Method for improving the utilization rate of an autonomous driving closed test site

Abstract

The present invention discloses a method for improving the utilization rate of an autonomous driving closed test site, which includes: digitizing the road network of the closed test site into sections, and the status bit of each section is: 1 or 0; when the status bit of a section is 0, testing is allowed. According to the status bits of each section, the occupied space sequence of the test vehicle is determined, and the status bits of all the occupied sections are set from 0 to 1; a time dictionary is constructed to record the section duration that the test vehicle will occupy on each section during its test; the test platform obtains the vehicle position and matches it to the section with the shortest distance; when the test vehicle travels to a section, a countdown is performed on the section duration occupied on this section. When the countdown ends, the status bit of this section is reset from 1 to 0. This method can not only realize the real-time and automatic allocation of each section of the test site, achieve multi-vehicle collaborative testing, but also reduce the vacant time of the site and improve the utilization rate of the test site.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving testing. More specifically, the present invention relates to a method for improving the utilization rate of an autonomous driving closed testing site. Background Art

[0002] In the process of autonomous driving technology research and development, testing in a closed site is an indispensable and crucial link. The reasonable utilization and management of the closed site are essential for the development of the technology. However, there are many drawbacks in the current mainstream management and use methods of closed testing sites, which seriously restrict the testing efficiency and the effective utilization of site resources.

[0003] Currently, the use planning of closed testing sites is relatively extensive. Most sites use a simple area division method, using cone barrels for isolation to define different testing areas. This division method lacks precision and is difficult to improve the space utilization rate of the site. From the perspective of space utilization, when divided into large areas, the actual available space of each part within the site is not considered, resulting in many areas that could accommodate more vehicles for simultaneous testing being left idle. For example, in some large closed testing sites, some edge areas or transitional zones between adjacent testing areas are not effectively utilized because they are simply used as isolation spaces or not fully planned, causing a waste of space resources.

[0004] In terms of site information sharing, communication mainly relies on walkie-talkies or WeChat groups. This method has serious information transmission defects. On the one hand, the information description is not precise enough. The description of the use area and time period is often vague. The description of the use area may be too general to accurately define the specific testing scope, and the time period information may also be exaggerated, inaccurate, or not updated in a timely manner. On the other hand, the communication is unstable. Walkie-talkies are easily affected by factors such as signal interference and equipment aging, resulting in poor or lost information transmission. Although WeChat groups can, to a certain extent, expand the information dissemination range, members may miss important information, resulting in untimely and inaccurate information reception. For example, when a test group notifies other groups through a walkie-talkie that a certain area is occupied, but due to walkie-talkie signal problems, other test groups do not receive this information and mistakenly think the site is idle and enter the test, which is very likely to cause test conflicts.

[0005] The currently widely used WeChat mini-program venue reservation system, although it standardizes the time arrangement of venue use to a certain extent, still has obvious deficiencies. In the time dimension, the method of reserving in fixed time periods cannot flexibly adapt to complex test requirements such as enterprise vehicle debugging code. During the code debugging process of enterprises, the vehicles do not continuously use the venue, and there will be a large number of idle periods. However, these periods cannot be effectively utilized by other test groups, resulting in the venue being vacant during these time periods and wasting precious venue resources. Most critically, these problems have led to serious safety hazards. Due to inaccurate information, poor communication, and unreasonable venue planning, test conflicts occur frequently. In autonomous driving tests, the vehicle speed is relatively high, and the speed of some tests can reach 140 Km / h. Once a vehicle collision occurs, it will cause serious casualties and equipment damage. There is an urgent need for a more efficient, accurate, and safe method for the management and use of closed test venues to improve venue utilization and ensure the smooth progress of tests. Summary of the Invention

[0006] The present invention provides a method for improving the utilization rate of an autonomous driving closed test venue, which can not only realize the real-time and automatic allocation of each section of the test venue, achieve multi-vehicle collaborative testing, but also reduce the idle time of the venue, improve the utilization rate of the test venue, and at the same time improve the safety of the test.

[0007] To achieve these and other advantages of the present invention, the present invention provides a method for improving the utilization rate of an autonomous driving closed test venue, including:

[0008] Digitize the road network of the closed test venue into sections. Each section includes: ID, section longitude and latitude information, direction, and two end nodes. The status bit of each section is: 1 or 0, where 1 indicates busy and 0 indicates idle;

[0009] When the status bit of a section is 0, testing is allowed. According to the status bits of each section, determine the occupied space sequence of the test vehicle, and set the status bits of all occupied sections from 0 to 1. The occupied space sequence is composed of the section IDs of the test vehicle's driving trajectory;

[0010] Construct a time dictionary to record the section duration that the test vehicle will occupy on each section during its test;

[0011] Test the test vehicle according to the occupied space sequence. The test platform obtains the vehicle's position through the GPS device of the test vehicle, and matches it to the section closest to it according to the distance between the vehicle's position and each section; when the test vehicle travels to a section, simultaneously count down the section duration occupied on this section. When the countdown ends, reset the status bit of this section from 1 to 0.

[0012] Preferably, it further includes a safety warning step:

[0013] If, according to the time dictionary, the countdown ends, but it is determined by the GPS device of the test vehicle that the test vehicle has not completed the test of this section, that is, the status bit of this section has been reset to 0, but this section is still occupied, then a first-level alarm is issued and the authority is exceeded, and the section status bit is prohibited from being set to 0;

[0014] If the test vehicle does not travel according to the occupied space sequence of the test scenario but travels to an irrelevant section, an alarm is prompted and the status bit of the irrelevant section is set to 1; if the status bit of the irrelevant section is already 1, a second-level alarm is issued to prompt the test personnel to stop the test.

[0015] Preferably, it further includes a target vehicle, which is a vehicle that conflicts with the test vehicle. The target vehicle is also equipped with a vehicle-end GPS device, and the target vehicle is also set with its own occupied space sequence and occupied section duration;

[0016] Among them, the test vehicle includes multiple test groups, and each test group includes multiple test vehicles; the target vehicle also includes multiple target vehicle groups, and each target vehicle group includes multiple target vehicles.

[0017] Preferably, when constructing the time dictionary to record the occupied section duration of each test vehicle on each section to be tested, the calculation method of the occupied section duration is:

[0018] Section duration = section length / preset speed + buffer time,

[0019] Among them, the preset speed is the expected driving speed of the test vehicle on this section, and the buffer time is used to cover possible delays or abnormal situations during the test process.

[0020] Preferably, to determine the occupied space sequence of the test vehicle according to the status bits of each section, specifically:

[0021] Sort the test tasks according to the task priority. The test tasks with higher task priority are allocated sections first, where the task priority is determined according to the urgency, complexity, and importance of the test tasks;

[0022] For each test task, traverse each section in the occupied space sequence to be used in the test and check the status bits of each section;

[0023] If the status bits of each section are 0, it is allocated to this test task, and the status bits of each section are set to 1;

[0024] If at least one of the road section status bits is 1, calculate the available time of the road section according to the countdown information in the time dictionary, and decide whether to wait or select an alternative road section;

[0025] If multiple test tasks compete for the same road section, conflict resolution is performed according to the task priorities. The high-priority tasks occupy the road section first, and the low-priority tasks choose to wait or re-plan the road section.

[0026] Preferably, the duration of each test vehicle occupying the road section recorded in the time dictionary is a dynamic time dictionary, specifically:

[0027] Through the speed sensor and acceleration sensor installed on the test vehicle, the speed and acceleration of the test vehicle are obtained in real time, and at the same time, the obtained speed and acceleration data are transmitted to the test platform in real time;

[0028] Through the cameras and road condition sensors deployed in the test site, the road condition information is collected in real time;

[0029] Construct a prediction road section duration training model;

[0030] Input the obtained real-time speed, acceleration and road condition information into the prediction road section duration training model to obtain the predicted road section duration;

[0031] If the difference between the predicted road section duration and the preset road section duration in the time dictionary is greater than the threshold, the test platform updates the road section duration in the time dictionary.

[0032] Preferably, the construction of the prediction road section duration training model is specifically:

[0033] Collect the historical data information of the test vehicle, including the driving speed, acceleration, road condition information and actual road section duration on different road sections;

[0034] Extract the road section type features, weather condition features and time period features from the collected historical data information. Among them, the road section types are divided into: curves, straight roads, uphill and downhill; the weather conditions are divided into: sunny, rainy, snowy, cloudy; the time periods are divided into morning, afternoon, evening;

[0035] Use a long short-term memory network model for road section duration prediction.

[0036] Preferably, the determination of the task priority includes: calculating the priority score of each test task through a weighted algorithm, where different weights are assigned to the urgency, complexity and importance; sorting the test tasks according to the priority score, and the tasks with higher priority scores are allocated road sections first.

[0037] Preferably, for the first-level alarm and unauthorized access, the prohibited road section status position is 0. Specifically, after the first-level alarm is triggered, the test platform automatically takes unauthorized access and forcibly maintains the road section status bit as 1. Only after manually confirming the vehicle status and unlocking can the road section be reallocated. The unauthorized operation record is in the test platform. Among them, the first-level alarm includes: prohibiting the road section status position from being 0 and forcibly maintaining the status bit as 1; sending alarm information to the tester, including the current vehicle position and abnormal road section occupancy information. The test platform locks the road section and prohibits other test tasks from allocating this road section.

[0038] The second-level alarm includes: sending emergency alarm information to the tester, including conflict road section information and the current vehicle position; suggesting that the tester immediately stop the test and check the vehicle status; the test platform automatically pauses all relevant test tasks.

[0039] Preferably, it further includes: after the digitization of the road section is completed, the test platform generates a road network topology map for visualizing the connection relationship and status bits of each road section; the road network topology map is updated in real time to reflect the current status of each road section.

[0040] The present invention has at least the following beneficial effects:

[0041] First, through the digital management of road sections, the real-time status (busy or idle) of each road section can be accurately grasped, so that the driving routes of test vehicles can be reasonably planned, reducing the situation of road section idleness or overuse. When determining the occupancy space sequence of test vehicles, it is allocated according to the road section status bit to ensure that each available road section can be fully utilized, avoiding resource waste caused by unclear information in the traditional management method, and greatly improving the utilization rate of the site space. The construction of the time dictionary and the countdown mechanism can accurately control the occupancy duration of vehicles on each road section. Countdown is carried out according to the preset road section duration, and after the countdown ends, the road section status is reset in time, enabling other test vehicles to use this road section as soon as possible, improving the utilization rate of the closed test site in the time dimension and reducing the vacant time of the closed test site. Therefore, through digital and status bit management, the automatic and real-time allocation of road section resources is realized, significantly improving the utilization rate of the test site.

[0042] Second, in the first-level alarm mechanism in the safety warning step, when the countdown ends but the test vehicle has not completed the test, the road section status bit is prohibited from being reset to 0, avoiding other test vehicles or target vehicles from accidentally entering the still occupied road section, effectively preventing the occurrence of vehicle collision accidents. When the test vehicle deviates from the predetermined driving route and enters an irrelevant road section, the tester is prompted and the test is stopped, which can timely detect and stop abnormal driving behaviors, reducing the conflict risk between different test vehicles and ensuring the safe progress of the test process.

[0043] Thirdly, consider the existence of the target vehicle and set a management method similar to that of the test vehicle for it, namely the in-vehicle GPS device, the occupied space sequence, and the occupied road section duration, so that in the complex scenario of simultaneous testing of multiple vehicles, the driving routes and time arrangements between different vehicles can be better coordinated. Group management of the test vehicle and the target vehicle helps to uniformly schedule and monitor vehicles in different groups. Through the centralized management of vehicles in each group, site resources can be better allocated, the test process can be optimized, and the utilization rate of the site and the test efficiency can be further improved. Therefore, through the management of the target vehicle group and multiple test groups, multi-vehicle collaborative testing in complex scenarios is achieved, and the utilization rate of the test site is further improved.

[0044] Other advantages, objectives, and features of the present invention will be partially reflected by the following description and partially understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a schematic flowchart of the method for improving the utilization rate of the closed test site for autonomous driving of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The following further describes the present invention in detail with reference to the accompanying drawings, so that those skilled in the art can implement it according to the description in the specification.

[0047] It should be understood that terms such as "having", "comprising", and "including" used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0048] As Figure 1 shown, an embodiment of the present invention provides a method for improving the utilization rate of a closed test site for autonomous driving, including:

[0049] S10. Digitize the road network of the closed test site into road sections. Each road section includes: ID, road section longitude and latitude information, direction, and two end nodes. The status bit of each road section is: 1 or 0, where 1 indicates busy and 0 indicates idle;

[0050] S20. When the status bit of a road section is 0, allow testing. According to the status bits of each road section, determine the occupied space sequence of the test vehicle, and set the status bits of all occupied road sections from 0 to 1. The occupied space sequence is composed of the road section IDs of the test vehicle's driving trajectory;

[0051] S30. Construct a time dictionary to record the road section duration that the test vehicle will occupy on each road section during its test;

[0052] S40. Test the test vehicle according to the occupied space sequence; S401. The test platform obtains the position of the vehicle based on the GPS device of the test vehicle, and matches it to the nearest road section according to the distance between the vehicle's position and each road section; S402. When the test vehicle travels onto a road section, simultaneously count down the time duration of the road section occupied by the vehicle; S403. When the countdown ends, reset the status bit of this road section from 1 to 0.

[0053] In the above embodiment, the closed test site may include various types of roads, such as straight roads, curves, uphill and downhill road sections, etc. When digitizing the road network of the closed test site for each road section, assign a unique ID to each road section. Exemplarily, adopt the coding method of area code - road section serial number, divide the site into multiple areas such as A, B, C, etc., and set the ID of the first road section in area A as A - 001. Record the longitude and latitude information of the road section in detail, accurate to six decimal places, such as the longitude and latitude of road section A - 001 is (30.123456 N, 118.654321 E). Define the direction of the road section, such as north - south direction, east - west direction, etc., and also the two end nodes, marked as node A1, node B1, etc., respectively, and set the initial status bit of all road sections to 0, indicating idle. Exemplarily, a test vehicle (numbered V001) submits a test application. The test platform obtains the current status bits of each road section and finds that the status bits of road sections A - 001, A - 002, and A - 003 are 0 (idle). According to the preset driving trajectory plan of the test vehicle, determine its occupied space sequence as [A - 001, A - 002, A - 003]. The test platform automatically sets the status bits of these three road sections from 0 to 1, indicating that these road sections have been occupied by test vehicle V001.

[0054] According to the test plan and vehicle performance, it is estimated that the driving speed of the V001 test vehicle on section A-001 is 30 km / h (about 8.33 m / s). The length of this section is 500 meters. Considering possible start-up acceleration, temporary deceleration, etc., a buffer time of 10 seconds is set. By using the formula: section duration = section length / preset speed + buffer time, the occupied duration of section A-001 is 500÷8.33 + 10 ≈ 70 seconds; similarly, the occupied duration of section A-002 is calculated to be 80 seconds, and the occupied duration of section A-003 is 90 seconds. The test platform constructs a time dictionary and records it as: [A-001:70, A-002:80, A-003:90]. For example, when the test starts at 9:00 and the V001 test vehicle enters the site, the test platform obtains the vehicle position information in real time at a frequency of 10 times per second through the high-precision GPS device installed on the test vehicle. When the test vehicle starts and approaches section A-001, the test platform matches it to section A-001 according to the distance between the vehicle position and each section. At this time, the test platform counts down the occupied duration of 70 seconds for section A-001. During the countdown, the test platform continuously monitors the vehicle position to ensure that the vehicle is driving normally on section A-001. When the countdown ends, if the vehicle has left section A-001, the test platform automatically resets the status bit of section A-001 from 1 to 0 again. The vehicle continues to drive to section A-002, and the test platform counts down the 80-second occupied duration of section A-002, and completes the entire test according to this process.

[0055] In one specific implementation, it further includes a safety warning step:

[0056] If, according to the time dictionary, the countdown ends, but it is judged by the GPS device of the test vehicle that the test vehicle has not completed the test of this section, that is, the status bit of this section has been reset to 0 again, but this section is still occupied, a first-level alarm and override are performed, and the status bit of the section is prohibited from being set to 0;

[0057] If the test vehicle does not drive according to the occupied space sequence of the test scenario but drives to an irrelevant section, an alarm prompt is given, and the status bit of the irrelevant section is set to 1; if the status bit of the irrelevant section is already 1, a second-level alarm is issued to prompt the tester to stop the test.

[0058] Specifically, for the first-level alarm and unauthorized access, the prohibited section status position is 0. Specifically: after the first-level alarm is triggered, the test platform automatically takes unauthorized access and forces the section status bit to be 1. Only after manually confirming the vehicle status and unlocking can the section be reallocated. The unauthorized operation records are stored in the test platform. Among them, the first-level alarm includes: prohibiting the section status position from being set to 0 and forcing the status bit to be 1; sending alarm information to the tester, including the current vehicle position and abnormal section occupancy information, locking the section by the test platform, and prohibiting other test tasks from allocating the section; the second-level alarm includes: sending emergency alarm information to the tester, including conflict section information and the current vehicle position; suggesting that the tester immediately stop the test and check the vehicle status; the test platform automatically pauses all related test tasks.

[0059] In the above implementation, in the management of the autonomous driving closed test site, adding the safety warning step can significantly improve safety. The first-level alarm can effectively prevent vehicle collision accidents. When the countdown ends but the test vehicle has not completed the test, the prohibited section status bit is prohibited from being reset to 0, avoiding other test vehicles or target vehicles from accidentally entering the still occupied section. In the case of a relatively high driving speed of the autonomous driving test vehicle, this alarm method can greatly reduce the risk of vehicle collision and ensure the safety of the tester's life and test equipment. For the situation where the test vehicle deviates from the predetermined driving route and enters an irrelevant section, the test platform will give an alarm prompt in time and take different measures according to the status bit of the irrelevant section. If the status bit is 0, it is set to 1. If it is already 1, a second-level alarm is issued and the tester is prompted to stop the test. This processing method can timely detect and stop abnormal driving behaviors, avoid driving route conflicts between different test vehicles, and ensure the orderliness and safety of vehicle driving within the entire test site. Therefore, the safety warning step adds abnormal handling capabilities to the test platform of the entire autonomous driving closed test site, enhancing the stability and reliability of the test platform. When sudden situations such as vehicle test timeout or abnormal driving occur, the test platform can automatically respond and take corresponding measures, reducing the cost of manual intervention and ensuring the smooth progress of the test work.

[0060] For example, there are multiple test groups conducting test work simultaneously, namely test group A and test group B. In test group A, there is a test vehicle numbered A-01, and the occupied space sequence planned for its test task is [R-001, R-002, R-003]; in test group B, there is a test vehicle numbered B-01, and its occupied space sequence is [R-004, R-005, R-006]. When test vehicle A-01 starts the test and enters section R-001, the test platform starts a countdown according to the time dictionary, and it is expected that the occupancy duration of A-01 in section R-001 is 60 seconds. When the countdown ends, according to the normal process, the status bit of section R-001 should be reset from 1 (busy) to 0 (idle). However, at this time, the test platform discovers through the GPS device of A-01 that the vehicle is still within section R-001 and has not completed the test of this section, which results in an anomaly where the section status bit does not match the actual occupancy situation. The test platform immediately triggers a first-level alarm and automatically performs an unauthorized operation: on the one hand, it forcibly maintains the status bit of section R-001 as 1, prohibiting other test vehicles from occupying this section; on the other hand, it sends an alarm message to all testers in test group A, and the message content includes the current position of vehicle A-01 and the abnormal occupancy information of section R-001.

[0061] After test vehicle A-01 completes the test of section R-001, it should enter section R-002 according to the occupied space sequence. However, due to an abnormality in the vehicle's autonomous driving system, A-01 does not drive towards section R-002 but towards section R-007, which is not related to the current test task. The test platform detects the deviation of the driving trajectory of A-01 and immediately issues an alarm prompt. At the same time, it sets the status bit of section R-007 from 0 to 1, indicating that this section is abnormally occupied. If the status bit of section R07 was originally 1, that is, it was already occupied by other test vehicles, the test platform will further issue a second-level alarm. The second-level alarm sends an emergency alarm message to the testers in test group A, including the conflicting section information (section R-007) and the current position of vehicle A-01 (accurate longitude and latitude), and advises the testers to immediately stop the test and check the status of vehicle A-01. In addition, the test platform automatically suspends all relevant test tasks within test group A to avoid more serious safety problems and test conflicts caused by the abnormal driving of A-01. After receiving the second-level alarm, the testers respond quickly, conduct a fault diagnosis and handling of vehicle A-01. After the problem is solved, they re-plan the test task and driving route and resume the test under the condition of ensuring safety.

[0062] In one specific embodiment, it further includes a target vehicle, which is a vehicle that conflicts with the test vehicle. The target vehicle is also equipped with a vehicle-end GPS device, and the target vehicle also has its own occupied space sequence and occupied road section duration. Among them, the test vehicle includes multiple test groups, and each test group includes multiple test vehicles; the target vehicle also includes multiple target vehicle groups, and each target vehicle group includes multiple target vehicles.

[0063] In the above embodiment, in the management of the autonomous driving closed test site, introducing the relevant settings of the target vehicle and grouping and managing the test vehicle and the target vehicle can effectively improve the test safety, optimize the utilization of site resources, and improve the test efficiency. The target vehicle is equipped with a vehicle-end GPS device, an occupied space sequence, and an occupied road section duration, enabling the test platform to monitor the driving trajectories and time arrangements of the target vehicle and the test vehicle in real time. Grouping and managing the test vehicle and the target vehicle enables the test platform to uniformly schedule and monitor vehicles in different groups. Through the centralized management of vehicles within each group, the site road section resources can be more reasonably allocated according to the test requirements and driving plans of each vehicle. The setting of multiple test groups and multiple target vehicle groups can adapt to complex test scenarios. Vehicles in different groups can test in the site orderly under the unified coordination of the test platform, realizing the collaborative operation among multiple vehicles. The test platform can reasonably arrange the test sequence and time interval according to the occupied space sequence and occupied road section duration of each vehicle. For test tasks that may cause conflicts, adjustments are made in advance to avoid vehicles waiting for each other or avoiding each other, reducing delays during the test process.

[0064] In one specific embodiment, when constructing the time dictionary to record the road section duration that each test vehicle will occupy on each road section during its test, the calculation method of the occupied road section duration is as follows:

[0065] Road section duration = Road section length / Preset speed + Buffer time,

[0066] Among them, the preset speed is the expected driving speed of the test vehicle on this road section, and the buffer time is used to cover possible delays or abnormal situations during the test process.

[0067] In the above embodiments, by accurately calculating the occupancy duration of each section, the stay time of the test vehicle in each section is reasonably planned. The countdown method can ensure that the section can be released in time after the vehicle has finished using it, and other vehicles can use it as soon as possible, reducing the overall vacant time of the site and improving the utilization rate of the site in the time dimension. The setting of the buffer time takes into account the uncertainties in the test process, such as vehicle startup acceleration, temporary deceleration, equipment response delay, etc. It reserves time for these potential delay factors, avoiding the actual stay time of the vehicle on the section exceeding the expectation due to unexpected situations, thus ensuring that the test proceeds as planned and reducing the chaos and conflicts caused by insufficient time estimation.

[0068] In one specific embodiment, determining the occupied space sequence of the test vehicle according to the status bits of each section is specifically as follows:

[0069] Sort the test tasks according to the task priority. The test tasks with higher task priority are allocated sections first, where the task priority is determined according to the urgency, complexity, and importance of the test task;

[0070] For each test task, traverse each section in the occupied space sequence to be used in the test and check the status bits of each section;

[0071] If the status bits of each section are 0, allocate it to the test task and set the status bits of each section to 1;

[0072] If at least one of the status bits of each section is 1, calculate the available time of the section according to the countdown information in the time dictionary and decide whether to wait or select an alternative section;

[0073] If multiple test tasks compete for the same section, conflict resolution is performed according to the task priority. The high-priority task occupies the section first, and the low-priority task chooses to wait or re-plan the section.

[0074] Among them, the determination of the task priority includes: calculating the priority score of each test task through a weighted algorithm, where different weights are assigned to the urgency, complexity, and importance respectively; sorting the test tasks according to the priority score, and the tasks with higher priority scores are allocated sections first.

[0075] In the above embodiments, determining the task priority according to the urgency, complexity, and importance of the test task can ensure the execution of critical test tasks first. For tasks involving new technology verification and emergency safety issue testing, due to their high priority, they can obtain road section resources first, avoiding delays in important tests due to insufficient resources and improving the rationality of resource utilization. When allocating road sections, traverse and check the status bits of the road sections. If the road section is idle, it is directly allocated, reducing the waiting time. When the road section is occupied, calculate the available time according to the time dictionary, which helps the test task to reasonably plan the waiting time or select an alternative road section, avoiding time waste caused by blind waiting, making the test process more compact and efficient, and accelerating the overall test progress. When multiple tasks compete for the same road section, allocate according to the priority. High-priority tasks occupy the road section first, and low-priority tasks choose to wait or re-plan, which avoids low-priority tasks occupying road section resources for a long time. Exemplarily, the test task can be a functional test, such as verifying the accuracy, reliability, and stability of various sensors under different environmental conditions; verifying the path planning algorithm and control algorithm of an autonomous vehicle, including simulating different traffic scenarios, such as passing through intersections, driving around roundabouts, overtaking, etc., to check whether the algorithm can accurately analyze the road conditions, plan a reasonable driving path, and stably control the vehicle to drive, etc. The test task can also be a safety test, a compatibility test, etc. For the determination of the priority, for example: there are three test tasks, namely test task A, test task B, and test task C. Test task A: The urgency score is 9 points (out of 10), the complexity score is 8 points, and the importance score is 9 points. Then the priority score = 9×0.5 + 8×0.3 + 9×0.2 = 4.5 + 2.4 + 1.8 = 8.7 points. Test task B: The urgency score is 3 points, the complexity score is 4 points, and the importance score is 5 points. The priority score = 3×0.5 + 4×0.3 + 5×0.2 = 1.5 + 1.2 + 1 = 3.7 points. Test task C: The urgency score is 4 points, the complexity score is 8 points, and the importance score is 8 points. The priority score = 4×0.5 + 8×0.3 + 8×0.2 = 2 + 2.4 + 1.6 = 6 points. Sorted by the priority score: test task A > test task C > test task B. For the determination of the occupied space sequence and road section allocation, for example: Test task A: Its occupied space sequence plan is [A-001, A-003, A-005]. The test platform traverses these road sections and finds that the status bits of A-001, A-002, and A-003 are all 0. So these road sections are allocated to test task A, and the status bits of A-001, A-002, and A-003 are set from 0 to 1. The test vehicle of test task A starts to enter the site for testing. Test task B: The occupied space sequence plan is [A-002, A-004].The test platform checks that the status bits of A-002 and A-004 are both 0. These two sections are assigned to test task B, and the status bits are also set from 0 to 1. The test vehicle for test task B starts the test. Test task C: The occupied space sequence is planned as [A-003, A-004]. When the test platform checks, it is found that the status bit of A-003 is 1 (occupied by test task A), and the status bit of A-004 is also 1 (occupied by test task B). The test platform calculates according to the countdown information of test task A on A-003 in the time dictionary that A-003 will be available in 5 minutes, and test task C decides to wait. At the same time, for A-004, test task C calculates that test task B still needs 10 minutes to complete the test, and the waiting time is too long, so it selects an alternative section and re-plans the occupied space sequence as [A-003, A-006] (assuming the status bit of A-006 is 0). When test task A completes the test on section A-003 and the status bit of A-003 becomes 0, the test vehicle of test task C enters A-003 for testing.

[0076] In one specific embodiment, the duration of each test vehicle occupying a section recorded in the time dictionary is a dynamic time dictionary, specifically:

[0077] S301. Through the speed sensor and acceleration sensor installed on the test vehicle, the speed and acceleration of the test vehicle are obtained in real time, and at the same time, the obtained speed and acceleration data are transmitted to the test platform in real time.

[0078] S302. Through the cameras and road condition sensors deployed in the test site, the road condition information is collected in real time.

[0079] S303. Build a training model for predicting the section duration.

[0080] Among them, the building of the training model for predicting the section duration is specifically:

[0081] A. Collect the historical data information of the test vehicle, including the driving speed, acceleration, road condition information, and actual section duration on different sections;

[0082] B. Extract the section type features, weather condition features, and time period features from the collected historical data information. Among them, the section types are divided into: curved roads, straight roads, uphill and downhill; the weather conditions are divided into: sunny, rainy, snowy, cloudy; the time periods are divided into morning, afternoon, and evening;

[0083] C. Use the long short-term memory network model for section duration prediction.

[0084] S304. Input the obtained real-time speed, acceleration, and road condition information into the training model for predicting the section duration to obtain the predicted section duration;

[0085] S305. If the difference between the predicted road segment duration and the preset road segment duration in the time dictionary is greater than the threshold, the test platform updates the road segment duration in the time dictionary. The threshold can be specifically set according to the actual situation. Exemplarily, the threshold can be 10 seconds.

[0086] In the above embodiment, the vehicle speed, acceleration, and road condition information are obtained in real time, and analyzed using the predicted road segment duration training model. When the predicted duration and the preset duration have a large difference, the time dictionary is updated in a timely manner, enabling testers to be aware of possible changes in the test progress in advance, adjust the test plan in advance, avoid waiting or delays caused by inaccurate time estimation, and improve the overall test efficiency. The dynamic update of the time dictionary enables the test platform to have a more accurate grasp of the occupancy duration of the vehicle on each road segment. The test platform can arrange the routes and times of other test vehicles more reasonably according to the real-time situation, reduce the idle and waste of site resources, and improve the utilization rate of the site in terms of time and space. Considering the influence of various factors such as road segment type, weather conditions, and time periods on the vehicle driving duration, through feature extraction of historical data and prediction using the long short-term memory network model, the road segment driving duration of the vehicle under different conditions can be estimated more accurately, providing a more accurate time reference for the test, and making the test results more reliable and scientific. The dynamic time dictionary can adapt to various changes during the test process, such as vehicle performance fluctuations and temporary road condition changes. Updating the time dictionary in a timely manner to ensure that the system is always managed and scheduled based on the latest information enhances the adaptability and flexibility of the entire test management system.

[0087] Specifically, the test platform has collected a large amount of driving data of past test vehicles on different sections of this site, including different weather conditions, time periods, and road section types. For example, in the morning on a sunny day, a test vehicle was driving on a certain straight section at a speed of 30 m / s, with a stable acceleration of 0, and the actual road section duration was 60 seconds; in the afternoon on a rainy day, another test vehicle had a speed of 15 m / s on a curved section, with fluctuating acceleration, and the actual road section duration was 80 seconds, etc. Feature extraction is performed on the collected historical data. For the road section type, the road sections are divided into curved sections, straight sections, uphill and downhill sections; the weather conditions are divided into sunny, rainy, snowy, and cloudy; the time periods are divided into morning, afternoon, and evening. For example, for the data of driving on a straight section in the morning on a sunny day, the feature of the road section type is a straight section, the feature of the weather condition is sunny, and the feature of the time period is morning. The long short-term memory network model (LSTM) is used to predict the road section duration. The extracted feature data and the corresponding actual road section duration are used as training data to train the LSTM model so that it can learn the relationship between different factors and the road section duration. The current real-time speed (25 m / s), acceleration (0.5 m / s²) of test vehicle V001, and the collected road condition information (there is water accumulation on the road surface) are input into the trained prediction road section duration training model. After calculation, the model predicts that the driving duration of V001 on the current road section is 70 seconds. In the time dictionary, the originally preset driving duration of V001 on this road section was 60 seconds. Since the difference between the predicted road section duration of 71 seconds and the preset duration of 60 seconds is greater than the set threshold (assuming the threshold is 10 seconds), the test platform determines that the road section duration in the time dictionary needs to be updated. Therefore, the test platform updates the duration of V001 on this road section in the time dictionary to 71 seconds.

[0088] In one specific embodiment, it further includes: after the digitization of the road sections is completed, the test platform generates a road network topology map for visually displaying the connection relationships and status bits of each road section; the road network topology map is updated in real time to reflect the current status of each road section.

[0089] In the above embodiment, the road network topology map can display the connection relationships and status bits of each road section in an intuitive graphical form, and the test site management personnel can clearly understand the usage situation of the entire site at a glance. By observing the road network topology map, the management personnel can quickly judge which road sections are in an idle state and which road sections are being used, and then reasonably plan new test tasks and optimize the allocation of site resources.

[0090] The number of devices and the processing scale described here are used to simplify the description of the present invention. The applications, modifications, and variations of the present invention will be obvious to those skilled in the art.

[0091] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those skilled in the art, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the illustrated examples described herein.

Claims

1. A method for improving the utilization rate of a closed test site for autonomous driving, characterized in that: include: The road network of the closed test site is digitized into sections. Each section includes: ID, section longitude and latitude information, direction and two end nodes. The status bit of each section is: 1 or 0, where 1 means busy and 0 means idle; When the status bit of the road section is 0, the test is allowed, and the occupied space sequence of the test vehicle is determined according to the status bits of each road section, and the status bits of all occupied road sections are set from 0 to 1, wherein the occupied space sequence is composed of the road section ID of the driving trajectory of the test vehicle; Construct a time dictionary to record the length of time that the test vehicle will occupy on each road section during its test; The test vehicle is tested according to the occupied space sequence. The test platform obtains the location of the vehicle based on the GPS device of the test vehicle, and matches it to the nearest road section based on the distance between the location of the vehicle and each road section. When the test vehicle moves to a road section, the length of time occupied on the road section is counted down at the same time. When the countdown ends, the status bit of the road section is reset from 1 to 0. The time dictionary is constructed to record the duration of each test vehicle's occupancy on each road section during its test. The duration of the occupancy is calculated as follows: Section duration = section length / preset speed measurement + buffer time, The preset speed measurement is the expected driving speed of the test vehicle on the road section, and the buffer time is used to cover possible delays or abnormal situations that may occur during the test; The duration of each test vehicle occupying the road section recorded in the time dictionary is a dynamic time dictionary, specifically: The speed and acceleration of the test vehicle are acquired in real time by using the speed sensor and acceleration sensor installed on the test vehicle, and the acquired speed and acceleration data are transmitted to the test platform in real time; The cameras and road condition sensors deployed at the test site collect road condition information in real time; Construct a training model for predicting road section duration; Inputting the acquired real-time speed, acceleration and road condition information into the predicted section duration training model to obtain the predicted section duration; If the difference between the predicted section duration and the preset section duration in the time dictionary is greater than a threshold, the test platform updates the section duration in the time dictionary.

2. The method for improving the utilization rate of a closed autonomous driving test site according to claim 1, characterized in that: Also includes safety warning steps: If the countdown ends according to the time dictionary, but the test vehicle's GPS device determines that the test vehicle has not completed the test of the road section, that is, the status bit of the road section has been reset to 0, but the road section is still occupied, a level 1 alarm is issued and the authority is exceeded, prohibiting the road section status bit from being 0; If the test vehicle does not drive according to the occupied space sequence of the test scenario, but drives to an irrelevant section of road, an alarm will be issued and the status bit of the irrelevant section will be set to 1; if the status bit of the irrelevant section is already 1, a secondary alarm will be issued to prompt the tester to stop the test.

3. The method for improving the utilization rate of the closed test site for autonomous driving according to claim 2, characterized in that: Also included is a target vehicle, which is a vehicle that conflicts with the test vehicle, the target vehicle is also provided with a vehicle-side GPS device, and the target vehicle is also provided with its own occupied space sequence and occupied road section duration; The test vehicle includes a plurality of test groups, each of which includes a plurality of test vehicles; the target vehicle also includes a plurality of target vehicle groups, each of which includes a plurality of target vehicles.

4. The method for improving the utilization rate of the closed test site for autonomous driving according to claim 3, characterized in that: The space occupied sequence of the test vehicle is determined according to the status bit of each road section, specifically: The test tasks are sorted according to the task priority. The test tasks with high task priority are assigned to the road sections first. The task priority is determined according to the urgency, complexity and importance of the test task. For each test task, traverse each road segment in the occupied space sequence to be used for its test, and check the status bit of each road segment; If the status bit of each road section is 0, it is assigned to the test task and the status bit of each road section is set to 1; If at least one of the road segments in the status bit of each road segment is 1, the available time of the road segment is calculated according to the countdown information in the time dictionary, and it is decided whether to wait or choose an alternative road segment; If multiple test tasks compete for the same road section, conflicts are resolved based on task priority. High-priority tasks take up the road section first, while low-priority tasks wait or re-plan the road section.

5. The method for improving the utilization rate of the closed test site for autonomous driving according to claim 1, characterized in that: The construction of the predicted road section duration training model is specifically as follows: Collect historical data information of the test vehicle, including driving speed, acceleration, road condition information and actual road duration on different road sections; The collected historical data information is used to extract road section type features, weather condition features, and time period features. The road section types are divided into: curves, straight roads, uphill and downhill; the weather conditions are divided into: sunny, rainy, snowy, and cloudy; and the time periods are divided into morning, afternoon, and evening. The long short-term memory network model is used to predict the road section duration.

6. The method for improving the utilization rate of a closed autonomous driving test site according to claim 4, characterized in that: The determination of the task priority includes: calculating the priority score of each test task through a weighted algorithm, wherein the urgency, complexity and importance are respectively assigned different weights; sorting the test tasks according to the priority scores, and assigning road sections first to tasks with high priority scores.

7. The method for improving the utilization rate of the closed test site for autonomous driving according to claim 2, characterized in that: The first-level alarm is carried out and the authority is exceeded, and the road section status position is prohibited to be 0. Specifically, after the first-level alarm is triggered, the test platform automatically exceeds the authority and forces the road section status bit to be kept at 1. The road section can be reallocated only after the vehicle status is manually confirmed and unlocked. The unauthorized operation is recorded in the test platform; wherein, the first-level alarm includes: prohibiting the road section status position to be 0, forcibly keeping the status bit at 1; sending an alarm message to the tester, including the current position of the vehicle and abnormal road section occupancy information, the test platform locks the road section, and prohibits other test tasks from allocating the road section; The secondary alarm includes: sending emergency alarm information to the tester, including conflicting road section information and the current location of the vehicle; suggesting that the tester stop the test immediately and check the vehicle status; the test platform automatically suspends all related test tasks.

8. The method for improving the utilization rate of a closed autonomous driving test site according to claim 1, characterized in that: Also includes: After the digitization of the road section is completed, the test platform generates a road network topology map to visualize the connection relationship and status of each road section; The road network topology map is updated in real time to reflect the current status of each road section.

Images