An automatic driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response

Abstract

The application discloses a kind of multi-dimension fusion and hierarchical response's automatic driving ODD monitoring and control system, including multi-dimension detection unit, deviation degree evaluation unit, space-time edge prediction unit, response decision unit, safety control vehicle execution unit, dynamic ODD update unit, the monitoring dimension of the present application is comprehensive, wide coverage, deviation evaluation is refined, quantification degree is high, hierarchical response is gentle, driving safety is strong, risk pre-judgment is front, reserves sufficient takeover time, double-layer adaptive optimization, adaptation ability is excellent, data interaction cooperation, scene adaptability is stronger, architecture is in line with specification, and practicality is high: the present application meets the latest L3 level automatic driving relevant standard requirements, especially the requirement of ODD monitoring and minimum risk strategy, system architecture is mature and stable, and it is convenient for real vehicle to carry and mass production landing application.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology for intelligent connected vehicles, and in particular to an autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response. Background Technology

[0002] With the formal implementation of national standards such as GB / T44721-2024 "Safety Requirements for Automated Driving Systems of Intelligent Connected Vehicles" and GB / T45312-2025 "Design and Operation Conditions for Automated Driving Systems of Intelligent Connected Vehicles," the Design Operating Domain (ODD) has become a core fundamental requirement for the access and operation of Level 3 automated driving systems. The standards clearly stipulate that automated driving systems can only activate and execute dynamic driving tasks within their clearly defined ODD boundaries. Once outside the ODD boundaries, a takeover request must be triggered or the Minimum Risk Decision (MRM) strategy must be executed.

[0003] However, current ODD monitoring technologies in the industry have the following shortcomings:

[0004] 1. Single monitoring dimension: Most systems only focus on static factors such as environmental conditions, road type and speed range, ignoring the comprehensive evaluation of the system's own health status, driver takeover capability status and V2X collaborative information;

[0005] 2. Simple response strategy: It usually adopts "binary" judgment logic, with only two states: "normal operation within ODD" and "immediate exit outside ODD". It lacks intermediate transition states, which can easily lead to panic and misoperation when the driver suddenly takes over.

[0006] 3. The incremental degradation mechanism is imperfect: The existing incremental degradation mechanism is just a simple functional degradation, without considering the combined impact of different ODD elements' deviation degrees, and cannot achieve refined security control;

[0007] 4. Static ODD Boundaries: ODD boundaries are preset and fixed, and cannot be dynamically adjusted according to actual operating data and environmental changes, which may cause the system to be too conservative or too aggressive in some edge scenarios. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a multi-dimensional fusion and hierarchical response monitoring and control system for autonomous driving ODD.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0010] An autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response includes a multi-dimensional detection unit, a deviation evaluation unit, a spatiotemporal edge prediction unit, a response decision unit, a safe vehicle control execution unit, and a dynamic ODD update unit.

[0011] The multi-dimensional detection unit is used to collect multi-dimensional ODD elements, which include environmental ODD elements, road ODD elements, system ODD elements, and human-machine ODD elements, and sends the multi-dimensional ODD elements to the deviation assessment unit.

[0012] The dynamic ODD update unit is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element based on the ODD edge scene data it collects in real time, thereby obtaining the updated static ODD boundary information, response threshold, and threshold corresponding to each ODD element. The updated static ODD boundary information is the current static ODD boundary information, the updated response threshold is the current response threshold, and the updated threshold corresponding to each ODD element is the current threshold corresponding to each ODD element. The dynamic ODD update unit sends the current static ODD boundary information to the spatiotemporal edge prediction unit, the dynamic ODD update unit sends the current response threshold to the response decision unit, and the dynamic ODD update unit sends the current threshold corresponding to each ODD element to the deviation evaluation unit.

[0013] The deviation assessment unit is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element through the dual logic of "threshold comparison + grade quantification". Then, combined with the element weight coefficient and the combined influence coefficient, the comprehensive deviation is finally calculated and sent to the spatiotemporal edge prediction unit and the response decision unit.

[0014] The spatiotemporal edge prediction unit is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information. Then, it extracts the static ODD boundary information on the predicted future driving path from the map, and then corrects the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data to obtain the dynamic ODD boundary. Then, it calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the real-time vehicle status, and sends the estimated time to the response decision unit.

[0015] The response decision unit is used to generate a graded response control command based on the current response threshold, comprehensive deviation, and estimated time, and send the graded response control command to the safety vehicle control execution unit;

[0016] The safety vehicle control execution unit is used to perform corresponding vehicle speed adjustments, autonomous driving function downgrades, multimodal human-machine interaction, and minimum-risk parking operations according to the graded response control commands.

[0017] Preferably, the multi-dimensional detection unit includes an environmental dimension detection module, a road dimension detection module, a system dimension detection module, and a human-machine dimension detection module. The environmental dimension detection module is used to collect environmental ODD elements, including weather ODD elements, illumination ODD elements, and visibility ODD elements. The road dimension detection module is used to collect road ODD elements, including road type ODD elements, lane line clarity ODD elements, curve curvature ODD elements, slope ODD elements, construction area ODD elements, and traffic flow density ODD elements. The system dimension detection module is used to collect system ODD elements, including sensor health status ODD elements, computing platform status ODD elements, drive-by-wire system status ODD elements, and high-precision map matching degree ODD elements. The human-machine dimension detection module is used to collect human-machine ODD elements, including driver attention status ODD elements, hand position ODD elements, eye gaze ODD elements, seat belt status ODD elements, and whether the driver is in the driver's seat ODD elements.

[0018] Preferably, the weather ODD elements include rainfall ODD elements, snowfall ODD elements, fog concentration ODD elements, and wind speed ODD elements, and the illumination ODD elements include daytime illumination intensity ODD elements, dusk or dawn illumination intensity ODD elements, nighttime illumination intensity with streetlights ODD elements, nighttime illumination intensity without streetlights ODD elements, and backlight intensity ODD elements.

[0019] As a preferred option, the road type ODD elements include highway ODD elements, urban expressway ODD elements, urban arterial road ODD elements, and ordinary highway ODD elements.

[0020] Preferably, the sensor health status ODD elements include camera ODD elements, millimeter-wave radar ODD elements, lidar ODD elements, and GNSS global navigation satellite system positioning status ODD elements.

[0021] Preferably, the deviation assessment unit includes a deviation calculation module, an element weight module, a combined influence coefficient calculation module, and a comprehensive deviation calculation module.

[0022] The deviation calculation module is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element, through the dual logic of "threshold comparison + grade quantification", and send the deviation value of each ODD element to the comprehensive deviation calculation module.

[0023] Preferably, the element weighting module is used to assign a weight coefficient to the impact of each ODD element on system security, and send the weight coefficient of each ODD element on system security to the comprehensive deviation calculation module.

[0024] As a preferred embodiment, the combined influence coefficient calculation module is used to calculate the combined influence coefficient when multiple ODD elements deviate simultaneously, and then sends the combined influence coefficient to the comprehensive deviation calculation module.

[0025] Preferably, the comprehensive deviation calculation module is used to calculate the comprehensive deviation based on the deviation value of each ODD element, the weight coefficient of each ODD element's impact on system security, and the combined impact coefficient, and then send the comprehensive deviation to the spatiotemporal edge prediction unit and the response decision unit.

[0026] Preferably, the spatiotemporal edge prediction unit includes a path prediction module, an ODD boundary extraction module, a dynamic boundary correction module, and a time calculation module;

[0027] Preferably, the path prediction module is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information, and then send the future driving path of the vehicle to the ODD boundary extraction module.

[0028] As a preferred option, the ODD boundary extraction module is used to extract static ODD boundary information from the map on the predicted future driving path of the vehicle.

[0029] As a preferred embodiment, the dynamic boundary correction module is used to correct the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data, so as to obtain the dynamic ODD boundary, and send the dynamic ODD boundary to the time calculation module.

[0030] Preferably, the time calculation module calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the vehicle's real-time status, and sends the estimated time to the response decision unit.

[0031] Preferably, the static ODD boundary information includes static ODD boundary information corresponding to lane lines, static ODD boundary information corresponding to road curvature, static ODD boundary information corresponding to elevated road sections, static ODD boundary information corresponding to tunnel sections, and static ODD boundary information corresponding to traffic light sections.

[0032] As a preferred embodiment, the safety vehicle control execution unit includes a vehicle speed smooth adjustment module, a function deactivation module, a multimodal human-machine interaction module, and a minimum risk parking execution module;

[0033] Preferably, the vehicle speed smoothing adjustment module is used to perform vehicle speed adjustment according to the graded response control command;

[0034] The function shutdown module is used to degrade the autonomous driving function according to the graded response control command;

[0035] The multimodal human-computer interaction module is used to execute multimodal human-computer interaction based on hierarchical response control commands;

[0036] The minimum risk parking execution module is used to perform minimum risk parking operations according to the graded response control instructions.

[0037] As a preferred embodiment, the dynamic ODD update unit includes an edge scene data acquisition module, a cloud big data analysis module, an ODD boundary adjustment module, and an OTA upgrade module;

[0038] The edge scene data acquisition module is used to collect ODD edge scene operation data in real time and send the ODD edge scene operation data to the cloud big data analysis module. The ODD edge scene operation data includes data on the area where the vehicle approaches the lane line disappears, data on the area where the road curvature exceeds the limit, data on the entrance and exit of the elevated road, data on the entrance and exit of the tunnel, data on the boundary of the traffic signal light control, and data on sudden environmental anomalies.

[0039] The cloud-based big data analysis module is used to analyze real-time ODD edge scene operation data and historical ODD edge scene operation data to obtain static ODD boundary information, response threshold, and adjustment strategy for the threshold corresponding to each ODD element, and then sends the adjustment strategy to the ODD boundary adjustment module.

[0040] The ODD boundary adjustment module is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element according to the adjustment strategy, so as to obtain the current static ODD boundary information, current response threshold, and current threshold corresponding to each ODD element, and send the above information to the OTA upgrade module.

[0041] The OTA upgrade module sends the current static ODD boundary information to the spatiotemporal edge prediction unit, sends the current response threshold to the response decision unit, and sends the current threshold corresponding to each ODD element to the deviation evaluation unit.

[0042] The beneficial effects of this invention are as follows:

[0043] 1. Comprehensive monitoring dimensions and wide coverage: This invention establishes a four-dimensional integrated ODD monitoring architecture encompassing environment, road, system, and human-machine interaction. Through multi-dimensional detection units, it completes the collection of all categories of operational design domain elements, fully encompassing various influencing factors such as natural environment, road conditions, vehicle equipment status, and driver and passenger status. It can comprehensively determine the degree of matching between the actual operating conditions of the vehicle and the safe operating domain. Compared with single-dimensional monitoring methods, the evaluation results are more comprehensive and accurate.

[0044] 2. Refined and highly quantified deviation assessment: This invention relies on the deviation assessment unit to construct a multi-level deviation evaluation system. It can not only calculate the independent deviation value of each type of ODD element, but also introduce the element safety weight coefficient and the multi-element combination influence coefficient to comprehensively judge the superimposed risks brought about by the synchronous deviation of multiple elements. This achieves refined quantitative classification of ODD deviation status and solves the defects of traditional assessment methods that are rough and ignore the combined risks.

[0045] 3. Smooth, graded response for enhanced driving safety: This invention employs a graded response decision-making system combined with a progressive vehicle control execution strategy. The response decision-making unit matches corresponding level control commands based on the comprehensive deviation and the estimated time for the vehicle to reach the dynamic ODD boundary. The safety vehicle control execution unit sequentially completes the smooth adjustment of vehicle speed, downgrade of autonomous driving functions, human-machine interaction warning, and minimum risk parking, eliminating the emergency forced exit mode. This effectively avoids driver panic and misoperation, significantly improving driving safety and passenger comfort during autonomous driving operation.

[0046] 4. Risk prediction in advance, with sufficient takeover time: This invention uses a spatiotemporal edge prediction unit to predict driving paths and dynamically correct ODD boundaries. It combines navigation information to predict the vehicle's future driving trajectory, extracts the static ODD boundary from the map, and then integrates real-time environmental data, real-time traffic data, and ODD edge scene operation data to complete dynamic boundary correction. It accurately calculates the estimated time for the vehicle to reach the dynamic ODD boundary, provides early warning of boundary crossing risks, and allows drivers sufficient time to prepare for manual takeover, thus achieving proactive risk prevention and control.

[0047] 5. Dual-layer adaptive optimization with excellent adaptability: This invention has dual adaptability capabilities: real-time vehicle-side correction and cloud-based self-learning updates. On the one hand, the dynamic boundary correction module temporarily corrects the static ODD boundary based on real-time on-site conditions to adapt to short-term sudden scene changes. On the other hand, the dynamic ODD update unit collects ODD edge scene operation data over a long period of time, and then combines it with historical ODD edge scene operation data to obtain adjustment strategies. Based on the adjustment strategies, the basic static ODD boundary, response threshold, and various element judgment thresholds are optimized, and the entire system parameters are synchronously updated through the OTA upgrade module. This allows it to adapt to different roads, different climates, and different vehicle usage scenarios, reasonably expanding the usable operating range of autonomous driving while adhering to the bottom line of safety.

[0048] 6. Enhanced data interaction and collaboration, resulting in stronger scenario adaptability: This invention integrates ODD edge scene operation data into the ODD boundary correction process, and merges all-domain traffic interaction data such as vehicle-to-vehicle, vehicle-to-infrastructure, vehicle-to-pedestrian, and intersection facilities. This compensates for the limitations of single-vehicle perception field of view, improves the accuracy of ODD boundary determination and driving risk prediction in complex intersections, dense traffic sections, and sudden road conditions, and adapts to intelligent transportation collaborative operation scenarios.

[0049] 7. Compliant and standardized architecture with high practicality: This invention complies with the latest national standards for Level 3 autonomous driving, especially the requirements for ODD monitoring and minimum risk strategies. The system architecture is mature and stable, making it easy to install in real vehicles and apply in mass production. Attached Figure Description

[0050] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0051] The technical solution of the present invention will be further described below with reference to the accompanying drawings:

[0052] like Figure 1 As shown, an autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response includes a multi-dimensional detection unit 100, a deviation evaluation unit 200, a spatiotemporal edge prediction unit 300, a response decision unit 400, a safety vehicle control execution unit 500, and a dynamic ODD update unit 600.

[0053] The multi-dimensional detection unit 100 is used to collect multi-dimensional ODD elements, which include environmental ODD elements, road ODD elements, system ODD elements, and human-machine ODD elements, and sends the multi-dimensional ODD elements to the deviation evaluation unit 200.

[0054] The dynamic ODD update unit 600 is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element based on the ODD edge scene data it collects in real time, thereby obtaining the updated static ODD boundary information, response threshold, and threshold corresponding to each ODD element. The updated static ODD boundary information is the current static ODD boundary information, the updated response threshold is the current response threshold, and the updated threshold corresponding to each ODD element is the current threshold corresponding to each ODD element. The dynamic ODD update unit 600 sends the current static ODD boundary information to the spatiotemporal edge prediction unit 300, sends the current response threshold to the response decision unit 400, and sends the current threshold corresponding to each ODD element to the deviation evaluation unit 200.

[0055] The deviation evaluation unit 200 is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element through the dual logic of "threshold comparison + grade quantification". Then, combined with the element weight coefficient and the combined influence coefficient, the comprehensive deviation is finally calculated and sent to the spatiotemporal edge prediction unit 300 and the response decision unit 400.

[0056] The spatiotemporal edge prediction unit 300 is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information. Then, it extracts the static ODD boundary information on the predicted future driving path from the map, and then corrects the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data to obtain the dynamic ODD boundary. Then, it calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the real-time vehicle status, and sends the estimated time to the response decision unit 400.

[0057] The spatiotemporal edge prediction unit combines comprehensive deviation, real-time vehicle operating status including current vehicle position, speed, and heading angle, navigation planning information, and dynamically updated static ODD boundary information to predict the vehicle's future path. It then extracts static ODD boundary information from the high-precision map, covering points such as road type change points, tunnel entrances, toll stations, and construction areas along the predicted path. Simultaneously, relying on real-time environmental change information and actual traffic conditions, it dynamically corrects the established static ODD boundary information, flexibly adjusting the safe operating domain. For example, when a sudden traffic accident is detected ahead, the accident area is directly marked as a non-ODD area, thus generating a dynamic ODD boundary adapted to the on-site conditions. Finally, combining the vehicle's real-time driving status and the dynamic ODD boundary, it accurately calculates the estimated time for the vehicle to reach the dynamic ODD boundary and transmits this estimated time to the response decision unit, providing a time-dimensional basis for subsequent tiered response strategy formulation.

[0058] The response decision unit 400 is used to generate a graded response control instruction based on the current response threshold, the comprehensive deviation, and the estimated time, and send the graded response control instruction to the safety vehicle control execution unit 500; the current response threshold is the comprehensive deviation boundary threshold for the system to classify risk levels.

[0059] ①L0 (normal operation) response control command: The overall deviation is ≤0.2, and the system normally executes the L3 level automatic driving function;

[0060] ②L1 (Warning and Reminder) Response Control Command: 0.2 < Overall Deviation ≤ 0.4, issues a mild warning and reminder to the driver to pay attention to changes in the environment;

[0061] ③L2 (Function Degradation) Response Control Command: 0.4 < Comprehensive Deviation ≤ 0.6, the system automatically degrades to Enhanced L2 Level Assisted Driving, limits the maximum vehicle speed, and disables the automatic lane change function;

[0062] ④L3 (Forced Takeover) Response Control Command: 0.6 < Overall Deviation ≤ 0.8, expected to enter the dynamic ODD boundary in 15 seconds, the system issues a strong multimodal takeover request (sound, vision, touch), while maintaining vehicle control for 10 seconds.

[0063] ⑤L4 (Minimum Risk Strategy) Response Control Instruction: If the overall deviation is >0.8 or the driver does not take over within the expected time, such as 10 seconds, execute the minimum risk stop strategy (MRM).

[0064] 0.2, 0.4, 0.6, and 0.8 are the current response thresholds.

[0065] The safety vehicle control execution unit 500 is used to perform corresponding vehicle speed adjustment, autonomous driving function downgrade, multimodal human-machine interaction, and minimum risk parking operations according to the graded response control command.

[0066] L0 (normal operation) response control command: The safety vehicle control execution unit 500 normally executes the L3 level autonomous driving function, supporting automatic following, automatic lane changing, and automatic overtaking. Human-machine interaction: The HMI displays "Autonomous driving in progress" and the green indicator light is always on.

[0067] L1 (Warning and Reminder) Response Control Command: The safety vehicle control execution unit 500 continues to execute the L3 level autonomous driving function, but increases the safety margin of perception and decision-making. Human-machine interaction: The HMI displays "Attention to environmental changes", the yellow indicator light flashes, the voice prompt "Attention to road conditions ahead", and the following distance is appropriately increased, and the frequency of automatic lane change is reduced.

[0068] L2 (Function Degradation) Response Control Command: The safety vehicle control execution unit 500 automatically downgrades the vehicle to enhanced L2 level assisted driving. The driver needs to take over the monitoring human-machine interface: The HMI displays "Downgraded to assisted driving, please pay attention to road conditions", limits the maximum speed to 80% of the current road speed limit, disables automatic lane changing and automatic overtaking functions, and only retains ACC and LKA functions.

[0069] L3 (Forced Takeover) Response Control Command: The safety vehicle control execution unit 500 maintains vehicle control for 10 seconds, strongly requesting the driver to take over. Human-Machine Interaction: The HMI displays "Please take over the vehicle immediately," the red indicator light flashes rapidly, the voice prompt "Please take over the vehicle immediately!", the steering wheel vibrates continuously, the vehicle speed is gradually reduced, the hazard lights are activated, and the vehicle maintains its lane.

[0070] L4 (Minimum Risk Strategy) Response Control Command: The safety vehicle control execution unit 500 executes the minimum risk parking strategy. Human-machine interaction: The HMI displays "Emergency parking is being performed," the red indicator light remains on, and the voice prompt "Emergency parking is being performed, please take over the vehicle!" activates the hazard warning lights, and decelerates smoothly at a rate of 0.5g. When the vehicle speed drops below 60km / h, it observes whether the right emergency lane is safe. If the emergency lane is safe, it slowly changes lanes to the emergency lane and continues to decelerate until it comes to a complete stop; it engages the electronic parking brake, unlocks the doors, and automatically dials the emergency rescue number.

[0071] like Figure 1 As shown, the multi-dimensional detection unit 100 includes an environmental dimension detection module 101, a road dimension detection module 102, a system dimension detection module 103, and a human-machine dimension detection module 104. The environmental dimension detection module 101 is used to collect environmental ODD elements, including weather ODD elements, illumination ODD elements, and visibility ODD elements. The road dimension detection module 102 is used to collect road ODD elements, including road type ODD elements, lane line clarity ODD elements, curve curvature ODD elements, and slope ODD elements. The system dimension detection module 103 is used to collect system ODD elements, including sensor health status ODD elements, computing platform status ODD elements, drive-by-wire system status ODD elements, and high-precision map matching degree ODD elements. The human-machine dimension detection module 104 is used to collect human-machine ODD elements, including driver attention status ODD elements, hand position ODD elements, eye gaze ODD elements, seat belt status ODD elements, and whether the driver is in the driver's seat ODD elements.

[0072] The weather ODD elements include rainfall ODD elements, snowfall ODD elements, fog concentration ODD elements, and wind speed ODD elements. The illumination ODD elements include daytime illumination intensity ODD elements, dusk or dawn illumination intensity ODD elements, nighttime illumination intensity with streetlights ODD elements, nighttime illumination intensity without streetlights ODD elements, and backlight intensity ODD elements.

[0073] Road type ODD elements include highway ODD elements, urban expressway ODD elements, urban arterial road ODD elements, and ordinary highway ODD elements.

[0074] Sensor health status ODD elements include camera ODD elements, millimeter-wave radar ODD elements, lidar ODD elements, and GNSS global navigation satellite system positioning status ODD elements.

[0075] The deviation assessment unit 200 includes a deviation calculation module 201, an element weight module 202, a combined influence coefficient calculation module 203, and a comprehensive deviation calculation module 204.

[0076] The deviation calculation module 201 is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element, through the dual logic of "threshold comparison + grade quantification", and send the deviation value of each ODD element to the comprehensive deviation calculation module 204.

[0077] For ODD elements with pre-set grading standards, the deviation of a single element is calculated based on the grading quantification logic. For ODD elements without pre-set grading standards, the deviation of a single element is calculated based on the threshold comparison logic.

[0078] Environmental ODD elements:

[0079] ①Rainfall ODD elements: Level 0 (no rain), Level 1 (light rain ≤ 14.9 mm / h), Level 2 (moderate rain 15-29.9 mm / h), Level 3 (heavy rain ≥ 30 mm / h);

[0080] ② Fog Concentration ODD Element: Level 0 (No Fog), Level 1 (Light Fog, Visibility 500-1000m), Level 2 (Dense Fog, Visibility 200-500m), Level 3 (Dense Fog, Visibility <200m);

[0081] ③ Illumination intensity ODD elements: Level 0 (daytime illumination intensity), Level 1 (dusk / dawn illumination intensity), Level 2 (nighttime illumination intensity with streetlights), Level 3 (nighttime illumination intensity without streetlights), Level 4 (strong backlight illumination intensity).

[0082] Road ODD elements:

[0083] ①Road type ODD elements: Level 0 (expressway ODD elements), Level 1 (urban expressway ODD elements), Level 2 (urban arterial road ODD elements), Level 3 (ordinary road ODD elements);

[0084] ② Lane line clarity ODD element: Level 0 (clear), Level 1 (slight wear), Level 2 (severe wear), Level 3 (no lane lines);

[0085] ③ Curvature radius of curves (ODD element): ≥150m (Level 0), 100-150m (Level 1), 50-100m (Level 2), <50m (Level 3);

[0086] ④ Traffic flow density (ODD) elements: low density (level 0), medium density (level 1), high density (level 2), congestion (level 3);

[0087] System ODD elements:

[0088] ① Sensor health status ODD elements: all normal (level 0), 1 non-critical sensor abnormal (level 1), 1 critical sensor abnormal (level 2), multiple sensors abnormal (level 3);

[0089] ② Global Navigation Satellite System Signal Quality ODD Elements in Sensor Health Status ODD Elements: RTK Fixed Solution (Level 0), RTK Floating Solution (Level 1), Single Point Positioning (Level 2), No Signal (Level 3);

[0090] ③ High-precision map matching degree ODD elements: ≥95% (Level 0), 80-95% (Level 1), 60-80% (Level 2), <60% (Level 3);

[0091] Human-Machine ODD elements:

[0092] ① Driver attention status ODD elements: High concentration (Level 0), Mild distraction (Level 1), Moderate distraction (Level 2), Severe distraction (Level 3);

[0093] ② Hand position ODD elements: both hands on the steering wheel (level 0), one hand on the steering wheel (level 1), both hands off the steering wheel (level 2);

[0094] ③ Seatbelt status ODD elements: Fastened (Level 0), Not Fastened (Level 3);

[0095] ④ Whether the driver is in the driver's seat (ODD element): Yes (Level 0), No (Level 3);

[0096] Formula for calculating deviation of a single element:

[0097] For ODD elements with pre-defined grading standards, the deviation of a single element = real-time grade / maximum grade.

[0098] For example: if the rainfall ODD element is level 2 (moderate rain) and the maximum level is level 3, then the rainfall deviation = 2 / 3 ≈ 0.67;

[0099] For ODD elements without a preset grading standard, the deviation of a single element is calculated based on threshold comparison logic.

[0100] The element weighting module 202 is used to assign a weight coefficient to the impact of each ODD element on system safety, and send the weight coefficient of each ODD element on system safety to the comprehensive deviation calculation module 204; the weight coefficient can be determined based on real vehicle test data;

[0101] For example, some numerical examples of the weighting coefficients are shown in the table below:

[0102]

[0103] The combined influence coefficient calculation module 203 is used to calculate the combined influence coefficient when multiple ODD elements deviate simultaneously, and sends the combined influence coefficient to the comprehensive deviation calculation module 204.

[0104] The calculation method for the combined influence coefficient is as follows:

[0105] When the deviation of n elements is greater than 0.5, the combined influence coefficient = 1 + 0.2 × (n-1).

[0106] For example, if two factors have a deviation greater than 0.5, the combined influence coefficient = 1 + 0.2 × 1 = 1.2;

[0107] Three factors have a deviation greater than 0.5, and the combined influence coefficient is 1 + 0.2 × 2 = 1.4.

[0108] And so on.

[0109] The comprehensive deviation calculation module 204 is used to calculate the comprehensive deviation based on the deviation value of each ODD element, the weight coefficient of each ODD element's impact on system security, and the combined impact coefficient, and then sends the comprehensive deviation to the spatiotemporal edge prediction unit 300 and the response decision unit 400. Comprehensive deviation = Σ(single element deviation × corresponding element weight coefficient) × combined impact coefficient

[0110] like Figure 1 As shown, the spatiotemporal edge prediction unit 300 includes a path prediction module 301, an ODD boundary extraction module 302, a dynamic boundary correction module 303, and a time calculation module 304.

[0111] The path prediction module 301 is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information, and send the future driving path of the vehicle to the ODD boundary extraction module 302.

[0112] The ODD boundary extraction module 302 is used to extract static ODD boundary information from the map on the predicted future driving path of the vehicle.

[0113] The dynamic boundary correction module 303 is used to correct the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data, to obtain the dynamic ODD boundary, and send the dynamic ODD boundary to the time calculation module 304.

[0114] The time calculation module 304 calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the real-time vehicle status, and sends the estimated time to the response decision unit 400.

[0115] Static ODD boundary information includes static ODD boundary information corresponding to lane lines, static ODD boundary information corresponding to road curvature, static ODD boundary information corresponding to elevated road sections, static ODD boundary information corresponding to tunnel sections, and static ODD boundary information corresponding to traffic light sections.

[0116] like Figure 1 As shown, the safety vehicle control execution unit 500 includes a vehicle speed smoothing adjustment module 501, a function shutdown module 502, a multimodal human-machine interaction module 503, and a minimum risk parking execution module 504;

[0117] The vehicle speed smoothing adjustment module 501 is used to perform vehicle speed adjustment according to the graded response control command, and adjust it gradually to avoid sudden acceleration or deceleration.

[0118] The function shutdown module 502 is used to perform autonomous driving function downgrade according to the graded response control command, that is, to gradually shut down the autonomous driving function in a preset order, from high-level function to low-level function.

[0119] The multimodal human-machine interaction module 503 is used to execute multimodal human-machine interaction according to the hierarchical response control command, such as gradually turning off the autonomous driving function in a preset order, from high-level functions to low-level functions, that is, controlling the vehicle to decelerate smoothly, turn on the hazard lights, and safely stop in the emergency lane or the lane itself.

[0120] The minimum risk parking execution module 504 is used to execute the minimum risk parking operation according to the graded response control command.

[0121] like Figure 1 As shown, the dynamic ODD update unit 600 includes an edge scene data acquisition module 601, a cloud big data analysis module 602, an ODD boundary adjustment module 603, and an OTA upgrade module 604.

[0122] The edge scene data acquisition module 601 is used to collect ODD edge scene operation data in real time and send the ODD edge scene operation data to the cloud big data analysis module 602. The ODD edge scene operation data includes data on the area where the vehicle approaches the lane line disappears, data on the area where the road curvature exceeds the limit, data on the elevated entrance and exit, data on the tunnel entrance and exit, data on the traffic signal control boundary, and data on sudden environmental anomalies.

[0123] The cloud-based big data analysis module 602 is used to analyze and obtain static ODD boundary information, response threshold, and adjustment strategy for the threshold corresponding to each ODD element based on real-time ODD edge scene operation data and historical ODD edge scene operation data, and sends the adjustment strategy to the ODD boundary adjustment module 603.

[0124] The ODD boundary adjustment module 603 is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element according to the adjustment strategy, so as to obtain the current static ODD boundary information, current response threshold, and current threshold corresponding to each ODD element, and send the above information to the OTA upgrade module 604; for example, if the system is found to perform well under moderate rain conditions, the rainfall threshold mentioned above can be adjusted from level 1 to level 2.

[0125] The OTA upgrade module 604 sends the current static ODD boundary information to the spatiotemporal edge prediction unit 300, sends the current response threshold to the response decision unit 400, and sends the current threshold corresponding to each ODD element to the deviation evaluation unit 200.

[0126] The system control flow of this invention is as follows:

[0127] 1. System Initialization

[0128] Once the vehicle's autonomous driving system is powered on and initialized, the system loads preset ODD element classification standards, element weight coefficients for each dimension, multi-element combination influence coefficient calculation rules, and comprehensive deviation classification response strategies.

[0129] 2. Real-time acquisition and preprocessing of multi-dimensional data

[0130] Through the environmental dimension detection module, road dimension detection module, system dimension detection module, and human-machine dimension detection module within the multi-dimensional detection unit, the original data of the four-dimensional ODD elements of environment, road, system, and human-machine are collected synchronously and in real time, and uniformly transmitted to the deviation assessment unit.

[0131] 3. Calculate the deviation of single elements by type.

[0132] The deviation assessment unit 200 is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element, through a dual logic of "threshold comparison + grade quantification".

[0133] 4. The element weight module retrieves the safety weight coefficient of each ODD element, the combined influence coefficient calculation module counts the number of elements with a deviation greater than 0.5 and calculates the combined influence coefficient, and the comprehensive deviation calculation module calculates the comprehensive deviation according to the formula Comprehensive Deviation = Σ(Single Element Deviation × Corresponding Weight Coefficient) × Combined Influence Coefficient, and finally sends the comprehensive deviation to the spatiotemporal edge prediction unit 300 and the response decision unit 400.

[0134] 5. Spatiotemporal fusion prediction of ODD edge out-of-bounds duration

[0135] The path prediction module combines comprehensive deviation, real-time vehicle driving status, navigation information, and current static ODD boundary information to predict the vehicle's future driving path; the ODD boundary extraction module extracts the static ODD boundary information corresponding to the predicted path; the dynamic boundary correction module combines real-time environmental data and real-time traffic data to correct and obtain the dynamic ODD boundary; the time calculation module accurately calculates the estimated arrival time of the vehicle entering the non-ODD area based on the dynamic ODD boundary and the vehicle driving status, and sends the estimated time and comprehensive deviation to the response decision unit simultaneously.

[0136] 6. Match graded response levels

[0137] The response decision unit retrieves the current response threshold issued by the dynamic ODD update unit. This current response threshold is the comprehensive deviation boundary threshold for classifying risk levels. Combined with the real-time comprehensive deviation value and the estimated time for the vehicle to arrive at the dynamic ODD boundary, a five-level response level is matched.

[0138] L0 Normal operation, L1 Early warning and reminder, L2 Functional degradation, L3 Forced takeover, L4 Minimum risk strategy.

[0139] 7. Perform tiered vehicle control and human-machine interaction actions.

[0140] Based on the matched hierarchical response control commands, the safety vehicle control execution unit sequentially performs actions such as smooth vehicle speed adjustment, gradual degradation of autonomous driving functions, HMI multimodal driver warning reminders, and automatic minimum risk parking after timeout without intervention through the vehicle speed smoothing adjustment module, function degradation module, multimodal human-machine interaction module, and minimum risk parking execution module.

[0141] 8. Vehicle-mounted terminal collects ODD edge scene data

[0142] The edge scene data acquisition module of the dynamic ODD update unit collects ODD edge scene operation data in real time during vehicle driving, including lane line disappearance areas, road curvature exceeding limits areas, tunnel and elevated road entrances and exits, traffic control areas and sudden abnormal environments, etc. The data is temporarily stored locally and uploaded to the cloud big data platform in real time.

[0143] 9. Cloud-based self-learning dynamic parameter optimization and vehicle-wide deployment

[0144] The cloud-based big data analysis module integrates real-time edge scene data with massive historical edge scene data to generate optimization and adjustment strategies for static ODD boundaries, ODD element thresholds, and comprehensive deviation response thresholds. After the ODD boundary adjustment module completes the parameter update, it pushes the updated current static ODD boundary information, current response threshold, and current threshold of ODD elements to the spatiotemporal edge prediction unit, response decision unit, and deviation evaluation unit through the OTA upgrade module, respectively, to complete the adaptive iterative update of the vehicle system parameters and achieve closed-loop control of the entire process.

[0145] This invention constructs a monitoring and control scheme for the design and operation domain of autonomous driving, based on the integrated monitoring of four dimensions: environment, road, system, and human-machine interaction, coupled with a hierarchical response mechanism. This scheme can accurately assess the vehicle's operating status in real time during normal autonomous driving operations, continuously verifying whether the vehicle's driving conditions are within the legal and compliant ODD (Operational Design Domain) safe operating range. Once a potential risk of the vehicle leaving the design operating domain is detected, multi-level warning prompts and progressive vehicle control strategies can be triggered sequentially according to the severity of the risk. This scheme fully complies with relevant domestic regulations and standards for L3-level autonomous driving mass production access and emergency response, and effectively improves the safety, stability, and driving comfort of high-level autonomous driving vehicles, effectively facilitating the large-scale deployment and commercial application of L3-level autonomous driving technology in real-world scenarios.

[0146] It should be noted that the above examples are merely one specific embodiment of the present invention. Obviously, the present invention is not limited to the above embodiments and many variations are possible. In short, all variations that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should be considered within the scope of protection of this invention.

Claims

1. A multi-dimensional fusion and hierarchical response-based monitoring and control system for autonomous driving ODD (Optical Discharge Domain) is characterized in that, It includes a multi-dimensional detection unit (100), a deviation evaluation unit (200), a spatiotemporal edge prediction unit (300), a response decision unit (400), a safety vehicle control execution unit (500), and a dynamic ODD update unit (600). The multi-dimensional detection unit (100) is used to collect multi-dimensional ODD elements, including environmental ODD elements, road ODD elements, system ODD elements, and human-machine ODD elements, and send the multi-dimensional ODD elements to the deviation evaluation unit (200). The dynamic ODD update unit (600) is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element according to the ODD edge scene data collected in real time, so as to obtain the updated static ODD boundary information, response threshold, and threshold corresponding to each ODD element. The updated static ODD boundary information is the current static ODD boundary information, the updated response threshold is the current response threshold, and the updated threshold corresponding to each ODD element is the current threshold corresponding to each ODD element. The dynamic ODD update unit (600) sends the current static ODD boundary information to the spatiotemporal edge prediction unit (300), the dynamic ODD update unit (600) sends the current response threshold to the response decision unit (400), and the dynamic ODD update unit (600) sends the current threshold corresponding to each ODD element to the deviation evaluation unit (200). The deviation assessment unit (200) is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element through the dual logic of "threshold comparison + grade quantification". Then, combined with the element weight coefficient and the combined influence coefficient, the comprehensive deviation is finally calculated and sent to the spatiotemporal edge prediction unit (300) and the response decision unit (400). The spatiotemporal edge prediction unit (300) is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information. Then, it extracts the static ODD boundary information on the predicted future driving path from the map, and then corrects the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data to obtain the dynamic ODD boundary. Then, it calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the real-time vehicle status, and sends the estimated time to the response decision unit (400). The response decision unit (400) is used to generate a graded response control command based on the current response threshold, comprehensive deviation, and estimated time, and send the graded response control command to the safety vehicle control execution unit (500). The safety vehicle control execution unit (500) is used to perform corresponding vehicle speed adjustment, autonomous driving function downgrade, multimodal human-machine interaction and minimum risk parking operations according to the graded response control command.

2. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 1, characterized in that, The multi-dimensional detection unit (100) includes an environmental dimension detection module (101), a road dimension detection module (102), a system dimension detection module (103), and a human-machine dimension detection module (104). The environmental dimension detection module (101) is used to collect environmental ODD elements, which include weather ODD elements, illumination ODD elements, and visibility ODD elements. The road dimension detection module (102) is used to collect road ODD elements, which include road type ODD elements, lane line clarity ODD elements, curve curvature ODD elements, and slope ODD elements. The system dimension detection module (103) is used to collect system ODD elements, including sensor health status ODD elements, computing platform status ODD elements, drive-by-wire system status ODD elements, and high-precision map matching degree ODD elements. The human-machine dimension detection module (104) is used to collect human-machine ODD elements, including driver attention status ODD elements, hand position ODD elements, eye gaze ODD elements, seat belt status ODD elements, and whether the driver is in the driver's seat ODD elements.

3. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 2, characterized in that, The weather ODD elements include rainfall ODD elements, snowfall ODD elements, fog concentration ODD elements, and wind speed ODD elements. The illumination ODD elements include daytime illumination intensity ODD elements, dusk or dawn illumination intensity ODD elements, nighttime illumination intensity with streetlights ODD elements, nighttime illumination intensity without streetlights ODD elements, and backlight intensity ODD elements.

4. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 2, characterized in that, The road type ODD elements include highway ODD elements, urban expressway ODD elements, urban arterial road ODD elements, and ordinary highway ODD elements.

5. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 2, characterized in that, The sensor health status ODD elements include camera ODD elements, millimeter-wave radar ODD elements, lidar ODD elements, and GNSS global navigation satellite system positioning status ODD elements.

6. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 1, characterized in that, The deviation assessment unit (200) includes a deviation calculation module (201), an element weight module (202), a combined influence coefficient calculation module (203), and a comprehensive deviation calculation module (204). The deviation calculation module (201) is used to calculate the deviation of a single element based on each ODD element, the current threshold corresponding to each ODD element, and the preset grading standard of each ODD element through the dual logic of "threshold comparison + grade quantification", and send the deviation value of each ODD element to the comprehensive deviation calculation module (204). The element weighting module (202) is used to assign a weight coefficient to the impact of each ODD element on system security, and send the weight coefficient of each ODD element on system security to the comprehensive deviation calculation module (204). The combined influence coefficient calculation module (203) is used to calculate the combined influence coefficient when multiple ODD elements deviate simultaneously, and sends the combined influence coefficient to the comprehensive deviation calculation module (204). The comprehensive deviation calculation module (204) is used to calculate the comprehensive deviation based on the deviation value of each ODD element, the weight coefficient of the impact of each ODD element on system security, and the combined impact coefficient, and send the comprehensive deviation to the spatiotemporal edge prediction unit (300) and the response decision unit (400).

7. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 1, characterized in that, The spatiotemporal edge prediction unit (300) includes a path prediction module (301), an ODD boundary extraction module (302), a dynamic boundary correction module (303), and a time calculation module (304). The path prediction module (301) is used to predict the future driving path of the vehicle based on the comprehensive deviation, real-time vehicle status, navigation information, and current static ODD boundary information, and send the future driving path of the vehicle to the ODD boundary extraction module (302). The ODD boundary extraction module (302) is used to extract static ODD boundary information from the map on the predicted future driving path of the vehicle; The dynamic boundary correction module (303) is used to correct the static ODD boundary information on the future driving path of the vehicle based on real-time environmental data and real-time traffic data, to obtain the dynamic ODD boundary, and send the dynamic ODD boundary to the time calculation module (304). The time calculation module (304) calculates the estimated time for the vehicle to reach the dynamic ODD boundary based on the dynamic ODD boundary and the real-time status of the vehicle, and sends the estimated time to the response decision unit (400).

8. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 7, characterized in that, The static ODD boundary information includes static ODD boundary information corresponding to lane lines, static ODD boundary information corresponding to road curvature, static ODD boundary information corresponding to elevated road sections, static ODD boundary information corresponding to tunnel sections, and static ODD boundary information corresponding to traffic light sections.

9. The autonomous driving ODD monitoring and control system with multi-dimensional fusion and hierarchical response as described in claim 1, characterized in that, The safety vehicle control execution unit (500) includes a vehicle speed smoothing adjustment module (501), a function shutdown module (502), a multimodal human-machine interaction module (503), and a minimum risk parking execution module (504). The vehicle speed smoothing adjustment module (501) is used to perform vehicle speed adjustment according to the graded response control command; The function shutdown module (502) is used to perform autonomous driving function downgrade according to the graded response control command; The multimodal human-computer interaction module (503) is used to perform multimodal human-computer interaction according to the hierarchical response control command; The minimum risk parking execution module (504) is used to execute the minimum risk parking operation according to the graded response control command.

10. The multi-dimensional fusion and hierarchical response autonomous driving ODD monitoring and control system according to claim 1, characterized in that, The dynamic ODD update unit (600) includes an edge scene data acquisition module (601), a cloud big data analysis module (602), an ODD boundary adjustment module (603), and an OTA upgrade module (604). The edge scene data acquisition module (601) is used to collect ODD edge scene operation data in real time and send the ODD edge scene operation data to the cloud big data analysis module (602). The ODD edge scene operation data includes vehicle approach lane line disappearance area data, road curvature exceeding limit area data, elevated entrance and exit data, tunnel entrance and exit data, traffic signal control boundary data and sudden environmental anomaly data. The cloud-based big data analysis module (602) is used to analyze and obtain static ODD boundary information, response threshold, and adjustment strategy of threshold corresponding to each ODD element based on real-time ODD edge scene operation data and historical ODD edge scene operation data, and send the adjustment strategy to the ODD boundary adjustment module (603). The ODD boundary adjustment module (603) is used to dynamically adjust the static ODD boundary information, response threshold, and threshold corresponding to each ODD element according to the adjustment strategy, so as to obtain the current static ODD boundary information, current response threshold, and current threshold corresponding to each ODD element, and send the above information to the OTA upgrade module (604). The OTA upgrade module (604) sends the current static ODD boundary information to the spatiotemporal edge prediction unit (300), sends the current response threshold to the response decision unit (400), and sends the current threshold corresponding to each ODD element to the deviation evaluation unit (200).

Images