Ground unmanned vehicle autonomous following performance test and evaluation method

Abstract

The application provides a ground unmanned vehicle autonomous following performance test and evaluation method, which can realize ground unmanned vehicle autonomous following performance test and multi-dimensional index quantitative evaluation. The application installs a satellite signal receiver and a non-contact automobile tester on the vehicle to be tested, respectively completes straight road following, curved road following and slope mutation following autonomous following scene test, and collects test data on the road. On this basis, a five-dimensional evaluation index system including risk analysis and key functional index is proposed, and the five evaluation indexes are dimensionless processed to obtain the autonomous following performance comprehensive evaluation result. The application realizes ground unmanned vehicle autonomous following performance test and multi-dimensional index quantitative evaluation based on unstructured scene risk analysis, and can provide effective reference for development and improvement of autonomous following performance of the vehicle to be tested.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle testing technology, specifically to a method for testing and evaluating the autonomous following performance of ground-based unmanned vehicles. Background Technology

[0002] With the rise of artificial intelligence technology and unmanned mobile platforms, unmanned ground vehicles have received continuous attention and development in the military field. Unmanned squad support vehicles can transport supplies and wounded soldiers, and follow manned vehicles and soldiers, effectively reducing logistical support work. As a transitional phase for fully autonomous unmanned vehicles, remote-controlled squad support vehicles are currently a research hotspot. When conducting long-distance rapid transport and off-road maneuvers, autonomous following is an essential function to reduce the workload of the operator. Effective following on straight roads, curves, and slopes, and avoiding collisions during emergency braking of the guide vehicle are two key technical challenges. To verify and improve the robustness and adaptability of the designed algorithms, current research focuses on improving and verifying the performance of subsystems such as environmental perception algorithms, real-time map construction and navigation positioning, trajectory prediction, and trajectory tracking. However, systematic analysis and quantitative evaluation of the overall autonomous following performance of the vehicle still lack relevant theories and methods. Summary of the Invention

[0003] In view of this, the present invention proposes a method for testing and evaluating the autonomous following performance of ground unmanned vehicles, which can realize the autonomous following performance testing and multi-dimensional quantitative evaluation of ground unmanned vehicles.

[0004] To achieve the above objectives, the technical solution of the present invention is as follows:

[0005] A method for testing and evaluating the autonomous following performance of ground-based unmanned vehicles includes the following steps:

[0006] Step 1: Select an unstructured road containing straight sections, curves, and slopes as the test site;

[0007] Step 2: Conduct autonomous following scenario tests for straight-line following, curve following, and gradient change following, and collect test data on the road. Specifically, the test data includes the speed, mileage, braking trigger signal, latitude and longitude, and time of the test vehicle and the accompanying vehicle.

[0008] Step 3: Based on physics and logic, quantify the risks of losing track, collisions, and the severity of harm after an accident, which are the key indicators for autonomous following risk assessment; secondly, incorporate functional indicators for autonomous following of unmanned vehicles, including average following speed and following stability, to construct a complete evaluation indicator system.

[0009] Step 4: Based on the evaluation index system, calculate the collision risk, following loss risk, collision hazard level, average autonomous following speed, and following stability indexes based on the test data.

[0010] Step 5: Dimensionless processing is performed on the obtained collision risk, following loss risk, collision hazard level, average autonomous following speed, and following stability indicators to obtain a comprehensive evaluation result of autonomous following performance.

[0011] In step 2, during the straight-line autonomous following test scenario, from t0 to t... i Real-time tracking distance L within a time period sf for:

[0012]

[0013] Among them, V sf0 With V sr0 Let a be the initial speeds of the car in front and the car behind when traveling on a straight road. f With a r The real-time accelerations of the front and rear vehicles, L, are respectively. sf0 This represents the initial following distance between the two vehicles;

[0014] In a curve-following test scenario, assuming both the lead vehicle and the autonomous vehicle are traveling in the center of the road, the shortest distance L from the center of the curve to the following vehicles' line of sight is measured. C for:

[0015]

[0016] R1 to R3 represent curves with different radii, L cf1 This represents the following distance when cornering;

[0017] In the test scenario of following a sudden change in slope, when the real-time visual angle is greater than the limit recognition angle of the perception system, there is a perception blind spot, and the autonomous following function fails.

[0018] In step 3, the risk of losing track of a target is assessed using the following indicators:

[0019]

[0020] In the formula, L f It is the real-time following distance, L max L is the maximum sensing distance for an autonomous vehicle to follow another vehicle. cfmax It is the maximum following distance when following curves with different curvatures; K s With K c θ1 and θ2 represent the maximum limit following distance safety margin coefficients when following on straight roads and curves, respectively; θ1 and θ2 represent the upper and lower limit recognition angles of the autonomous vehicle's perception system; α1 and α2 represent the real-time visual angles between the guide vehicle and the autonomous vehicle when going uphill and downhill, respectively.

[0021] Regarding collision risk, the risk of a collision during autonomous following is measured by the proportion of time during which the time to ... and time to which the time to which the time to which the time and time to a certain critical warning value (TTC0) are less than or equal to the total driving distance (S).

[0022] RC = (count(t|TTC) <TTC0)) / S

[0023] Regarding the severity of a collision, the energy level at the time of the collision is:

[0024] EC = m r ·(V sr -V sf +(a r -a f )·TTC) 2 / 2

[0025] Where, m r For the weight of the following vehicle, V sf With V sr These are the real-time speeds of the vehicle in front and the vehicle behind, respectively. f With a r These are the real-time accelerations of the vehicle in front and the vehicle behind, respectively.

[0026] Among them, the average speed of autonomous following is determined by the total following distance S and the total time t. all The conversion yields the following:

[0027]

[0028] Regarding follow-up stability, the autonomous follow-up stability index is as follows:

[0029]

[0030] Where N is the total number of failures in the follow function, N0 is the maximum number of failures allowed when executing the follow task, and N1, N2, and N3 are the number of times the follower is lost, the number of collisions, and the number of times human intervention occurs, respectively.

[0031] In step 5, each indicator is processed to be dimensionless, that is, the collision risk value, the following loss risk value, the collision hazard degree, and the following stability are divided by the maximum value calculated by each indicator, and the following speed is divided by the expected maximum speed to obtain a performance network diagram, which is used to evaluate the autonomous following performance of the ground unmanned vehicle.

[0032] In step 1, the test site requires that the total length of the straight road the vehicle passes through be no less than 1000m, the number of curves be no less than 4, and that the vehicle complete at least one uphill or downhill maneuver. The road slope angle is determined based on the maximum gradeability of the test vehicle.

[0033] In step 2, test data is collected by installing a satellite signal receiver and a non-contact vehicle tester on the vehicle under test. Specifically, a satellite signal receiver and a non-contact vehicle tester are installed on the guide vehicle and the unmanned ground vehicle under test, respectively. If the guide vehicle is a manned vehicle, an autonomous driving robot is installed.

[0034] Beneficial effects:

[0035] 1. This invention successfully completed autonomous following scenario tests in straight-line following, curve following, and gradient change following, and collected test data on these routes. Based on this, a five-dimensional evaluation index system incorporating risk analysis and key functional indicators is proposed. The five evaluation indicators are dimensionless to obtain a comprehensive evaluation result of autonomous following performance. This invention, based on unstructured scenario risk analysis, realizes the testing and multi-dimensional quantitative evaluation of autonomous following performance of ground-based unmanned vehicles, providing a valuable reference for the development and improvement of the autonomous following performance of the tested vehicles.

[0036] 2. In this invention, the five-dimensional evaluation index system based on risk analysis includes the risk of losing track, the risk of collision, the degree of collision damage, the average speed of autonomous following, and the following stability index, which covers all aspects of risk and can achieve a comprehensive evaluation.

[0037] 3. In this invention, each indicator is dimensionless, that is, the collision risk value, the following loss risk value, the collision hazard degree, and the following stability are divided by the maximum value calculated by each indicator, and the following speed is divided by the expected maximum speed, to obtain a performance network diagram, which can achieve accurate evaluation.

[0038] 4. In this invention, an unstructured road containing straight sections, curves, and slopes is selected as the test site. The design of the unmanned ground vehicle autonomous following unstructured test scenario mainly includes: straight-line following, curve following, and slope change following test scenario design to achieve comprehensive testing and evaluation.

[0039] 5. This invention employs a distributed experimental data acquisition and processing method. A satellite signal receiver and a non-contact automotive testing instrument are installed on the vehicle under test. Information such as speed, mileage, braking trigger signals, latitude, longitude, and time of the tested and accompanying vehicles is collected and processed using a scene kinematics modeling-based data processing method. The data acquisition is comprehensive, and the processing methods are scenario-appropriate, achieving accurate testing and evaluation. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the method for testing and evaluating the autonomous following performance of unmanned ground vehicles according to the present invention.

[0041] Figure 2 This is a schematic diagram of the test equipment and its installation in an embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of autonomous vehicle following on a straight road in an embodiment of the present invention.

[0043] Figure 4 This is a schematic diagram of autonomous following on a curve in an embodiment of the present invention.

[0044] Figure 5 This is a schematic diagram of autonomous vehicle following on uphill and downhill in an embodiment of the present invention.

[0045] Figure 6 This is a schematic diagram of the five-dimensional evaluation index system in an embodiment of the present invention.

[0046] Figure 7 This is a performance mesh diagram in an embodiment of the present invention. Detailed Implementation

[0047] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0048] This invention provides a method for testing and evaluating the autonomous following performance of unmanned ground vehicles. Through experimental scenario setup, experimental data acquisition and processing, and the construction of an evaluation index system, it achieves the testing and evaluation of the autonomous following performance of unmanned ground vehicles, applicable to unstructured scenarios. Specifically, it is based on autonomous following scenario testing on unstructured roads, including straight-line following, curve following, and following following sudden slope changes. Then, based on risk and key function analysis, a five-dimensional evaluation index system for autonomous following performance is established. Finally, the indexes are dimensionless to achieve a comprehensive evaluation of the autonomous following performance of unmanned ground vehicles. The method flow of this invention is as follows: Figure 1 As shown, the specific steps include the following:

[0049] Step 1: Select an unstructured road containing straight sections, curves, and slopes as the test site;

[0050] For the test site, the total length of the straight road that the vehicle passes through should be no less than 1000m, the number of curves should be no less than 4, and the vehicle should complete at least one uphill or downhill maneuver. The road slope angle should be determined based on the maximum gradeability of the test vehicle, and generally should not be less than 50% of the maximum gradeability. The vehicle should be able to complete the autonomous following test tasks for straight roads, curves, and slopes.

[0051] Step 2: Complete autonomous following scenario tests for straight-line following, curve following, and gradient change following respectively, and collect test data on the road.

[0052] This embodiment collects test data by installing a satellite signal receiver and a non-contact vehicle testing instrument on the vehicle under test. Specifically, a satellite signal receiver and a non-contact vehicle testing instrument are installed on both the guide vehicle and the unmanned ground vehicle under test. If the guide vehicle is a manned vehicle, an autonomous driving robot should be installed if conditions permit. The test equipment and installation in this embodiment are as follows: Figure 2 As shown.

[0053] The test data specifically includes information such as the speed, mileage, braking trigger signal, latitude and longitude, and time of the test and companion vehicles. The data collection process for each test scenario is as follows:

[0054] Specifically, in the straight-line following test scenario, autonomous following of another vehicle on a straight road is as follows: Figure 3 As shown, where V sf0 With V sr0 V represents the initial speed of the vehicle in front and the vehicle behind when traveling on a straight road. sf With V sr These are the real-time speeds of the vehicle in front and the vehicle behind, respectively. f With a r The real-time accelerations of the front and rear vehicles, respectively, are L. sf0 With L sd These represent the initial following distance and the real-time following distance between the two vehicles, respectively. Real-time following distance L sf for:

[0055]

[0056] In the cornering follow test scenario, the autonomous vehicle follows another vehicle in a corner. Figure 4 As shown, for ease of analysis, it is initially assumed that both the guide vehicle and the autonomous vehicle are traveling in the exact center of the road. R1 to R3 represent curves of different radii, and W... R W is the width of the road. V For the width of the driverless car, L C L is the shortest distance from the center of the curve to the line of sight of the vehicles in front and behind. cf1 With L cf2 V represents different following distances when cornering. cf With V cr These are the real-time speeds of the vehicle in front and the vehicle behind when driving on a curve. By the Pythagorean theorem, we can obtain:

[0057]

[0058] After analyzing the horizontal direction, the next step is to build an autonomous vehicle-following model in the vertical direction. This includes testing autonomous vehicle-following on inclines and declines, such as in a gradient change following test scenario. Figure 5 As shown, θ1 and θ2 represent the upper and lower limit recognition angles of the autonomous vehicle's perception system, while α1 and α2 are the real-time visual angles between the guide vehicle and the autonomous vehicle when going uphill and downhill, respectively. When the real-time visual angle is greater than the limit recognition angle of the perception system, a perception blind spot exists, which can easily lead to target loss and failure of the autonomous following function.

[0059] Step 3: This invention constructs an evaluation index system for evaluation. Based on physics and logic, it quantifies the risks of losing track, collisions, and the severity of harm after a dangerous situation occurs; these are the key indicators for assessing the risk of autonomous following. Secondly, by incorporating functional indicators for autonomous vehicle following (including average following speed and following stability), a complete evaluation index system can be constructed, such as... Figure 6 As shown, the risk of losing track, the risk of collision, the degree of collision damage, the average speed of autonomous following, and the following stability are specifically as follows:

[0060] (1) Risk of losing track

[0061] Under the premise of autonomous following technology with real-time target locking, in order to ensure that the vehicle does not lose track of the other vehicle on curves, the view of the following vehicle from the front cannot be obstructed by roadside obstacles. Therefore, the following constraints exist:

[0062] L C ≥R3-W R / 2

[0063] Right now:

[0064]

[0065] Secondly, when the road slope changes abruptly, such as Figure 4 As shown, the real-time visual angle between the two vehicles should be within the limit angle recognition area of ​​the perception system. In summary, the risk index for losing tracking is constructed as follows:

[0066]

[0067] In the formula, L f It is the real-time following distance, L max L is the maximum sensing distance for an autonomous vehicle to follow another vehicle. cfmax K represents the maximum following distance when following curves with different curvatures. s With K c These are the safety margin coefficients for the maximum following distance when following on a straight road and when following on a curve, respectively.

[0068] (2) Collision risk

[0069] First, calculate the time remaining until the two vehicles collide. Assume that the driving states of the vehicles remain unchanged in the following moments, i.e., the accelerations of the vehicles are 'a'. f With a r Let L be a constant value. sf =0, therefore:

[0070] (a f ·TTC 2 / 2+V sf0 ·TTC)-(a r ·TTC2 / 2+V sr0 ·TTC)+L sf0 =0

[0071] Throughout the autonomous following process, the risk of a collision is measured by the proportion of time during which the time TTC is lower than a certain critical alarm value TTC0 (this parameter is determined by the development or design requirements of the unmanned vehicle) relative to the total driving distance S.

[0072] RC = (count(t|TTC) <TTC0)) / S

[0073] (3) Collision damage level

[0074] Analyzing the relative velocities of the front and rear vehicles at the time of the collision, the energy level at the time of the collision can be obtained as follows:

[0075] EC = m r ·(V sr -V sf +(a r -a f )·TTC) 2 / 2

[0076] Where, m r For the quality of the following vehicle.

[0077] (4) Autonomous following average speed

[0078] This can be achieved by following the total mileage S and total time t. all The conversion is as follows:

[0079]

[0080] (5) Follow stability

[0081] The number of times a vehicle loses its following position per unit distance, the number of collisions, and the number of times human intervention is required directly reflects the stability of the following process. Secondly, the fluctuation range of vehicle speed during the following process also reflects the quality of following stability. This can be achieved by analyzing the relative speed difference between the vehicles in front and behind and the speed fluctuation of the vehicle in front, using the ratio of their root mean square deviations to measure the stability of autonomous following. The autonomous following stability index is constructed as follows:

[0082]

[0083] S represents the total mileage of the following route, N represents the total number of times the following function failed throughout the entire process, N0 represents the maximum number of failures allowed when performing the following task, N1, N2, and N3 represent the number of times the vehicle lost track of the vehicle, the number of collisions, and the number of times human intervention occurred, respectively; FS represents the following stability, with a larger value indicating greater stability.

[0084] Step 4: Based on the evaluation index system, calculate the collision risk, following loss risk, collision hazard level, average autonomous following speed, and following stability indexes based on the test data.

[0085] Step 5 involves dimensionlessly processing the obtained collision risk, follow-loss risk, collision severity, average autonomous following speed, and following stability indices to obtain a comprehensive evaluation result of autonomous following performance. Specifically, dimensionless processing of each index—dividing the collision risk value, follow-loss risk value, collision severity, and following stability by their calculated maximum values, and dividing the following speed by the expected maximum speed—results in the following: Figure 7 The performance mesh diagram shown is used to evaluate the autonomous following performance of unmanned ground vehicles.

[0086] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for testing and evaluating the autonomous following performance of ground-based unmanned vehicles, characterized in that, Includes the following steps: Step 1: Select an unstructured road containing straight sections, curves, and slopes as the test site; Step 2: Conduct autonomous following scenario tests for straight-line following, curve following, and gradient change following, and collect test data on the road. Specifically, the test data includes the speed, mileage, braking trigger signal, latitude and longitude, and time of the test vehicle and the accompanying vehicle. Step 3: Based on physics and logic, quantify the risks of losing track, collisions, and the severity of harm after an accident, which are the key indicators for autonomous following risk assessment; secondly, incorporate functional indicators for autonomous following of unmanned vehicles, including average following speed and following stability, to construct a complete evaluation indicator system. Step 4: Based on the evaluation index system, calculate the collision risk, following loss risk, collision hazard level, average autonomous following speed, and following stability indexes based on the test data. Step 5: Dimensionless processing is performed on the obtained collision risk, follow-loss risk, collision hazard level, average autonomous following speed, and following stability indexes to obtain a comprehensive evaluation result of autonomous following performance. In step 2, during the straight-line following test scenario, the vehicle autonomously follows another vehicle on a straight road. arrive Real-time tracking distance within a time period for: in, and These are the initial speeds of the vehicle in front and the vehicle behind when traveling on a straight road. and These are the real-time accelerations of the vehicle in front and the vehicle behind, respectively. This represents the initial following distance between the two vehicles; In a curve-following test scenario, assuming both the lead vehicle and the autonomous vehicle are traveling in the center of the road, the shortest distance from the center of the curve to the following vehicles' line of sight is determined. for: to Representing curves with different radii. This represents the following distance when cornering; In the test scenario of following a sudden change in slope, when the real-time visual angle is greater than the limit recognition angle of the perception system, there is a perception blind spot, and the autonomous following function fails. In step 3, the risk of losing track of a target is assessed using the following indicators: In the formula, It is the real-time following distance. It is the maximum sensing distance when an unmanned vehicle is autonomously following another vehicle. It is the maximum following distance when following curves with different curvatures; and These are the safety margin coefficients for the maximum following distance when following on a straight road and when following on a curve, respectively. , This represents the upper and lower limit recognition angles of the autonomous vehicle's perception system. , These are the real-time visual angles between the guide vehicle and the driverless vehicle when going uphill and downhill, respectively. Regarding collision risk, during the entire autonomous following process, the TTC (Tracking Time Control) is lower than a certain critical alarm value. The duration of the journey accounts for a percentage of the total mileage followed. The proportion is used to measure the risk of a collision during autonomous following: Regarding the severity of a collision, the energy level at the time of the collision is: in, For the quality of the following vehicle, and These are the real-time speeds of the vehicle in front and the vehicle behind, respectively. and These are the real-time accelerations of the vehicle in front and the vehicle behind, respectively.

2. The method as described in claim 1, characterized in that, For autonomous following average speed, the total following mileage is used. Total time The conversion yields the following: 。 3. The method as described in claim 1, characterized in that, Regarding follow-up stability, the autonomous follow-up stability index is as follows: in, To track the total number of function failures throughout the entire process, This is the maximum number of failures allowed when performing a follow task. , , These are respectively: losing track of the target, collision occurring, and number of times personnel intervened.

4. The method as described in claim 1 or 3, characterized in that, In step 5, each indicator is processed to be dimensionless, that is, the collision risk value, the following loss risk value, the collision hazard degree, and the following stability are divided by the maximum value calculated by each indicator, and the following speed is divided by the expected maximum speed, to obtain a performance network diagram, which is used to evaluate the autonomous following performance of the ground unmanned vehicle.

5. The method as described in claim 1 or 3, characterized in that, In step 1, the test site requires that the total length of the straight road the vehicle passes through is not less than 1000m, the number of curves is not less than 4, and the vehicle completes at least one uphill or downhill maneuver. The road slope angle is determined based on the maximum gradeability of the test vehicle.

6. The method as described in claim 1 or 3, characterized in that, In step 2, test data is collected by installing satellite signal receivers and non-contact vehicle testing instruments on the vehicle under test. Specifically, satellite signal receivers and non-contact vehicle testing instruments are installed on the guide vehicle and the unmanned ground vehicle under test, respectively. If the guide vehicle is a manned vehicle, an autonomous driving robot is installed.

Images