Mesh wireless network device conflict detection processing method and system for intelligent driving test scene

Abstract

The application discloses a mesh wireless network device conflict detection processing method and system for an intelligent driving test scene. Before entering the test, active detection is performed on devices in an active state in the mesh wireless network, a mapping relationship between device IP and MAC is constructed, device conflicts of repeated IP, unregistered MAC or out-of-bound network segments are identified, a first detection result is obtained; during the test, address resolution protocol ARP packets in the mesh wireless network are passively monitored, device conflicts occurring during the test are detected, a second detection result is obtained; a multi-dimensional conflict feature vector is constructed based on the first detection result and / or the second detection result, a conflict level is determined by using a conflict determination model based on a gradient boosting decision tree GBDT, and a conflict risk classification result is output; and control instructions bound with the test process are executed according to the conflict risk classification result. The application completely solves the problem of network paralysis caused by unapproved new device access.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle testing technology, and in particular to a method and system for conflict detection and processing of mesh wireless network devices for intelligent driving testing scenarios. Background Technology

[0002] In closed-course real-vehicle testing of Advanced Driver Assistance Systems (ADAS), the equipment involved, such as test vehicles, base stations, various high-speed moving targets, and vehicle-mounted inertial navigation systems and robot control boxes, relies on a closed test LAN composed of a dedicated wireless mesh network to achieve low-latency communication. This closed test LAN spans multiple service network segments, and all devices use statically configured IP addresses. The number of devices typically exceeds 20, involving various roles such as wireless mesh network nodes, vehicle-end nodes, various types of mobile platforms (maximum speed 120km / h), inertial navigation equipment, driving robot control boxes, and auxiliary data acquisition equipment.

[0003] The most dangerous IP conflicts in actual testing stemmed from the access of new devices not registered in the device log. These conflicts fall into three categories: 1) An unregistered test computer connects and is configured with the same IP address as the core node, instantly triggering an ARP mapping flip that paralyzes the entire network; 2) After a device is replaced, the new MAC address uses the old IP address, triggering a latent conflict if that old IP address is already in use; 3) After a legitimate device is shut down, its IP address is occupied by an unregistered device, rendering traditional ARP detection completely ineffective (a detection blind spot) due to the original device being offline. These scenarios collectively lead to device disconnection, test interruptions, troubleshooting time of several hours, and pose a security risk of high-speed targets going out of control.

[0004] The existing IPv4 address conflict detection (ACD) defined in RFC5227 only supports self-testing within a single subnet when a new device connects, and cannot detect across network segments or perform pre-testing. Existing DHCP conflict protection relies on dynamic IP allocation, which is completely ineffective in ADAS testing scenarios. General scanning tools (such as Nmap) lack semantic awareness of the test device type and cannot accurately locate specific physical device labels. Furthermore, traditional ARP detection creates a security blind spot in scenario three, where a new device steals an IP address after the original device is powered off.

[0005] In summary, there is an urgent need for a proactive detection and prevention technology for inter-device IP conflicts that is designed for the testing environment of Advanced Driver Assistance Systems (ADAS), is forcibly integrated with the testing process, and can solve the detection blind spots in scenario three, in order to address the aforementioned problems. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings and defects of existing technologies and provide a method and system for conflict detection and processing of mesh wireless network devices for intelligent driving testing scenarios. It is used in closed wireless mesh test local area networks containing various high-speed moving targets, base stations, and vehicle-mounted devices to actively detect, prevent, and locate conflicts between unknown device access and static IP configurations.

[0007] One aspect of the present invention provides a method for conflict detection and processing of mesh wireless network devices for intelligent driving testing scenarios, comprising the steps of:

[0008] Before entering the test, active detection is performed on the active devices in the mesh wireless network to build the mapping relationship between device IP and MAC, compare it with the multi-dimensional device fingerprint baseline, identify device conflicts such as duplicate IP, unregistered MAC or out-of-bounds network segment, and obtain the first detection result;

[0009] During the entry test, the Address Resolution Protocol (ARP) packets in the mesh wireless network are passively monitored. The monitored device MAC information is compared with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result.

[0010] Based on the multidimensional conflict feature vector constructed from the first and / or second detection results, the conflict level is determined using the conflict determination model based on gradient boosting decision tree (GBDT), and the corresponding conflict risk classification result is output.

[0011] Based on the conflict risk classification results, control instructions bound to the testing process are executed automatically.

[0012] As a preferred option, the multi-dimensional device fingerprint baseline includes the device's function label, IP address, MAC address, role type, legal network segment, priority, and responsible person.

[0013] Preferably, the role types include base station nodes, vehicle-end nodes, various types of platform nodes, inertial navigation devices, robot control boxes, and platform controllers.

[0014] Preferably, the legitimate network segment includes the main communication network segment, the main control network segment, and the auxiliary network segment.

[0015] As a preferred option, the multidimensional conflict feature vector includes active scan conflict volume, number of conflicting devices, IP segment compliance anomalies, ARP flip-flop frequency, MAC ledger matching degree, conflicting device priority, test mode classification code, and number of currently online devices.

[0016] Preferably, the conflict determination model based on gradient boosting decision tree (GBDT) determines the conflict level and outputs the corresponding conflict risk classification result, including:

[0017] The multidimensional conflict feature vectors are processed using a conflict determination model based on gradient boosting decision tree (GBDT), which outputs different preset classification probability vectors. The determination is made based on the different preset classification probability vectors, and the corresponding conflict risk classification results are output.

[0018] Preferably, the control commands bound to the testing process include the issuance, verification, or revocation of test release tokens, as well as triggering stop control of physical devices in the testing scenario when a high-risk conflict is determined.

[0019] A second aspect of the present invention provides a mesh wireless network device conflict detection and processing system for intelligent driving testing scenarios, comprising:

[0020] The device fingerprint baseline module is used to store and maintain a multi-dimensional device fingerprint ledger baseline for legitimate devices in the test scenario;

[0021] The active detection module is used to actively detect active devices in the mesh wireless network before entry testing, obtain the mapping relationship between device IP and MAC, compare it with the multi-dimensional device baseline, identify duplicate IPs, unregistered devices or network segment out-of-bounds behavior, and obtain the first detection result.

[0022] The passive listening module is used to passively listen to the Address Resolution Protocol (ARP) packets in the mesh wireless network during the entry test. It compares the listened device MAC information with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result.

[0023] The conflict adjudication module is used to determine the conflict level based on the multi-dimensional conflict feature vector constructed by the first detection result and / or the second detection result, and output the corresponding conflict risk classification result by using the conflict adjudication model based on gradient boosting decision tree (GBDT).

[0024] The execution control module is used to automatically execute control instructions bound to the test process based on the conflict risk results.

[0025] This invention completely solves the problem of network paralysis caused by the access of unapproved new devices by conducting tests at different stages before and during vehicle testing. Through multi-dimensional fingerprint verification and detection, it breaks through the three detection blind spots in intelligent testing scenarios, ensuring that unregistered devices cannot disrupt the stable operation of the network topology. It can also convert the IP and MAC mapping of devices into device role labels, output the responsible person and modification suggestions, and reduce the troubleshooting time from several hours to minutes, achieving extremely fast troubleshooting.

[0026] This invention uses a preset conflict determination model to automatically revoke the token and issue an electronic stop command when a high-risk conflict is detected. The entire process is unmanned, eliminating the safety risks caused by delays in manual response and ensuring the physical safety of the experiment.

[0027] This invention achieves full coverage before testing through active scanning and millisecond-level response during testing through passive monitoring, realizing full-time adaptive safety protection for closed intelligent driving test sites. It is applicable to network security control of all closed intelligent driving test sites using wireless mesh networks. Attached Figure Description

[0028] Figure 1 This is a flowchart of the mesh wireless network device conflict detection and processing method of the present invention.

[0029] Figure 2 This is a schematic diagram of the mesh wireless network device conflict detection and processing system of the present invention. Detailed Implementation

[0030] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0031] See Figure 1 As shown in the exemplary embodiment of this application, the mesh wireless network device conflict detection and processing method for intelligent driving test scenarios includes the following steps:

[0032] Before entering the test, active detection is performed on the active devices in the mesh wireless network to build the mapping relationship between device IP and MAC, compare it with the multi-dimensional device fingerprint baseline, identify device conflicts such as duplicate IP, unregistered MAC or out-of-bounds network segment, and obtain the first detection result;

[0033] During the entry test, the Address Resolution Protocol (ARP) packets in the mesh wireless network are passively monitored. The monitored device MAC information is compared with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result.

[0034] Based on the multidimensional conflict feature vector constructed from the first and / or second detection results, the conflict level is determined using the conflict determination model based on gradient boosting decision tree (GBDT), and the corresponding conflict risk classification result is output.

[0035] Based on the conflict risk classification results, control instructions bound to the testing process are executed automatically.

[0036] In this application, proactive conflict detection is performed before the test to identify device conflicts such as duplicate IPs, unregistered MAC addresses, or out-of-bounds network segments. This ensures that the IPs of active devices accessing the preset mesh wireless network are not duplicated or conflicting. If a conflict is detected, it is modified, such as changing the conflicting device IP to a valid free address and synchronizing the recorded data of the multi-dimensional device fingerprint baseline.

[0037] During the experiment, ARP packets in the mesh wireless network were passively monitored. The monitored device MAC information was compared with the multi-dimensional device fingerprint baseline in real time to detect device conflicts, such as MAC conflicts with preset IP addresses. Abnormal MACs were captured in milliseconds. This achieved real-time protection during the test, preventing unauthorized devices from accessing the test and affecting its normal progress or causing serious losses.

[0038] Before or during the test, this application uses multi-dimensional feature vectors obtained through detection to determine and adjudicate conflicts. When a high-risk conflict is detected through multi-dimensional feature vector adjudication, physical device control commands are automatically triggered (such as revoking the token, stopping the acceleration of the target device during the test, controlling the access control to perform corresponding actions, etc.) to prevent equipment network collapse, test interruption, and loss of control of high-speed target objects.

[0039] This application achieves full coverage through proactive pre-test detection, enabling access control and preventing conflicting devices from entering the test environment. During the test, passive monitoring provides millisecond-level real-time response and continuous test protection, ensuring smooth operation. This dual-modal operation of proactive detection and passive monitoring generates a multi-dimensional feature vector as the basis for conflict resolution.

[0040] In one embodiment, the multidimensional device fingerprint baseline includes preset device function tags (general functional descriptions), IP address, MAC address, role type, legal network segment, priority, and responsible person. Priorities are set such that device one has a higher priority than device two, and device two has a higher priority than device three.

[0041] For example, in this application, the role types include base station nodes, vehicle-end nodes, various platform nodes, inertial navigation devices, robot control boxes, and platform controllers, all of which are devices or gateway devices involved in intelligent driving test scenarios. The specific types can be determined based on the various devices in the test scenario. For example, the legally belonging network segment includes the main communication network segment, the main control network segment, and the auxiliary network segment.

[0042] The base station node is a reference positioning base station deployed in a fixed location in an enclosed area; it provides high-precision differential positioning reference coordinates for RTK differential correction of vehicles and robots; it synchronizes the time of the entire network, outputs position reference data, and monitors the position status of all nodes in the area; it is the central reference for the position and time of the entire test site.

[0043] The vehicle-side node is an on-board edge node installed on the actual autonomous driving / assisted driving vehicle under test; it collects vehicle driving status, speed, steering, braking, radar, and camera perception data; receives differential positioning information from base stations, calculates the vehicle's high-precision position and attitude; uploads real-time vehicle operation data to the platform, and performs ADAS scenario test data collection and interaction.

[0044] The various platform nodes are the control nodes corresponding to various simulation / motion platforms, mobile target vehicles, dummy platforms, and trajectory simulation platforms in the closed venue, excluding real vehicles; they are responsible for simulating the movement trajectories of road participants (pedestrians, non-motorized vehicles, and social vehicles); they receive instructions from the main platform and automatically drive, change lanes, and cross traverses according to the test conditions, cooperating with real vehicles to complete the reproduction of ADAS collision avoidance, warning, and other scenarios.

[0045] The inertial navigation device (INS) can continuously output high-precision attitude, angular velocity, acceleration, heading angle, and position without relying on satellites or base stations; it can provide supplementary positioning when satellite signals are weak or blocked (under bridges, in the shade of trees, or next to buildings); when installed on a vehicle, it outputs instantaneous attitude data of the vehicle / moving platform, providing attitude, heading, and lateral and longitudinal displacement references for ADAS algorithms, ensuring continuous and uninterrupted test positioning.

[0046] The robot control box is a core control unit specifically designed to manage the field testing robot, walking dummy, and track motion mechanism; it receives operating condition commands from the host computer, drives the servo mechanism, and precisely controls start / stop, walking speed, and motion trajectory; it also transmits the robot's posture, position, and operating status back in real time to ensure consistent reproduction of actions in the test scenario.

[0047] The platform controller is the core control unit / central management platform of the entire closed test site; it uniformly schedules the timing and operating conditions of base stations, vehicle terminals, various platform nodes, inertial navigation systems, and robots; it issues test tasks, configures test scenarios, and summarizes all test data; it completes process start-up and shutdown, status monitoring, anomaly alarms, test data storage and diagnosis, and coordinates the entire ADAS test process closed loop.

[0048] In the intelligent driving test scenario, the preset mesh wireless network is the preset local area network of the test scenario. All devices related to the test are connected to this local area network to exchange information, so as to realize intelligent control and remote control of the test.

[0049] The active detection of devices in an active state in a mesh wireless network as described in this application includes:

[0050] The system constructs a mapping relationship between device IP and MAC addresses, initiates a full ARP request to the entire network segment, collects the responses from devices in the mesh wireless network, compares them with the multidimensional device fingerprint baseline, and identifies device conflicts such as duplicate IPs, unregistered MAC addresses, or devices crossing network segments based on the comparison results, thus obtaining the first detection result.

[0051] During proactive detection, the IP→MAC mapping in the mesh wireless network is compared with the stored multi-dimensional device fingerprint baseline data. If the MAC address is not in the pre-stored data table, even if the IP address corresponds to only this single MAC address, a high-risk alarm is triggered, and the device is locked to prevent potential conflicts from external devices. This process distinguishes between hard conflicts (multiple MAC addresses for the same IP address) and soft conflicts (unknown MAC addresses occupying legitimate IP addresses). The detection can be performed by traversing the multi-dimensional device fingerprint baseline storage table using the MAC address as an index to check the uniqueness of the IP address and MAC address. Simultaneously, the proactive detection process also automatically verifies the compliance of IP ranges and detects out-of-bounds behavior.

[0052] Furthermore, in this embodiment of the application, before the active detection before the test, IP static detection is also performed. The IP is used as an index to traverse the multi-dimensional device fingerprint baseline storage table to perform IP repeatability detection and IP compliance detection, which is completed in milliseconds. If a high-risk state is detected, the test is controlled to stop immediately.

[0053] In one feasible embodiment, the conflict determination model based on gradient boosting decision tree (GBDT) determines the conflict level and outputs the corresponding conflict risk classification result, including:

[0054] The multidimensional conflict feature vectors are processed using a conflict determination model based on gradient boosting decision tree (GBDT), which outputs different preset classification probability vectors. The determination is made based on the different preset classification probability vectors, and the corresponding conflict risk classification results are output.

[0055] The multidimensional conflict feature vector includes the active scan conflict quantity E_active (normalized [0,1]), the number of conflicting devices N_conflict, the IP segment compliance anomaly F_segment (binary), the ARP flip frequency E_passive (normalized value), the MAC ledger matching degree E_ledger (MAC ledger matching degree refers to whether a device's MAC exists in the preset ledger, which can take two values ​​0 / 1), the conflicting device priority P_device, the test mode classification code M_test, and the number of currently online devices N_online.

[0056] By employing a Gradient Boosting Decision Tree (GBDT)-based conflict resolution model, leveraging GBDT's inherent support for heterogeneous feature fusion, the model eliminates the need for manually preset weights. It iteratively constructs decision trees to automatically learn the contributions of each feature. The model exhibits low inference latency, meeting real-time resolution requirements, and its feature importance is interpretable. The GBDT conflict resolution model automatically learns weights by fusing multi-source evidence and adjusts the resolution strategy according to experimental modes. Incremental updates allow the system to continuously adapt to changes in the network environment. The GBDT conflict resolution model can be trained using supervised learning: historical network scan records and manually labeled conflict levels serve as the training set, with logarithmic loss and cross-validation for tuning. Periodic updates using incremental data adapt to changes in the network environment.

[0057] In one feasible embodiment, the step of making a judgment based on different preset classification probability vectors and outputting the corresponding conflict risk classification result includes:

[0058] Different classification probability vectors are compared with preset thresholds, and the corresponding conflict risk classification results are output based on the threshold comparison results. Finally, based on the output conflict risk classification results, control commands bound to the testing process are executed to perform corresponding operations and automatically control preset equipment, such as automatically controlling access control of the test site, to ensure the safe operation of the test scenario and prevent unauthorized equipment from entering the test site.

[0059] For example, the GBDT conflict resolution model outputs a three-class probability vector [P(CRITICAL), P(WARNING), P(INFO)]. The decision rules are as follows: The class with the highest probability is selected; if P(CRITICAL) ≥ 0.7, CRITICAL is prioritized and a trial lock and token revocation are triggered; if P(CRITICAL) ∈ [0.3, 0.7), WARNING requires manual confirmation, and a suspension control operation is performed based on the confirmation result; if P(CRITICAL) < 0.3 and P(INFO) is at its maximum, normal passage is allowed, and a token is issued.

[0060] The detection technology in this application can perform pre-experiment IP Consistency Registration Protocol (PIPP) verification before the experiment, unifying the pre-verification process and executing three verification stages sequentially. A token can only be issued if all three stages pass: Stage 1 (Static Ledger Scan): The ledger is traversed using IP addresses as indexes to detect duplicate conflicts and network segment compliance, completed in milliseconds. Stage 2 (Active ARP Scan and MAC Fingerprint Comparison): An ARP request (≥20 threads) is sent to the activated device's IP address, performing IP→MAC uniqueness detection and MAC→ledger matching detection, taking ≤30 seconds. Stage 3 (GBDT Conflict Determination): Multi-dimensional features are input into the GBDT model, which outputs classification probabilities and determines, according to the determination rules, whether to issue a token or lock the experiment, while simultaneously generating a structured diagnostic report, completed in milliseconds.

[0061] The data fields of the structured diagnostic report may include timestamps, conflict types, GBDT determination probabilities, device function tags, responsible persons, and modification suggestions.

[0062] Furthermore, the mesh network device conflict detection and processing method described in this application can realize the automated control of the test site based on the access control sign mechanism of physical devices in the network status automatic control test scenario.

[0063] For example, a standardized access control interface is provided, which automatically converts network anomaly determinations from the GBDT conflict determination model into electronic control commands for physical devices. The entire control process requires no manual intervention. The control procedure is set up according to the flow of token application - token carrying - token verification - token revocation in case of network anomaly, including:

[0064] Token application: Configure the scheduling software to call the pre-inspection interface, automatically execute the PIPP protocol, and if the GBDT conflict determination model passes the judgment, issue a timed token (default 10 minutes); otherwise, return "prohibit startup".

[0065] Token delivery: The scheduling software sends the token to the platform controller along with the startup command;

[0066] Token verification: The platform controller automatically calls the verification interface before acceleration. If the verification is successful, the process is allowed; otherwise, it is rejected.

[0067] Automatic emergency stop for network anomalies: In the experiment, after the passive monitoring detects an abnormal signal, it inputs the GBDT conflict determination model in real time for GBDT inference. If P(CRITICAL) exceeds the threshold, the token is automatically revoked, an emergency stop command is issued, and the platform controller controls the motion execution module to stop acceleration through electronic signals.

[0068] See Figure 2 As shown in the figure, a further embodiment of the present invention provides a mesh wireless network device conflict detection and processing system for intelligent driving test scenarios, comprising:

[0069] The device fingerprint baseline module is used to store and maintain a multi-dimensional device fingerprint ledger baseline for legitimate devices in the test scenario;

[0070] The active detection module is used to actively detect active devices in the mesh wireless network before entry testing, obtain the mapping relationship between device IP and MAC, compare it with the multi-dimensional device baseline, identify duplicate IPs, unregistered devices or network segment out-of-bounds behavior, and obtain the first detection result.

[0071] The passive listening module is used to passively listen to the Address Resolution Protocol (ARP) packets in the mesh wireless network during the entry test. It compares the listened device MAC information with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result.

[0072] The conflict adjudication module is used to determine the conflict level based on the multi-dimensional conflict feature vector constructed from the first detection result and / or the second detection result, and output the corresponding conflict risk classification result using the Gradient Boosting Decision Tree (GBDT) conflict adjudication model.

[0073] The execution control module is used to automatically execute control instructions bound to the test process based on the conflict risk results.

[0074] The multidimensional device fingerprint baseline includes the preset device's function label, IP address, MAC address, role type, legal network segment, priority, and responsibility.

[0075] The control commands bound to the testing process include the issuance, verification, or revocation of test release tokens, as well as triggering stop control of physical devices in the testing scenario when a high-risk conflict is identified.

[0076] In this application, the conflict resolution module comprehensively evaluates the multi-dimensional features of active detection and passive listening, compares the output high-risk probability vector P (CRITICAL) with a preset threshold, and triggers a test stop signal if the threshold is exceeded. The execution module automatically revokes the token and sends an emergency braking command to the platform controller to control the target device to accelerate to a stop.

[0077] The aforementioned modules are interdependent. For example, the baseline module provides a comparison benchmark for multi-dimensional device fingerprint baselines, the active detection module provides full coverage detection, the monitoring module passively monitors to enable real-time device responses, the conflict resolution module performs fusion judgment based on preset multi-dimensional features, and the execution module converts access control signs into physical controls. This achieves a closed loop of detection, feature extraction, model judgment, and control, greatly improving the security of testing in the test scenario.

[0078] Test Implementation Example:

[0079] Implementation Environment: Deployed and operated at an ADAS test site. The test LAN involves multiple service network segments and approximately 30 devices, including wireless mesh network nodes, inertial navigation equipment (dual-segment configuration), driving robot control boxes, various types of mobile platform controllers, and auxiliary data acquisition equipment, all using static IP configurations.

[0080] IP Conflicts Detected: During the first phase of PIPP deployment, two serious IP conflicts were detected. Conflict 1: The main communication network segment IP 192.168.1.28 was simultaneously occupied by a medium-sized mobile platform node and an in-vehicle inertial navigation device, causing a wireless mesh network outage. Conflict 2: The main control network segment IP 195.0.0.202 was simultaneously occupied by two driving robot control boxes, causing control signal corruption. Specific modification suggestions for each conflict output (e.g., changing the conflicting device IPs to valid, free addresses and synchronizing the logbook).

[0081] Listening Detection and Automatic Emergency Stop: In the verification test, an unregistered laptop connected to the network and was configured with a registered IP address. The listening module captured its Gratuitous ARP packets within approximately 0.5 seconds, and confirmed that the MAC address was not registered after comparison. The feature vector was input into the GBDT collision determination model, and the output P(CRITICAL) = 0.92 (exceeding the threshold of 0.7). Automatic actions were taken: the access token was revoked; an emergency stop command was issued, and the platform controller stopped acceleration via electronic signals; a structured diagnostic report was output. The end-to-end link took approximately 1 second.

[0082] The initial training dataset for the GBDT conflict resolution model came from approximately 500 network scan records and corresponding manually labeled data from the site over the past three months. Training was performed using the LightGBM framework, achieving a 94.3% accuracy rate with five-fold cross-validation. After deployment, incremental updates were performed weekly using newly added data to fine-tune the model and adapt it to changes in network topology and testing modes. The model inference latency was less than 10 milliseconds, which did not affect the experiment preparation process.

[0083] After deployment, the number of test interruptions caused by IP conflicts was reduced to zero; the average execution time of the three phases of PIPP was 3 to 8 seconds; troubleshooting time was reduced from several hours to less than 5 minutes; access control enabled automatic emergency shutdown of physical devices in case of network anomalies; and incremental updates of GBDT continuously improved the accuracy of new conflict identification.

[0084] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention.

[0085] Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the claims be included within the invention.

[0086] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for conflict detection and handling of mesh wireless network devices for intelligent driving testing scenarios, characterized in that, Including the following steps: Before entering the test, active detection is performed on the active devices in the mesh wireless network to build the mapping relationship between device IP and MAC, compare it with the multi-dimensional device fingerprint baseline, identify device conflicts such as duplicate IP, unregistered MAC or out-of-bounds network segment, and obtain the first detection result; During the entry test, the Address Resolution Protocol (ARP) packets in the mesh wireless network are passively monitored. The monitored device MAC information is compared with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result. Based on the multidimensional conflict feature vector constructed from the first and / or second detection results, the conflict level is determined using the conflict determination model based on gradient boosting decision tree (GBDT), and the corresponding conflict risk classification result is output. Based on the conflict risk classification results, control instructions bound to the testing process are executed automatically.

2. The collision detection and processing method for mesh wireless network devices according to claim 1, characterized in that, The multidimensional device fingerprint baseline includes the preset device's function label, IP address, MAC address, role type, legal network segment, priority, and responsible person.

3. The collision detection and processing method for mesh wireless network devices according to claim 2, characterized in that, The roles include base station nodes, vehicle-end nodes, various types of platform nodes, inertial navigation devices, robot control boxes, and platform controllers.

4. The collision detection and processing method for mesh wireless network devices according to claim 2, characterized in that, The legally owned network segments include the main communication network segment, the main control network segment, and the auxiliary network segment.

5. The collision detection and processing method for mesh wireless network devices according to claim 1, characterized in that, The multidimensional conflict feature vector includes active scan conflict quantity, number of conflicting devices, IP segment compliance anomalies, ARP flip-flop frequency, MAC ledger matching degree, conflicting device priority, test mode classification code, and number of currently online devices.

6. The collision detection and processing method for mesh wireless network devices according to claim 1, characterized in that, The conflict determination model based on gradient boosting decision tree (GBDT) determines the conflict level and outputs the corresponding conflict risk classification result. This includes: processing the multi-dimensional conflict feature vector using the GBDT-based conflict determination model and outputting preset different classification probability vectors; making a determination based on the preset different classification probability vectors and outputting the corresponding conflict risk classification result.

7. The collision detection and processing method for mesh wireless network devices according to claim 1, characterized in that, The control commands bound to the testing process include the issuance, verification, or revocation of test release tokens, as well as triggering stop control of physical devices in the testing scenario when a high-risk conflict is identified.

8. A mesh wireless network device conflict detection and processing system for intelligent driving testing scenarios, characterized in that, include: The device fingerprint baseline module is used to store and maintain a multi-dimensional device fingerprint ledger baseline for legitimate devices in the test scenario; The active detection module is used to actively detect active devices in the mesh wireless network before entry testing, obtain the mapping relationship between device IP and MAC, compare it with the multi-dimensional device baseline, identify duplicate IPs, unregistered devices or network segment out-of-bounds behavior, and obtain the first detection result. The passive listening module is used to passively listen to the Address Resolution Protocol (ARP) packets in the mesh wireless network during the entry test. It compares the listened device MAC information with the multi-dimensional device fingerprint baseline in real time to detect device conflicts that occur during the test and obtain a second detection result. The conflict adjudication module is used to construct a multi-dimensional conflict feature vector based on the first detection result and / or the second detection result, use the Gradient Boosting Decision Tree (GBDT) conflict adjudication model to determine the conflict level, and output the corresponding conflict risk classification result. The execution control module is used to automatically execute control instructions bound to the test process based on the conflict risk results.

9. The mesh wireless network device collision detection and processing system according to claim 8, characterized in that, The multidimensional device fingerprint baseline includes the preset device's function label, IP address, MAC address, role type, legal network segment, priority, and responsibility.

10. The mesh wireless network device collision detection and processing system according to claim 8, characterized in that, The control commands bound to the testing process include the issuance, verification, or revocation of test release tokens, as well as triggering stop control of physical devices in the testing scenario when a high-risk conflict is identified.

Images