Park low-speed automatic driving system based on RTK pre-acquisition path

Abstract

The invention relates to a park low-speed automatic driving system based on an RTK pre-acquisition path, which is applied to a structured scene and comprises a sensing positioning module, a decision planning module and a control execution module which are connected with one another, a scene model integrating a predefined path and real-time perception is built through a sensor fusion technology, and RTK is used for collecting the predefined path and providing vehicle state information together with an IMU, a chassis and accessories; the decision planning module predicts a target behavior in the scene model, makes a behavior decision according to a prediction result, control feedback and vehicle state information, and outputs a local trajectory planning result; and the control execution module receives the trajectory planning result, controls a chassis and accessories, and realizes automatic driving of the vehicle on a predefined path. Compared with the prior art, the method has the advantages that the path is acquired in advance through the RTK, the global path planning process can be avoided, the calculation complexity is reduced, and automatic driving under specific conditions is realized.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving, and in particular to a low-speed autonomous driving system for a park based on an RTK pre-collected path. Background Art

[0002] With the rapid development of autonomous driving technology, more and more application scenarios have begun to explore the implementation of autonomous driving systems, especially in closed or semi-closed environments such as parks, schools and airports. These scenarios usually have the characteristics of low vehicle speeds, relatively simple and fixed traffic structures, and controllable numbers of pedestrians and non-motor vehicles. They are one of the ideal scenarios for the application of autonomous driving systems. Introducing autonomous driving systems in these specific environments can significantly reduce labor costs, alleviate traffic congestion and improve operational efficiency. Therefore, the low-speed autonomous driving system for parks has great commercial potential, and has broad application prospects in terminal logistics, unmanned public transportation, unmanned sanitation, etc. It is an important carrier for the realization of higher-level autonomous driving systems.

[0003] However, the current low-speed autonomous driving system in the park relies on GPS positioning with limited accuracy, and the signal has certain delays and errors, and is easily affected by environmental occlusion, which cannot meet the high-precision positioning needs of autonomous driving vehicles; at the same time, although the road environment in the structured park is relatively simple, there are still certain dynamic obstacles and uncertain factors (such as road construction, temporary obstacles, etc.), which requires the low-speed autonomous driving system in the park to have a fast and accurate perception module; moreover, the traditional park autonomous driving system relies on the global path planning algorithm to generate the global path, which is complex in calculation and has high requirements for the system hardware.

[0004] Therefore, with the increasing demand for low-speed autonomous driving systems in the park, how to make the vehicle have low cost, high positioning accuracy and strong reliability at the same time has become a problem that needs to be solved urgently. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a low-speed autonomous driving system for a park based on RTK pre-collected path. By pre-collecting path information by RTK, the global path planning process can be avoided, the calculation complexity can be reduced, and stable and reliable autonomous driving can be achieved under specific conditions.

[0006] The present invention can be implemented through the following technical solutions: a low-speed automatic driving system for a park based on an RTK pre-collected path, applied to a structured scene, comprising a perception and positioning module, a decision-making and planning module, and a control execution module, the three of which are interconnected, the perception and positioning module uses a camera and a laser radar to collect environmental information, and uses sensor fusion technology to build a scene model that integrates a predefined path and real-time perception, and RTK is used for predefined path collection and provides vehicle status information together with IMU, chassis and accessories;

[0007] The decision planning module predicts the target behavior in the scenario model, makes behavior decisions based on the prediction results, control feedback and vehicle status information, and outputs the local trajectory planning results;

[0008] The control execution module receives the trajectory planning result, controls the chassis and accessories of the vehicle, and realizes automatic driving of the vehicle on a predefined path.

[0009] Furthermore, the scenario model of the fusion of the predefined path and real-time perception in the perception and positioning module specifically performs traffic sign and signal light detection, lane line detection, and target (including pedestrians, vehicles, and obstacles) detection on the predefined path through target detection algorithm and sensor fusion technology.

[0010] Furthermore, the vehicle status information in the perception and positioning module is obtained by RTK, IMU, chassis and accessories.

[0011] Furthermore, the sensor fusion technology is a post-fusion technology, which realizes fusion through the Hungarian algorithm and the Kalman filter algorithm.

[0012] Furthermore, the target detection algorithm is a YOLO algorithm.

[0013] Furthermore, the predefined path is a global path pre-collected by RTK, and the path is not generated by a global path planning algorithm.

[0014] Furthermore, the decision planning module predicts behavior based on physical parameters such as speed, acceleration, position, etc. of the target, and adopts a rule-based decision-making method.

[0015] Furthermore, the local trajectory planning algorithm in the decision planning module includes but is not limited to an A* algorithm, a Dijkstra algorithm, and a deep learning algorithm.

[0016] Furthermore, the motion control part in the control execution module adopts PID algorithm for longitudinal control and adopts LQR algorithm for lateral control.

[0017] Furthermore, the chassis and accessories include a wire-controlled brake, a wire-controlled steering gear and a wire-controlled throttle.

[0018] 1. Compared with the prior art, the present invention has the following advantages: RTK can achieve centimeter-level positioning accuracy through dual-frequency differential positioning of ground base stations and vehicle-mounted mobile stations, and is less affected by environmental occlusion, and can provide stable and accurate position information for the automatic driving system. In addition, the predefined path is collected in advance, and the path information is stored in the automatic driving system. The vehicle can match the predefined path in real time through RTK during driving, achieve high-precision path following, and improve the stability and safety of the automatic driving system.

[0019] Second, using RTK to pre-collect paths can avoid the global path planning process, thereby reducing computational complexity and lowering the cost of the autonomous driving system. At the same time, using post-fusion technology as a sensor fusion technology requires low computing power, is relatively intuitive, and can be implemented most quickly in engineering. It can minimize costs and improve stability while ensuring the accuracy of autonomous driving environment perception. Moreover, using the YOLO algorithm as a target detection algorithm can also enable the autonomous driving system to have rapid perception and response capabilities, thereby ensuring the safety of the autonomous driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a schematic diagram of the working process in the embodiment;

[0021] Figure 2 A schematic diagram of the sensor fusion technology of an embodiment;

[0022] Figure 3 Schematic diagram of the application framework of the embodiment. DETAILED DESCRIPTION

[0023] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] Example

[0025] This embodiment applies the above technical solution, such as Figure 1 As shown, a low-speed autonomous driving system for a campus based on an RTK pre-collected path is applied to structured scenes (including clear road boundaries, markings, traffic signs, and fixed pedestrian and vehicle flow directions, etc.), and includes a perception and positioning module, a decision-making planning module, and a control execution module. The three are interconnected. In this embodiment, the modules communicate based on ROS and CAN. The perception and positioning module uses cameras and lidars to collect environmental information, and uses sensor fusion technology to build a scene model that integrates a predefined path with real-time perception. Specifically, the scene model that integrates a predefined path with real-time perception is to perform traffic sign and signal light detection, lane line detection, and target (including pedestrians, vehicles, and obstacles) detection on a predefined path through target detection algorithms and sensor fusion technology. Figure 2This is a schematic diagram of the sensor fusion technology of this embodiment. The YOLO algorithm is used as the target detection algorithm, and the post-fusion technology is used as the sensor fusion technology. The fusion is achieved through the Hungarian algorithm and the Kalman filter algorithm. Specifically, the camera and the lidar are independently processed by their respective perception algorithms, and the final target tracking result is output. The Hungarian algorithm is used for optimal matching, and then the Kalman filter algorithm is used for state update to achieve final fusion. The post-fusion technology has low computing power requirements, is relatively intuitive, and can be implemented most quickly in engineering. RTK is used to collect predefined paths and provide vehicle status information together with IMU, chassis and accessories.

[0026] The decision-making planning module predicts the target behavior in the scenario model, makes behavioral decisions based on the prediction results, control feedback, and vehicle status information, and outputs the local trajectory planning results;

[0027] The control execution module receives the trajectory planning results, controls the chassis and accessories of the vehicle, and realizes automatic driving of the vehicle on a predefined path.

[0028] This embodiment applies the above technical solution, such as Figure 3 As shown, first, the predefined path is collected based on RTK as the global path of the autonomous driving system. During the driving process of the autonomous driving vehicle, the perception and positioning module uses cameras and lidar to collect environmental information, and builds a scene model that integrates the predefined path and real-time perception through sensor fusion technology to perform traffic sign and signal light detection, lane line detection, and target (including pedestrians, vehicles and obstacles) detection; the IMU in the perception and positioning module is composed of an accelerometer and a gyroscope, which can measure the acceleration and angular velocity of an object to help track the object's posture, speed and displacement. At the same time, combined with the data provided by RTK and the chassis and accessories, the vehicle status information can be accurately obtained. The vehicle status information in this embodiment includes position, speed, acceleration, angular acceleration, posture information, etc.

[0029] The decision-making planning module receives the environmental perception information provided by the perception and positioning module, and predicts the behavior of pedestrians and vehicles in the environment based on physical parameters such as speed, acceleration, and position. The advantages of this type of method are fast calculation speed and no training required. In structured scenarios, it can achieve better short-term trajectory prediction results, which helps to further save computing resources. In a relatively simple structured scenario such as a park, this embodiment adopts a rule-based decision-making method to respond to various driving scenarios through preset rules and conditions, make corresponding decisions, and receive control feedback from the control execution module for real-time adjustment. The following are some application scenarios of the rule-based decision-making method in this embodiment:

[0030] (1) Traffic sign and signal light control: The autonomous vehicle responds to the color of the traffic light. For example, the vehicle stops at a red light, goes at a green light, slows down at a yellow light and determines whether to stop; the vehicle speed is adjusted according to the speed limit sign;

[0031] (2) Obstacle avoidance and safe distance maintenance: The autonomous vehicle decelerates or stops based on the distance to the vehicle ahead detected by the LiDAR. When a static obstacle is detected, a detour is triggered;

[0032] (3) Pedestrian avoidance: When a pedestrian is detected entering a preset danger zone, the autonomous vehicle will stop driving until the pedestrian passes safely before continuing to drive.

[0033] After receiving the decision of the autonomous driving vehicle, the decision planning module will perform corresponding local trajectory planning. In this embodiment, the trajectory planning algorithm includes but is not limited to the A* algorithm, the Dijkstra algorithm and the deep learning algorithm.

[0034] The control execution module receives the trajectory planning results and performs corresponding motion control. In this embodiment, the longitudinal control adopts the PID algorithm, and the lateral control adopts the LQR algorithm. At the same time, other controls including but not limited to multi-vehicle coordinated motion control and extreme working condition stability control are performed. Finally, the control execution module outputs control signals to the corresponding chassis and accessories of the autonomous driving vehicle. Specifically, the wire control brake, wire control steering and wire control throttle are controlled to realize the autonomous driving of the vehicle on the predefined path.

[0035] Finally, it should be noted that the above description is only a preferred example of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A low-speed automatic driving system for a park based on an RTK pre-collected path, characterized in that: The system is applied to structured scenes and comprises a perception and positioning module (1), a decision-making and planning module (2) and a control and execution module (3), which are interconnected. The perception and positioning module (1) uses a camera and a laser radar to collect environmental information, and uses sensor fusion technology to build a scene model that integrates a predefined path and real-time perception. RTK is used to collect predefined paths and provide vehicle status information together with an IMU, a chassis and accessories. The decision planning module (2) predicts the target behavior in the scenario model, makes a behavior decision based on the prediction result, control feedback and vehicle state information, and outputs a local trajectory planning result; The control execution module (3) receives the trajectory planning result, controls the chassis and accessories of the vehicle, and realizes automatic driving of the vehicle on a predefined path.

2. According to claim 1, a low-speed automatic driving system for a park based on an RTK pre-collected path is characterized in that: The scenario model of the fusion of the predefined path and the real-time perception in the perception and positioning module (1) specifically performs traffic sign and signal light detection, lane line detection, and target (including pedestrians, vehicles, and obstacles) detection on the predefined path through target detection algorithms and sensor fusion technology.

3. According to claim 1, a low-speed automatic driving system for a park based on an RTK pre-collected path is characterized in that: The vehicle status information in the perception and positioning module (1) is obtained by RTK, IMU, chassis and accessories.

4. According to claim 2, a low-speed automatic driving system for a park based on an RTK pre-collected path is characterized in that: The sensor fusion technology is a post-fusion technology, which realizes fusion through the Hungarian algorithm and the Kalman filter algorithm.

5. According to claim 2, a low-speed automatic driving system for a park based on an RTK pre-collected path is characterized in that: The target detection algorithm is the YOLO algorithm.

6. The low-speed automatic driving system for a park based on an RTK pre-collected path according to claim 2, characterized in that: The predefined path is a global path pre-collected by RTK, and the path is not generated by a global path planning algorithm.

7. The low-speed automatic driving system for a park based on RTK pre-collected path according to claim 1, characterized in that: The decision planning module (2) performs behavior prediction based on physical parameters such as speed, acceleration, position, etc. of the target, and adopts a rule-based decision-making method.

8. The low-speed automatic driving system for a park based on RTK pre-collected path according to claim 1, characterized in that: The local trajectory planning algorithm in the decision planning module (2) includes but is not limited to the A* algorithm, the Dijkstra algorithm and the deep learning algorithm.

9. The low-speed automatic driving system for a park based on RTK pre-collected path according to claim 1, characterized in that: The motion control part in the control execution module (3) adopts a PID algorithm for longitudinal control and adopts an LQR algorithm for lateral control.

10. The low-speed automatic driving system for a park based on RTK pre-collected path according to claim 1, characterized in that: The chassis and accessories include wire-controlled brakes, wire-controlled steering and wire-controlled throttle.

Images