A cloud-edge collaborative based vehicle automatic driving function real vehicle test system

Abstract

The application discloses a kind of vehicle automatic driving function real vehicle test system based on cloud edge cooperation, it is related to intelligent network connected car function safety test technical field, including multi-functional obstacle chassis car and cloud end cooperative control platform;Chassis car integration drive-by-wire chassis ontology, car roof can be rolled ramp structure, quick change interface, positioning module, edge computing unit, communication module and state self-checking module, test method covers scene definition, task generation verification, issue execution, data processing and report generation whole link;The application realizes the repeated use of props by the rollable ramp, realizes the quick change of multiple scenes in combination with quick change interface, realizes the parallel automated testing of multiple vehicles relying on cloud edge cooperation architecture, built-in network outage autonomous safety mechanism, can greatly reduce test cost, improve test efficiency and scene coverage, ensure test operation safety.

Description

Technical Field

[0001] This invention relates to the field of functional safety testing technology for intelligent connected vehicles, and specifically to a real-vehicle testing system for autonomous driving functions based on cloud-edge collaboration. Background Technology

[0002] With the rapid development of the intelligent connected vehicle industry, automatic emergency braking (AEB), automatic emergency steering (AES), and adaptive cruise control (ACC) have become mandatory test items for mass-produced vehicles. Safety verification of these functions requires simulating dynamic obstacle scenarios in real-world road environments, such as sudden braking by the vehicle in front, pedestrians crossing, and vehicles cutting in, to verify the reliability of the vehicle's perception, decision-making, and execution systems. Currently, the industry mainly uses four types of testing schemes: foam target vehicles, remote-controlled chassis vehicles, towed target platforms, and fixed obstacles. However, all these schemes have significant technical shortcomings and cannot simultaneously balance testing cost, operational efficiency, scenario coverage, and operational safety.

[0003] The first category consists of foam vehicles and inflatable target vehicles. These rely on lightweight foam or inflatable structures to create simulated targets, using soft materials to avoid collision damage to the vehicle under test. Their core drawbacks are: the target vehicle is essentially a disposable consumable; after a single collision, it undergoes plastic deformation or structural breakage and cannot be reused. This results in high material costs per test, and large-scale testing requires extremely high material input. Furthermore, manual setup, cleaning, and replacement of the target vehicle are required before and after testing, leading to long downtime and extremely low operational efficiency. Moreover, these targets can only be statically placed or towed at low speeds, unable to simulate active dynamic behaviors such as vehicle acceleration, deceleration, lane changing, and cutting in; they only cover the most basic static braking test scenarios, resulting in very limited coverage of operating conditions.

[0004] The second category is the remote-controlled chassis vehicle solution, which uses a chassis equipped with a drive-by-wire steering and braking system. Test personnel visually control the vehicle using a handheld remote control. Some high-end products have a foam shell added to the top to simulate the appearance of different vehicle models. The shortcomings of this solution are: control accuracy relies entirely on the operator's experience and reaction speed; consistency in trajectory, speed, and braking timing is poor; test data repeatability is insufficient, failing to meet regulatory accuracy requirements; each device requires one operator, making multi-vehicle collaborative testing extremely difficult, and it's hard to simulate complex traffic flow scenarios such as multiple vehicles merging or platooning; the entire system lacks cloud scheduling and automated execution capabilities, with the entire testing process relying on manual control, making it impossible to achieve a closed-loop operation from scenario generation to report output, resulting in limited overall efficiency improvement; furthermore, ordinary remote-controlled chassis generally lack crush-resistant structures, making chassis components easily damaged after a collision, leading to high repair costs.

[0005] The third type is the towed target platform solution, in which the simulated target is fixed to a flatbed truck and driven by a tow vehicle via steel cables or tow bars. Before a collision, the platform is released by a release mechanism to allow it to glide freely. The main disadvantages of this solution are: the platform's trajectory is completely constrained by the tow vehicle, and it can only travel along a fixed path, making it impossible to plan arbitrary paths or change lanes flexibly, resulting in extremely poor scenario adaptability; after release, the platform is in a powerless and free state, and its speed and direction are uncontrollable, making it impossible to simulate active behaviors such as emergency braking or acceleration away from the vehicle in front, resulting in very limited test conditions; the tow vehicle itself occupies test site space, is prone to interference with the vehicle under test, and poses safety hazards; in addition, the layout of the tow cables and release mechanism is cumbersome, and switching between different scenarios is time-consuming, resulting in low test efficiency; the release mechanism is prone to wear and jamming after long-term use, resulting in a high failure rate and high daily maintenance costs.

[0006] The fourth category is the fixed obstacle scheme, which involves setting up fixed obstacles such as inflatable walls and foam columns within the test area. This scheme can only verify the vehicle's ability to recognize and brake on static obstacles, and cannot simulate the movement behavior of dynamic traffic participants. It cannot cover core test scenarios such as dynamic target AEB and vehicle cutting in, and can only serve as an auxiliary testing method, with very limited application scenarios.

[0007] In summary, existing testing technologies generally suffer from common pain points such as short lifespan of testing equipment, high consumable costs, low efficiency in scene switching, poor adaptability, reliance on manual operation, insufficient automation and data consistency, weak parallel testing capabilities, low site utilization, and a lack of automated closed-loop testing processes and data iteration capabilities. Currently, the industry lacks an engineering testing system that integrates repeated testing, rapid equipment replacement, cloud-edge collaborative scheduling, and fully automated testing, thus failing to meet the large-scale, low-cost, and high-coverage testing needs of the intelligent connected vehicle industry. Summary of the Invention

[0008] The purpose of this invention is to provide a vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration to solve the problems of high testing cost, low replacement efficiency, insufficient automation, weak parallel testing capability, and poor safety of extreme testing in the prior art.

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] According to the first aspect of this disclosure, a real-vehicle testing system for autonomous driving functions based on cloud-edge collaboration is proposed, including a multi-functional obstacle chassis vehicle and a cloud-based collaborative control platform that is wirelessly connected to the multi-functional obstacle chassis vehicle.

[0011] The multi-functional obstacle chassis vehicle includes a chassis body, a rollable ramp structure on the roof, a quick-change interface, a positioning module, an edge computing unit, a communication module, and a status self-checking module.

[0012] The chassis body integrates a steer-by-wire unit, a brake-by-wire unit, and a drive-by-wire unit, which are used to perform steering, braking, and driving actions.

[0013] The rollable ramp structure on the roof is fixedly installed on the top of the chassis body and is used to guide the vehicle being tested to roll over it.

[0014] The quick-change interface is arranged around the upper part of the chassis body and is used to connect and lock different types of simulation prop shells.

[0015] The positioning module is fixedly installed at the geometric center of the chassis body and is used to collect the position and attitude data of the chassis.

[0016] The edge computing unit is fixedly installed in the waterproof electrical control box of the chassis body. The drive-by-wire steering unit, drive-by-wire braking unit, drive-by-wire drive unit, positioning module, communication module and status self-test module are electrically connected to the edge computing unit and interact with each other.

[0017] The communication module is used to establish a wireless communication link between the edge computing unit and the cloud-based collaborative control platform.

[0018] The status self-check module is used to collect chassis operating status data and transmit it to the edge computing unit;

[0019] The cloud-based collaborative control platform is used for test scenario generation, multi-vehicle task scheduling, test data reception and analysis, and test report output.

[0020] Furthermore, the slope of the vehicle roof ramp structure has an angle of 15° with the horizontal plane, rubber buffer pads are provided around the bottom of the ramp, and anti-slip textures are processed on the surface of the ramp; the vehicle roof ramp structure is fixedly connected to the mounting bracket on the upper part of the chassis body by bolt groups.

[0021] Furthermore, the quick-change interface includes at least four mechanical positioning interfaces and a corresponding number of quick clamps; the mechanical positioning interface adopts a structure in which a conical positioning pin and a conical positioning hole are engaged, and the conical positioning pin is vertically fixed to the upper surface of the chassis body; the quick clamps are set next to the corresponding mechanical positioning interface and are used to lock the outer shell of the simulation prop.

[0022] Furthermore, the steer-by-wire unit consists of a steering motor, a steering angle sensor, and a controller, and receives steering commands via a bus; the brake-by-wire unit is an EMB electromechanical braking module, equipped with a pressure sensor and a position sensor; the drive-by-wire unit uses a hub motor or a central motor, and controls vehicle speed via torque commands; the self-testing module integrates an inertial measurement unit, a battery management system, and a steer-by-wire actuator status monitoring circuit, which are used to detect chassis attitude, monitor power battery parameters, and collect feedback signals from the steer-by-wire actuator, respectively.

[0023] Furthermore, the positioning module is a combined inertial navigation system that combines RTK-GNSS and IMU;

[0024] The edge computing unit uses an automotive-grade SoC chip and integrates an MPC trajectory tracking controller, an emergency strategy local adjudication module, and a network outage autonomous security module.

[0025] The communication module is a 5G module, equipped with dual SIM card slots to support dual-link hot standby for operators;

[0026] The cloud-based collaborative control platform has a built-in multi-vehicle scheduling module and an AI scene generation engine, which are used for multi-vehicle trajectory and time sequence allocation, and physical parameterization generation of natural language scenes, respectively.

[0027] According to the second aspect of this disclosure, an automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration is also proposed, using the real vehicle testing system of the first aspect, including the following steps:

[0028] S1. Enter the test scenario description or select a preset scenario template in the cloud-based collaborative control platform to generate the corresponding scenario parameters;

[0029] S2. The cloud-based collaborative control platform generates a sequence of motion trajectory points and behavior triggering conditions based on scene parameters, forms standardized test scripts, and verifies the legality of the parameters.

[0030] S3, the cloud-based collaborative control platform distributes test tasks to at least one multi-functional obstacle chassis vehicle via a wireless link;

[0031] S4, the edge computing unit of the multi-functional obstacle chassis vehicle analyzes the test task, drives the chassis body to perform test actions according to the predetermined trajectory and trigger conditions, and uploads vehicle status data in real time;

[0032] S5. During the test, the chassis control process data and the test vehicle operation data are synchronously uploaded to the cloud collaborative control platform.

[0033] The S6 cloud-based collaborative control platform performs timestamp alignment, data cleaning, and indicator calculation on the received data, and automatically generates a test report containing judgment conclusions.

[0034] Furthermore, step S1 also includes inputting a scene description in natural language form, which is automatically parsed by the AI ​​scene generation engine and outputs the corresponding physical test parameters.

[0035] Furthermore, in step S4, the execution of the edge computing unit also includes the following sub-steps:

[0036] S4.1 Analyze the test task instructions and extract the trajectory point sequence and behavior trigger conditions;

[0037] S4.2 Start the MPC trajectory tracking controller to calculate the steering angle, braking force and drive torque control quantities at a frequency of 100Hz;

[0038] S4.3, Drive the steer-by-wire unit, brake-by-wire unit and drive-by-wire unit to execute corresponding control commands;

[0039] S4.4 monitors the distance or TTC value of the vehicle under test in real time, and executes the corresponding preset test behavior when the preset trigger conditions are met;

[0040] S4.5. Upload vehicle status data to the cloud-based collaborative control platform at a set frequency.

[0041] Furthermore, after step S6, the test data is injected into the training dataset of the AI ​​scene generation engine to optimize the scene mutation algorithm and generate edge test cases, thus forming a test closed loop.

[0042] Furthermore, during the test, when the edge computing unit detects that the communication heartbeat loss with the cloud collaborative control platform exceeds the set threshold, it automatically performs deceleration, turns on the warning lights, and pulls over to the side of the road. It also stores the operating data during the network outage locally and automatically retransmits it to the cloud after the network is restored.

[0043] Compared with existing technologies, this invention provides a real-vehicle testing system for autonomous driving functions based on cloud-edge collaboration. By adopting a cloud-edge collaborative mode of global scheduling in the cloud and real-time control at the edge, the cloud can schedule more than 50 chassis vehicles in parallel, incorporating an AI scene generation engine and multi-vehicle collaborative algorithms. The edge device is equipped with an MPC trajectory tracking controller and an autonomous safety module for network outages, along with dual SIM card 5G communication hot standby. This balances the efficiency of large-scale parallel testing with site safety under network outage conditions, achieving full automation from scene generation, task issuance, real-vehicle execution to data feedback iteration and automatic report output, shortening the testing cycle and improving testing efficiency. Furthermore, the 15° rollable ramp design on the roof can withstand more than 500 rolling cycles, effectively reducing the cost per test. The accompanying symmetrical conical positioning quick-change interface allows switching between various simulation props such as cars, pedestrians, and bicycles within 5 minutes, reducing equipment procurement costs and enabling multi-purpose use of a single vehicle. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0045] Figure 1 A top-view perspective view of the overall structure of the vehicle autonomous driving function real-vehicle test system based on cloud-edge collaboration provided in an embodiment of the present invention;

[0046] Figure 2 Figure a is a top view of the chassis body provided in the embodiment of the present invention, and Figure b is a bottom view of the chassis body.

[0047] Figure 3 This is a physical diagram of a vehicle roof ramp structure provided in an embodiment of the present invention; the red circle indicates the opening portion of the vehicle roof ramp structure.

[0048] Figure 4 A flowchart of an automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration, provided in an embodiment of the present invention.

[0049] Explanation of reference numerals in the attached figures:

[0050] 1. Chassis body; 2. Roof ramp structure; 3. Quick-change interface; 4. Positioning module; 5. Edge computing unit; 6. Communication module; 7. Status self-test module. Detailed Implementation

[0051] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0052] Example 1:

[0053] This embodiment presents a vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration, including a multi-functional obstacle chassis vehicle and a cloud-based collaborative control platform connected to the multi-functional obstacle chassis vehicle via 5G wireless communication. The multi-functional obstacle chassis vehicle is a drive-by-wire chassis device that can be repeatedly rolled over, serving as a dynamic obstacle target to perform test actions; the cloud-based collaborative control platform is deployed on a cloud server to realize the scheduling and data management of global test tasks.

[0054] The multi-functional obstacle chassis vehicle mainly includes the chassis body 1, the roof crushable ramp structure 2, the quick-change interface 3, the positioning module 4, the edge computing unit 5, the communication module 6, and the status self-checking module 7.

[0055] The chassis body 1 adopts a fully domestically produced drive-by-wire chassis with four support points at the bottom. The overall structure is low-profile to reduce the lifting height during rolling. The chassis body 1 integrates a drive-by-wire steering unit, a drive-by-wire braking unit, and a drive-by-wire drive unit.

[0056] The steer-by-wire unit consists of a steering motor, a steering angle sensor, and a controller. It receives steering commands from the edge computing unit 5 via a CAN bus to achieve precise steering angle control. The brake-by-wire unit is an EMB electromechanical braking module with built-in pressure and position sensors, enabling precise linear adjustment of braking force. The drive-by-wire unit uses a hub motor or central motor solution to control vehicle speed via torque commands, meeting the acceleration and deceleration requirements of the test scenario.

[0057] The roof-mounted ramp structure 2 is installed on top of the chassis body 1. It features a gentle slope with a 15° angle to the horizontal plane. This angle ensures the tested vehicle can smoothly climb and roll over the ramp while avoiding excessive impact loads. Rubber buffer pads are installed around the bottom of the ramp to cushion the impact force during rolling. The ramp surface is textured with a 0.5mm deep diamond-shaped anti-slip pattern to prevent slippage during vehicle rolling. The roof-mounted ramp structure 2 is securely connected to the mounting bracket on the upper part of the chassis body 1 via bolts, ensuring a stable connection and allowing for disassembly and maintenance as needed. When the tested vehicle fails to brake and collides with the chassis, the front wheels drive up the ramp to the top of the chassis and smoothly pass over it, preventing damage to the chassis structure and achieving a repeatable rolling effect.

[0058] The quick-change interfaces 3 are symmetrically arranged around the upper part of the chassis body 1, with at least 4 sets, to ensure the stability of the simulation shell installation. Each set of quick-change interfaces 3 includes a mechanical positioning interface and quick clamps.

[0059] The mechanical positioning interface adopts a structure in which a tapered positioning pin and a tapered positioning hole cooperate. The tapered positioning pin is vertically fixed to the upper surface of the chassis body 1, and the positioning hole is correspondingly set at the bottom of the simulation prop shell, which can achieve quick alignment and positioning during installation. The quick clamp is set next to the corresponding mechanical positioning interface. After positioning, the quick clamp can be pulled to lock the simulation shell without the need for tools. When changing, only the clamp needs to be unlocked, the shell replaced, and alignment and locking required. The switching between different types of simulation props can be completed within 5 minutes.

[0060] The positioning module 4 employs a combined inertial navigation system (INS) combining RTK-GNSS and IMU, and is fixedly installed at the geometric center of the chassis to ensure the accuracy of positioning data. The positioning module 4 is connected to the edge computing unit 5 via a shielded cable, and can output high-precision latitude and longitude positions, vehicle speed, heading angle, attitude angle, and other data, providing a position reference for trajectory tracking control and cloud monitoring.

[0061] Edge computing unit 5 is fixedly installed in a waterproof electrical control box in the middle of the chassis body 1, with IP67 protection capability, suitable for outdoor testing environments. Edge computing unit 5 uses a domestically produced automotive-grade SoC chip with a computing power of no less than 30 TOPS, 8GB of onboard RAM and 64GB of eMMC storage, and runs a domestically produced real-time operating system to meet high real-time control requirements.

[0062] The interface connection relationship of the edge computing unit 5 is as follows: it connects to each wire-controlled actuator through a serial bus to issue control commands and collect actuator feedback; it connects to the communication module 6 through Ethernet to realize data interaction with the cloud; it connects to the positioning module 4 through a serial port to receive positioning data; and it connects to the status self-test module 7 through a GPIO interface to collect device status data.

[0063] The edge computing unit 5 is pre-installed with an MPC trajectory tracking controller, an emergency policy local decision-making module, communication middleware, and a network outage autonomous security module. The MPC trajectory tracking controller is used to achieve high-precision trajectory tracking control; the emergency policy local decision-making module is used for rapid local decision-making in emergency scenarios; and the network outage autonomous security module is used for security policy execution when communication is interrupted.

[0064] The communication module 6 adopts a 5G industrial-grade module and connects to the edge computing unit 5 via a PCIe interface. The 5G module is equipped with dual SIM card slots, supporting dual-link hot standby for different operators. It automatically switches to the backup link when the primary link signal is abnormal, ensuring communication continuity. The communication module 6 is responsible for uploading vehicle status data from the edge computing unit 5 to the cloud and receiving task instructions and control signals from the cloud.

[0065] The self-test module 7 integrates an inertial measurement unit, a battery management system, and a drive-by-wire actuator status monitoring circuit. The inertial measurement unit detects the chassis's attitude angle and impact acceleration in real time, identifying abnormal states such as collisions and rollovers. The battery management system monitors the voltage, current, temperature, and SOC status of the power battery in real time, ensuring power supply safety. The drive-by-wire actuator status monitoring circuit collects feedback signals from angle sensors, pressure sensors, and motor encoders to determine whether each drive-by-wire actuator is functioning correctly. All status data is collected via a serial bus to the edge computing unit 5, enabling real-time self-testing and fault reporting of the entire device's status.

[0066] The cloud-based collaborative control platform is deployed on a cloud server cluster and includes a multi-vehicle scheduling module, an AI scene generation engine, a task orchestration module, a data management module, and a report generation module. The multi-vehicle scheduling module employs a distributed collaborative algorithm to assign trajectories and timing sequences to multiple chassis vehicles, preventing collisions between them. The AI ​​scene generation engine can automatically generate physically parameterized test scripts based on natural language descriptions. The task orchestration module is responsible for generating standardized test cases and performing parameter verification. The data management module is responsible for storing, cleaning, and analyzing test data. The report generation module automatically outputs standardized test reports.

[0067] Working principle:

[0068] The working principle of this testing system is divided into four parts: normal testing mode, cloud-edge collaborative control mechanism, crushable function principle, and quick equipment change principle, as detailed below:

[0069] Normal test mode:

[0070] The cloud-based collaborative control platform sends test task commands to the multi-functional obstacle course chassis vehicle via a 5G link. These commands include the motion trajectory (latitude and longitude sequence, velocity sequence, and acceleration sequence) and behavioral trigger conditions (such as "execute emergency braking when 10m away from the test vehicle"). Upon receiving the commands, the edge computing unit 5 invokes the MPC trajectory tracking controller to calculate the required steering angle, braking force, and drive torque at a 100Hz control frequency. This data is then sent to each drive-by-wire actuator via a serial bus, driving the chassis vehicle to move precisely along the predetermined trajectory. Simultaneously, the positioning module 4 collects position and attitude data in real time and uploads it to the cloud for testing process monitoring and subsequent data analysis.

[0071] Cloud-edge collaborative control mechanism:

[0072] This system adopts a cloud-edge collaborative architecture of "global scheduling in the cloud and real-time control at the edge". The cloud is responsible for global task management: the multi-vehicle scheduling module realizes trajectory allocation and timing coordination of multiple chassis vehicles, supporting large-scale parallel testing; and the AI ​​scene generation engine realizes rapid parameterization of test scenarios. The edge is responsible for high real-time local control: based on the high-level task instructions issued by the cloud, it autonomously completes high-precision trajectory tracking, emergency avoidance, and fault handling.

[0073] When communication is normal, the edge device reports the vehicle status (location, speed, fault codes, etc.) to the cloud at a frequency of 50Hz, and at the same time receives real-time control commands (such as speed fine-tuning, global emergency stop) issued by the cloud. When communication is interrupted for more than 500ms (configurable), the edge device automatically activates the network disconnection autonomous safety mode, which can complete the safe parking operation without relying on cloud commands.

[0074] The principle behind the rollable function:

[0075] When the automatic emergency braking function of the tested vehicle fails and it collides head-on with the obstacle chassis, the front wheels of the tested vehicle will first contact the ramp surface of the roof that can be rolled over (slope 2), gradually rising along the 15° slope and passing over the top of the chassis. The gentle slope design avoids rigid impacts, effectively protecting the internal structure and components of the chassis. After the collision, the chassis only needs to undergo a visual inspection and functional self-test. Once confirmed to be normal, it can continue to be tested. It can withstand more than 500 rolling cycles, significantly extending its service life.

[0076] The principle of quick costume change

[0077] When changing the outer shell of the simulation prop, simply manually open all the quick-release clamps 3-2 to lift and remove the original simulation shell. Align the tapered positioning hole at the bottom of the new simulation shell with the tapered positioning pin on the chassis, place it downwards until the positioning pin is fully inserted, and then pull the quick-release clamps to lock it in place. This completes the entire replacement operation. The whole process requires no special tools and can be operated by a single person. It can quickly switch between various simulation targets such as cars, SUVs, pedestrians, and bicycles, achieving multi-purpose use of one vehicle.

[0078] Example 2:

[0079] This embodiment proposes an automated testing method based on the testing system of Embodiment 1, including the following steps:

[0080] Step 1: The user inputs a test scenario description in the graphical interface of the cloud-based collaborative control platform, such as "The vehicle in front is traveling at a constant speed of 20 km / h, and the vehicle under test approaches at 30 km / h, with an initial distance of 30m between the two vehicles." Alternatively, the user can directly select a preset scenario template built into the system to generate the corresponding scenario parameters. This step supports AI scenario generation: the user inputs a scenario description in natural language, such as "On an icy road, a vehicle in the oncoming lane skids and enters this lane." The AI ​​scenario generation engine automatically parses the semantics and outputs the corresponding physical test parameters such as the friction coefficient, relative speed, and entry angle, enabling the rapid construction of complex scenarios.

[0081] Step 2: The cloud-based task orchestration module calculates and generates a continuous sequence of motion trajectory points and behavior triggering conditions based on the parameters defined in the scenario, and outputs a standardized test script in OpenSCENARIO format. Simultaneously, the system automatically verifies the validity of the parameters, including whether the speed and acceleration are within the chassis vehicle's performance range, whether the trajectory points are continuous without jumps, and whether the triggering condition logic is reasonable, to avoid issuing invalid tasks.

[0082] Step 3: The cloud-based collaborative control platform sends test task instructions to one or more multi-functional obstacle chassis vehicles via a 5G wireless link. For multi-vehicle collaborative test tasks, the cloud simultaneously issues independent motion trajectories to each chassis vehicle, along with a globally synchronized timestamp, to ensure the temporal consistency of the actions of multiple vehicles.

[0083] Step 4: After receiving the test task, the edge computing unit 5 of the multi-functional obstacle chassis vehicle executes the following sub-steps:

[0084] S4.1. Parse the test task instructions, extract the trajectory point sequence, velocity curve and behavior trigger conditions, and load them into the local control cache;

[0085] S4.2 Start the MPC trajectory tracking controller, and calculate the required steering angle, braking force and drive torque control quantities at a control frequency of 100Hz, combined with real-time positioning data.

[0086] S4.3. The control quantity is sent to the steer-by-wire unit, brake-by-wire unit and drive-by-wire unit via the bus to drive the chassis to perform the corresponding action;

[0087] S4.4 Calculate the relative distance or time of collision (TTC) with the vehicle under test in real time. When the value reaches the preset trigger condition, execute the corresponding preset test behavior, such as emergency braking, acceleration away, lane change, etc.

[0088] S4.5. Upload vehicle status data, including position, speed, attitude, control parameters, fault codes, and other information, to the cloud-based collaborative control platform at a frequency of 50Hz.

[0089] Step 5: During the test, the edge computing unit 5 of the multi-functional obstacle chassis vehicle stores process data such as control quantities and tracking errors in real time on the local storage medium; at the same time, the sensor data and control response data of the vehicle under test are also uploaded to the cloud collaborative control platform through the vehicle terminal. The data sampling rate is configurable, and the commonly used configuration is 100Hz.

[0090] Step 6: After receiving all test data, the cloud-based data management module first aligns the multi-source data based on the global timestamp, then cleans the data to remove outliers, and calculates core test metrics, including minimum safe distance, maximum deceleration, time to collision (TTC), trajectory lateral deviation, and braking response delay. Finally, the system automatically generates a standardized test report in PDF format, containing complete information such as test parameters, data curves, and pass / fail conclusions.

[0091] Step 7: After the test is completed, all test data, especially the raw sensor data and control response data of failed cases, are automatically injected into the training dataset of the AI ​​scene generation engine to optimize the scene mutation algorithm, generate more edge test cases under boundary conditions, form a closed-loop iteration mechanism, and continuously improve test coverage.

[0092] Step 8: During the test, edge computing unit 5 continuously monitors the communication heartbeat with the cloud. When the detected communication heartbeat loss exceeds the set threshold (default 500ms, configurable), the network disconnection autonomous security policy is automatically triggered, and the following operations are performed in sequence:

[0093] S8.1 Control the vehicle to smoothly decelerate to below 5km / h;

[0094] S8.2 Turn on the hazard warning lights to alert personnel on site;

[0095] S8.3 Control the vehicle to perform a side-by-side parking maneuver along the edge of the road;

[0096] S8.4 Store all operational data during the network outage on local storage media, and automatically re-upload it to the cloud after the network is restored to ensure data integrity.

[0097] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration, characterized in that, This includes a multi-functional obstacle course chassis and a cloud-based collaborative control platform that wirelessly connects to the multi-functional obstacle course chassis; The multi-functional obstacle chassis vehicle includes a chassis body, a rollable ramp structure on the roof, a quick-change interface, a positioning module, an edge computing unit, a communication module, and a status self-checking module. The chassis body integrates a steer-by-wire unit, a brake-by-wire unit, and a drive-by-wire unit, which are used to perform steering, braking, and driving actions. The rollable ramp structure on the roof is fixedly installed on the top of the chassis body and is used to guide the vehicle being tested to roll over it. The quick-change interface is arranged around the upper part of the chassis body and is used to connect and lock different types of simulation prop shells. The positioning module is fixedly installed at the geometric center of the chassis body and is used to collect the position and attitude data of the chassis. The edge computing unit is fixedly installed in the waterproof electrical control box of the chassis body. The drive-by-wire steering unit, drive-by-wire braking unit, drive-by-wire drive unit, positioning module, communication module and status self-test module are electrically connected to the edge computing unit and interact with each other. The communication module is used to establish a wireless communication link between the edge computing unit and the cloud-based collaborative control platform. The status self-check module is used to collect chassis operating status data and transmit it to the edge computing unit; The cloud-based collaborative control platform is used for test scenario generation, multi-vehicle task scheduling, test data reception and analysis, and test report output.

2. The vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration according to claim 1, characterized in that, The slope of the vehicle roof ramp structure has an angle of 15° with the horizontal plane. Rubber buffer pads are provided around the bottom of the ramp, and the surface of the ramp is processed with anti-slip texture. The vehicle roof ramp structure is fixedly connected to the mounting bracket on the upper part of the chassis body by bolts.

3. The vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration according to claim 1, characterized in that, The quick-change interface includes at least four mechanical positioning interfaces and a corresponding number of quick clamps; the mechanical positioning interface adopts a structure in which a conical positioning pin and a conical positioning hole are engaged, and the conical positioning pin is vertically fixed to the upper surface of the chassis body; the quick clamps are set next to the corresponding mechanical positioning interface and are used to lock the outer shell of the simulation prop.

4. The vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration according to claim 1, characterized in that, The steer-by-wire unit consists of a steering motor, a steering angle sensor, and a controller, and receives steering commands via a bus. The brake-by-wire unit is an EMB electromechanical braking module, equipped with a pressure sensor and a position sensor. The drive-by-wire unit uses a hub motor or a central motor, and controls vehicle speed via torque commands. The self-testing module integrates an inertial measurement unit, a battery management system, and a steer-by-wire actuator status monitoring circuit, which are used to detect chassis attitude, monitor power battery parameters, and collect feedback signals from the steer-by-wire actuators, respectively.

5. A vehicle autonomous driving function real-vehicle testing system based on cloud-edge collaboration according to claim 1, characterized in that, The positioning module is a combined inertial navigation system that combines RTK-GNSS and IMU; The edge computing unit uses an automotive-grade SoC chip and integrates an MPC trajectory tracking controller, an emergency strategy local adjudication module, and a network outage autonomous security module. The communication module is a 5G module, equipped with dual SIM card slots to support dual-link hot standby for operators; The cloud-based collaborative control platform has a built-in multi-vehicle scheduling module and an AI scene generation engine, which are used for multi-vehicle trajectory and time sequence allocation, and physical parameterization generation of natural language scenes, respectively.

6. An automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration, employing a real-vehicle testing system according to any one of claims 1-5, characterized in that, Includes the following steps: S1. Enter the test scenario description or select a preset scenario template in the cloud-based collaborative control platform to generate the corresponding scenario parameters; S2. The cloud-based collaborative control platform generates a sequence of motion trajectory points and behavior triggering conditions based on scene parameters, forms standardized test scripts, and verifies the legality of the parameters. S3, the cloud-based collaborative control platform distributes test tasks to at least one multi-functional obstacle chassis vehicle via a wireless link; S4, the edge computing unit of the multi-functional obstacle chassis vehicle analyzes the test task, drives the chassis body to perform test actions according to the predetermined trajectory and trigger conditions, and uploads vehicle status data in real time; S5. During the test, the chassis control process data and the test vehicle operation data are synchronously uploaded to the cloud collaborative control platform. The S6 cloud-based collaborative control platform performs timestamp alignment, data cleaning, and indicator calculation on the received data, and automatically generates a test report containing judgment conclusions.

7. The automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration according to claim 6, characterized in that, Step S1 also includes inputting a scene description in natural language form, which is automatically parsed by the AI ​​scene generation engine and outputs the corresponding physical test parameters.

8. The automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration according to claim 6, characterized in that, In step S4, the execution of the edge computing unit further includes the following sub-steps: S4.1 Analyze the test task instructions and extract the trajectory point sequence and behavior trigger conditions; S4.2 Start the MPC trajectory tracking controller to calculate the steering angle, braking force and drive torque control quantities at a frequency of 100Hz; S4.3, Drive the steer-by-wire unit, brake-by-wire unit and drive-by-wire unit to execute corresponding control commands; S4.4 monitors the distance or TTC value of the vehicle under test in real time, and executes the corresponding preset test behavior when the preset trigger conditions are met; S4.

5. Upload vehicle status data to the cloud-based collaborative control platform at a set frequency.

9. The automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration according to claim 6, characterized in that, Following step S6, the test data is injected into the training dataset of the AI ​​scene generation engine to optimize the scene mutation algorithm and generate edge test cases, thus forming a test closed loop.

10. The automated testing method for vehicle autonomous driving functions based on cloud-edge collaboration according to claim 6, characterized in that, During the test, when the edge computing unit detects that the communication heartbeat loss with the cloud collaborative control platform exceeds the set threshold, it automatically performs deceleration, turns on the warning lights, and pulls over to the side of the road. It also stores the operation data during the network outage locally and automatically re-uploads it to the cloud after the network is restored.

Images