Rule-based scene library setting automatic driving decision-making system and method, and vehicle

Abstract

The invention discloses a rule-based scene library setting automatic driving decision-making system and method and a vehicle, and belongs to the technical field of automatic driving, and the system comprises a sensor module which is used for collecting environment information around the vehicle; the data processing module is connected with the sensor module, and the data processing module is used for processing and analyzing the environment information collected by the sensor module so as to identify an obstacle and extract feature information of the obstacle; the decision module is used for making an avoidance decision according to the feature information extracted by the data processing module on the basis of a predetermined rule; and the execution module is used for controlling the vehicle to execute the avoidance decision generated by the decision module. According to the invention, various obstacles and moving states thereof can be accurately identified in a complex urban traffic scene, sudden occurrence of the obstacles can be rapidly judged and processed, and decision rules are continuously optimized through real-time learning and experience accumulation so as to adapt to different urban environments and obstacle arrangement.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and more specifically, to a rule-based scenario library autonomous driving decision system, method and vehicle. Background Art

[0002] With the rapid development of autonomous driving technology, how vehicles can achieve safe and efficient decision-making in complex and ever-changing urban environments has become a key issue that needs to be tackled. At this stage, the decision-making process of autonomous driving mainly relies on the real-time processing of sensor data and machine learning algorithms, so that vehicles can recognize obstacles and traffic rules in the surrounding environment to a certain extent. However, when the scene is converted to a complex multi-obstacle scenario, these conventional methods expose the shortcomings of poor decision-making efficiency and questionable safety assurance. For example, in urban road scenarios, vehicles are very likely to need to quickly weigh and avoid multiple pedestrians, many other vehicles and static obstacles, which undoubtedly places strict requirements on the real-time and accuracy of the decision-making algorithm.

[0003] At present, the existing avoidance decision-making methods can be roughly divided into two categories: model-based control strategy and learning-based intelligent decision-making. These methods can indeed play a good role in specific scenarios, but once placed in a dynamic environment with multiple obstacles, it is difficult to break free from the following limitations:

[0004] First, decision-making delays. Given the need to perceive and analyze complex scenarios in all directions, existing methods often inevitably cause decision-making delays, which in turn pose a threat to driving safety.

[0005] Second, the lack of environmental adaptability. Most existing technologies rely on pre-trained models, and once they encounter a completely new and unfamiliar obstacle distribution pattern, they perform poorly.

[0006] Third, it consumes a lot of computing resources. Complex algorithms require a lot of computing resources, which limits their application in real-time systems.

[0007] The patent "Lane Decision Method Based on Multi-Objective Decision Matrix for Autonomous Driving Vehicles" (patent number CN201810383811.2) in the prior art proposes a lane decision method based on a multi-objective decision matrix for autonomous driving vehicles. This method uses the left lane, the lane where the autonomous driving vehicle is located, and the right lane as lanes to be decided. According to the position of the vehicle, the distance that can continue to travel and the speed limit information on each lane to be decided are collected for quantitative evaluation. At the same time, the obstacle information on the front and rear sides of each lane to be decided that is closest to the longitudinal distance of the vehicle is collected, and the distance of the front and rear obstacles relative to the vehicle at the current moment and after the set time is calculated, and the decision matrix is ​​obtained. The decision matrix is ​​processed through the two processes of decision information weight calculation and decision analysis to achieve real-time autonomous lane decision of autonomous driving vehicles.

[0008] However, the shortcomings of the above-mentioned existing technologies are as follows: The above-mentioned existing decision-making methods mainly focus on lane decisions. When facing multiple obstacle scenarios in complex urban environments, the avoidance decisions are not targeted enough, and it is difficult to make avoidance decisions quickly and effectively. On the other hand, although factors such as the relative distance between front and rear obstacles are taken into account, information such as the type and movement status of obstacles is not fully utilized, and the flexibility is poor when dealing with dynamically changing obstacles. In the virtual obstacle hypothesis problem, unreasonable decisions are caused by special circumstances, and the actual road conditions cannot be accurately mapped. Summary of the invention

[0009] One object of the present invention is to provide a rule-based scenario library autonomous driving decision system to solve the aforementioned problem that in urban environments, autonomous driving systems must cope with highly dynamic and complex traffic scenarios, and traditional perception and decision-making systems are often unable to effectively integrate information from multiple sensors, resulting in insufficient obstacle recognition, state prediction and path planning, thereby increasing the risk of collision.

[0010] According to a first aspect of the present invention, there is provided a rule-based automatic driving decision system with a set scenario library, comprising:

[0011] A sensor module, used to collect environmental information around the vehicle;

[0012] A data processing module connected to the sensor module, the data processing module is used to process and analyze the environmental information collected by the sensor module to identify obstacles and extract feature information of the obstacles;

[0013] A decision module connected to the data processing module, wherein the decision module makes an avoidance decision based on a predetermined rule and the feature information extracted by the data processing module;

[0014] An execution module is connected to the decision module, and the execution module controls the vehicle to execute the avoidance decision generated by the decision module.

[0015] Optionally, the sensor module includes a millimeter-wave radar, a lidar and a camera, wherein the millimeter-wave radar is used to detect speed information of the obstacle, the lidar is used to provide three-dimensional spatial information of the obstacle to help determine the exact position and shape of the obstacle, and the camera is used to collect image information of the obstacle.

[0016] Optionally, the data processing module uses a deep neural network model with a multi-layer convolutional neural network structure to perform real-time processing on the images captured by the camera, wherein the position and movement status of pedestrians are determined by detecting human feature points and contours, parked vehicles are identified based on the vehicle's appearance characteristics and color information, and the construction area is judged by identifying construction signs and fence features.

[0017] Optionally, the decision module has a built-in online learning module, which dynamically adjusts the decision rules through an online learning algorithm; the online learning module is equipped with an adaptive learning rate adjustment module, which automatically increases the learning rate when the vehicle enters an unfamiliar scene or detects a sharp increase in road complexity, so as to speed up the initial adjustment speed of the decision rules.

[0018] Optionally, the decision module adopts a graph-based path planning algorithm, performs random sampling in the space around the vehicle according to the multiple obstacle information detected in real time, and expands the path based on the location, size and movement status of the obstacles.

[0019] Optionally, the vehicle continuously senses environmental information around the vehicle and updates characteristic information of obstacles while driving according to the avoidance decision.

[0020] Optionally, the vehicle uploads its own experience data to a cloud server, and obtains experience data of other vehicles from the cloud, wherein the experience data includes obstacle type, obstacle location, processing method, and decision result.

[0021] Optionally, the rule-based scenario library automatic driving decision system also includes an evaluation module, which has several performance evaluation indicators built in. The several performance evaluation indicators include safety indicators, efficiency indicators and comfort indicators. Among them, the safety indicators include the number of collisions and the minimum distance to obstacles, the efficiency indicators include driving time and average speed, and the comfort indicators include acceleration change rate and steering angle change rate. The performance of the automatic driving system in multiple obstacle scenarios is evaluated and verified through several performance evaluation indicators.

[0022] According to a second aspect of the present invention, a rule-based automatic driving decision method for setting a scenario library is provided, comprising:

[0023] Collect environmental information around the vehicle;

[0024] Processing and analyzing the environmental information to identify obstacles and extract characteristic information of the obstacles;

[0025] Based on predetermined rules, making avoidance decisions according to the characteristic information extracted by the data processing module;

[0026] Control the vehicle to execute the avoidance decision generated by the decision module.

[0027] According to a third aspect of the present invention, there is provided a vehicle comprising the above-mentioned rule-based setting scenario library automatic driving decision system.

[0028] The rule-based scenario library automatic driving decision system, method and vehicle disclosed in the present invention have the following technical effects:

[0029] The decision module makes the best avoidance decision based on rule definition, perception technology and learning ability. The rule-based setting scenario library can provide predefined decision rules to quickly and effectively avoid obstacles in specific situations. The decision module has a built-in online learning module, which dynamically adjusts the decision rules through an online learning algorithm; the online learning module is equipped with an adaptive learning rate adjustment module, which automatically increases the learning rate when the vehicle enters an unfamiliar scene or detects a sharp increase in road complexity to speed up the initial adjustment of the decision rules. Not only can it significantly improve the decision-making speed, but it can also improve the adaptability and robustness of the system in complex environments by introducing flexible rule combinations. By building a scenario library covering common obstacle configurations and combining real-time sensor data, autonomous driving vehicles can make reasonable decisions quickly when facing complex scenarios, thereby improving safety and efficiency.

[0030] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0032] Figure 1 A logical schematic diagram of a rule-based scenario library automatic driving decision-making method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0034] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0035] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0036] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0037] The present invention proposes an embodiment of an automatic driving decision system based on a rule-based setting scenario library. Specifically, Figure 1 shown, including:

[0038] A sensor module, used to collect environmental information around the vehicle;

[0039] A data processing module connected to the sensor module, the data processing module is used to process and analyze the environmental information collected by the sensor module to identify obstacles and extract feature information of the obstacles;

[0040] Among them, the data processing module uses a high-performance processor to quickly process and analyze sensor data.

[0041] A decision module connected to the data processing module, wherein the decision module makes an avoidance decision based on a predetermined rule and according to the feature information extracted by the data processing module;

[0042] An execution module is connected to the decision module, and the execution module controls the vehicle to execute the avoidance decision generated by the decision module.

[0043] Specifically, the rule configuration in the present invention is to set a variety of dynamic rules, so that the system can make corresponding avoidance decisions according to the nature of different obstacles. For example, when encountering pedestrians, the system will automatically give priority to pedestrian safety and avoid according to the "pedestrian priority" rule. When the obstacle is a static construction area, the system will evaluate the best detour path to ensure driving safety. This flexibility helps to improve driving safety and passenger comfort.

[0044] The embodiment of the present invention implements independent design for the three core modules of perception, decision-making and execution. Each module is efficiently coordinated on the basis of independent operation with the help of standardized data interfaces and communication protocols. The perception module integrates multiple perception devices such as visual sensors, lidars and ultrasonic sensors. The decision-making module is closely connected with the perception module. Based on the accurate perception data received in real time, it uses built-in advanced algorithms to quickly analyze complex road conditions and instantly generate optimal avoidance path planning for different scenarios. The execution module strictly follows the instructions of the decision-making module and accurately controls every detail of the vehicle's movements through deep integration with the vehicle's acceleration, braking and steering control systems.

[0045] In an embodiment of the present invention, the sensor module includes a millimeter wave radar, a laser radar and a camera, wherein the millimeter wave radar is used to detect the speed information of the obstacle, the laser radar is used to provide the three-dimensional spatial information of the obstacle to help determine the accurate position and shape of the obstacle, and the camera is used to collect image information of the obstacle. The camera can provide rich image information based on the visual sensor, and the data detected by the millimeter wave radar, the laser radar and the camera can improve the accuracy and reliability of obstacle detection, especially in complex urban environments.

[0046] In an embodiment of the present invention, the data processing module uses a deep neural network model with a multi-layer convolutional neural network structure to perform real-time processing on the images collected by the camera, wherein the position and movement status of pedestrians are determined by detecting human feature points and contours, parked vehicles are identified based on the vehicle's appearance features and color information, and the construction area is judged by identifying construction signs and enclosure features.

[0047] In an embodiment of the present invention, the decision module is built with an online learning module, and the online learning module dynamically adjusts the decision rules through an online learning algorithm; the online learning module is equipped with an adaptive learning rate adjustment module, which automatically increases the learning rate when the vehicle enters an unfamiliar scene or detects that the complexity of the road conditions increases sharply, so as to speed up the initial adjustment speed of the decision rules. In the present invention, it is possible to make instant decisions for sudden obstacle scenes, rely on the perception module to continuously update data, and measure the distance between the vehicle and the obstacle, the relative speed, and the surrounding available avoidance space in real time. Once an emergency situation is encountered, such as an animal that suddenly breaks into the lane or a vehicle that changes lanes illegally, the system immediately starts an emergency assessment procedure. If it is determined that there is sufficient avoidance space around, the decision module quickly plans a new safe path, the execution module seamlessly connects, smoothly adjusts the speed and direction, and flexibly interweaves detours to ensure that the journey is not disturbed; if the assessment finds that the space is cramped and difficult to avoid, the system decisively triggers the emergency braking command, and the braking system responds with full force to make the vehicle stable and stationary in the shortest time.

[0048] In an embodiment of the present invention, the decision module adopts a graph-based path planning algorithm, performs random sampling in the space around the vehicle according to the multiple obstacle information detected in real time, and expands the path based on the position, size and movement status of the obstacles.

[0049] Specifically, when a pedestrian is identified, the system of the present invention immediately activates the "pedestrian priority" core rule, not only actively slowing down, but also accurately predicting the movement trajectory of pedestrians, reserving sufficient safe space, and ensuring that pedestrians can pass without worries; if encountering static obstacles such as construction areas, the system quickly switches to detour mode, comprehensively considering factors such as surrounding traffic flow, road width, intersection distribution, etc., and using complex path planning algorithms to calculate the best detour route in real time to ensure driving continuity and safety.

[0050] In the embodiment of the present invention, the vehicle continuously senses the environmental information around the vehicle and updates the characteristic information of the obstacle while driving according to the avoidance decision.

[0051] In an embodiment of the present invention, the vehicle uploads its own experience data to a cloud server and obtains experience data of other vehicles from the cloud, wherein the experience data includes obstacle type, obstacle location, processing method, and decision result. By analyzing and learning the experience data of other vehicles obtained from the cloud, the vehicle can accelerate its own learning speed and improve the accuracy and adaptability of decision-making. For example, when a vehicle encounters a rare obstacle arrangement in a new urban area, it can obtain the experience of other vehicles in similar areas from the cloud, so as to make correct decisions faster.

[0052] Specifically, at each decision moment, the system of the present invention automatically and exhaustively records key data such as obstacle type, overall environment, and final decision path, and builds a decision database. Based on the online learning algorithm, the system periodically back-analyzes past decision performance, rewards successful avoidance and efficient driving cases, and punishes collision risks and inefficient decision trajectories, thereby continuously fine-tuning and optimizing the built-in decision rules.

[0053] In an embodiment of the present invention, when the sensor module detects that an obstacle suddenly appears, the vehicle immediately activates the emergency decision-making mechanism. First, the type, position, speed, direction and other information of the obstacle are quickly determined using sensor information. Then, based on the vehicle's own state and the surrounding environment, it is determined whether there is enough space to avoid the obstacle. If there is space, a new avoidance path is quickly generated based on the current path planning algorithm, and the vehicle's driving direction and speed are adjusted; if there is insufficient space, the vehicle immediately takes braking measures, stops safely, and waits for the obstacle to be removed or for a suitable driving opportunity to appear.

[0054] In an embodiment of the present invention, the vehicle continuously collects data such as environmental information, obstacle characteristics, and decision results during driving. The decision rules are dynamically adjusted and optimized through an online learning algorithm. For example, when a vehicle encounters similar obstacle arrangements multiple times in a specific scenario, the online learning algorithm can adjust the decision strategy for the scenario based on these empirical data to improve the accuracy and efficiency of dealing with similar situations in the future. Among them, a scenario is set such as a specific intersection, where there are always a large number of electric vehicles passing through during rush hours and delivery trucks are often parked on the roadside. The online learning algorithm uses pattern recognition technology to identify common features of similar scenarios, such as the common driving trajectory of electric vehicles at the intersection in a specific time period, the probability position of delivery trucks opening doors to unload goods, etc. Then, based on the preset optimization goals (such as minimizing the risk of collision with obstacles, maximizing driving efficiency, and improving passenger comfort, etc.), the online learning algorithm adjusts the decision rules. For example, the original routine decision at this intersection was to maintain a fixed distance between vehicles and pass at a constant speed. However, it was found that sudden lane changes by electric vehicles caused emergency braking many times. The online learning algorithm was adjusted to slow down in advance and increase the lateral safety distance. At the same time, it collected information such as the status of the electric vehicle's turn signal and the rider's head rotation to predict the details of the lane change intention, and change the vehicle speed control and path fine-tuning strategies.

[0055] In an embodiment of the present invention, the rule-based scenario library automatic driving decision system also includes an evaluation module, which has several performance evaluation indicators built in it, including safety indicators, efficiency indicators and comfort indicators, wherein the safety indicators include the number of collisions and the minimum distance to obstacles, the efficiency indicators include the driving time and the average speed, and the comfort indicators include the acceleration change rate and the steering angle change rate. The performance of the automatic driving system in multiple obstacle scenarios is evaluated and verified by several performance evaluation indicators. In the simulation experiment, virtual reality technology is used to construct various complex urban environments and obstacle arrangements, and the system is tested and optimized on a large scale. In the field test, representative urban roads are selected for testing, actual data is collected and analyzed, and the system parameters and algorithms are continuously adjusted and optimized to improve the system performance and stability.

[0056] The execution module is connected to the vehicle's acceleration system, braking system and steering system by a wired or wireless communication link. Among them, the wired communication link is based on vehicle bus technology, such as CAN (Controller Area Network) bus, FlexRay bus, etc. The wireless communication link adopts vehicle wireless communication protocols, such as Bluetooth Low Energy (BLE), Wi-Fi6E and 5G vehicle networking technology. The execution module communication link not only controls the basic functions of acceleration, braking and steering, but also links the electronic stability program (ESP), adaptive suspension, etc. When cornering at high speed, the execution module instructs steering, and at the same time informs ESP through the communication link to prepare for intervention in advance, and fine-tune the braking force distribution of each wheel to prevent skidding; when encountering undulating road surface, the adaptive suspension is linked to adjust the damping coefficient, and the command is instantly transmitted through the link to ensure smooth driving.

[0057] The present invention also provides an embodiment of a rule-based automatic driving decision method for setting a scenario library, comprising:

[0058] Collect environmental information around the vehicle;

[0059] Processing and analyzing the environmental information to identify obstacles and extract characteristic information of the obstacles;

[0060] Based on predetermined rules, making avoidance decisions according to the characteristic information extracted by the data processing module;

[0061] Control the vehicle to execute the avoidance decision generated by the decision module.

[0062] It should be noted that the execution process and other contents between the above methods are based on the same concept as the system embodiment of the present application, and are methods corresponding to the above-mentioned rule-based setting scenario library automatic driving decision system. All implementation structures in the above-mentioned system embodiment are applicable to the embodiment of the method. Its specific functions and technical effects brought about can be found in the system embodiment part, which will not be repeated here.

[0063] The present invention also provides an embodiment of a vehicle, including a rule-based automatic driving decision system for setting a scenario library. Specifically, when the vehicle of the present invention encounters a special complex scenario, it can reasonably adjust the vehicle speed, maintain a safe distance, and select a suitable avoidance route based on the stored similar scenario experience and rules, so as to effectively deal with these complex and challenging traffic conditions.

[0064] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings, but the above is only a preferred embodiment of the present invention. It should be noted that the technical features involved in the above embodiments and their preferred methods can be reasonably combined and matched into a variety of equivalent schemes by those skilled in the art without departing from or changing the design ideas and technical effects of the present invention; therefore, the present invention is not limited to the scope of implementation shown in the drawings, and all changes made in accordance with the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the specification and drawings, should be within the protection scope of the present invention.

Claims

1. A rule-based scenario library automatic driving decision system, characterized in that: include: A sensor module, used to collect environmental information around the vehicle; A data processing module connected to the sensor module, the data processing module is used to process and analyze the environmental information collected by the sensor module to identify obstacles and extract feature information of the obstacles; A decision module connected to the data processing module, wherein the decision module makes an avoidance decision based on a predetermined rule and the feature information extracted by the data processing module; An execution module is connected to the decision module, and the execution module controls the vehicle to execute the avoidance decision generated by the decision module.

2. The rule-based scenario library automatic driving decision system according to claim 1, characterized in that: The sensor module includes a millimeter-wave radar, a laser radar and a camera, wherein the millimeter-wave radar is used to detect the speed information of the obstacle, the laser radar is used to provide the three-dimensional spatial information of the obstacle to help determine the exact position and shape of the obstacle, and the camera is used to collect image information of the obstacle.

3. The rule-based scenario library automatic driving decision system according to claim 2, characterized in that: The data processing module uses a deep neural network model with a multi-layer convolutional neural network structure to process the images collected by the camera in real time, wherein the position and movement status of pedestrians are determined by detecting human feature points and contours, parked vehicles are identified based on the vehicle's appearance features and color information, and the construction area is judged by identifying construction signs and enclosure features.

4. The rule-based scenario library automatic driving decision system according to claim 1, characterized in that: The decision module has a built-in online learning module, which dynamically adjusts the decision rules through an online learning algorithm; the online learning module is equipped with an adaptive learning rate adjustment module, which automatically increases the learning rate when the vehicle enters an unfamiliar scene or detects that the complexity of the road conditions increases sharply, so as to speed up the initial adjustment speed of the decision rules.

5. The rule-based scenario library automatic driving decision system according to claim 2, characterized in that: The decision module adopts a graph-based path planning algorithm to perform random sampling in the space around the vehicle based on the multiple obstacle information detected in real time, and expand the path based on the location, size and movement status of the obstacles.

6. The rule-based scenario library automatic driving decision system according to claim 5, characterized in that: While the vehicle is driving according to the avoidance decision, it continuously senses the environmental information around the vehicle and updates the characteristic information of the obstacle.

7. The rule-based scenario library automatic driving decision system according to claim 5, characterized in that: The vehicle uploads its own experience data to the cloud server and obtains experience data of other vehicles from the cloud, wherein the experience data includes obstacle type, obstacle location, processing method and decision result.

8. The rule-based scenario library automatic driving decision system according to claim 1, characterized in that: It also includes an evaluation module, which has several performance evaluation indicators built in. The several performance evaluation indicators include safety indicators, efficiency indicators and comfort indicators. Among them, the safety indicators include the number of collisions and the minimum distance to obstacles, the efficiency indicators include the driving time and the average speed, and the comfort indicators include the acceleration change rate and the steering angle change rate. The performance of the autonomous driving system in multiple obstacle scenarios is evaluated and verified through the several performance evaluation indicators.

9. A rule-based automatic driving decision-making method for setting a scenario library, characterized in that: include: Collect environmental information around the vehicle; Processing and analyzing the environmental information to identify obstacles and extract characteristic information of the obstacles; Based on predetermined rules, making avoidance decisions according to the characteristic information extracted by the data processing module; Control the vehicle to execute the avoidance decision generated by the decision module.

10. A vehicle, characterized in that: An autonomous driving decision system comprising a rule-based setting scenario library as described in any one of claims 1 to 8.

Images