost coefficient, yuan. Due to the parking lot capacity limit, assuming the parking lot capacity is M, then: N≤M There are restrictions on the time intervals between drone inspections: T min ≤x ij <T max 。 7. The method for dynamically calculating the inspection time of an unmanned aerial vehicle on-street parking fee management system according to claim 1 is characterized in that: In step five, the parking revenue maximization model is solved based on the double simulated annealing algorithm. On the basis of the classic simulated annealing algorithm, a second "simulated annealing" process is introduced to increase the depth and breadth of the search.

8. The method for dynamically calculating the inspection time of the unmanned aerial vehicle on-street parking fee management system according to claim 7 is characterized in that: The update of the algorithm solution of the double simulated annealing algorithm relies on the Metropolis sampling criterion. The change of the objective function from state i to state j is described as:

Images