Field unmanned vehicle reliable following system and method based on perception

Abstract

The invention discloses a sensing-based field unmanned vehicle reliable following system and method. The system comprises a vehicle operation sensing system, a tracking target sensing system, a tracking target monitoring system, a tracking target feature extraction system and a target following decision system. The vehicle operation sensing system senses and stores current vehicle driving information in real time; the tracking target sensing system continuously obtains a current vehicle following target and target attribute information; the tracking target monitoring system monitors the driving information of the tracking target vehicle according to the data of the vehicle operation sensing system and the target sensing system, and stores the latest driving data of the tracking target vehicle in real time; the tracking target feature extraction system analyzes the unique feature combination of the following target through a target recognition model according to the data of the vehicle operation sensing system, the tracking target sensing system and the tracking target monitoring system, associates the task number of the current vehicle with the unique feature combination of the following target, and stores the task number and the unique feature combination of the following target in a vehicle-mounted database in real time.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to a perception-based reliable following system and method for unmanned vehicles in the wild. Background Technology

[0002] With the development of the vehicle industry and the advancement of information technology, achieving high-level autonomous driving for unmanned ground vehicles based on intelligent technology has become an important research direction, with broad application prospects in both civilian and specialized fields. Intelligent technologies include perception and positioning technology, decision-making and planning technology, and control execution technology. Perception and positioning technology uses LiDAR, millimeter-wave radar, cameras, etc., to achieve real-time and accurate perception of the vehicle's surrounding environment. Combined with high-precision positioning information, it performs semantic-level feature matching with pre-made high-precision maps to determine the autonomous vehicle's own position and the positions of surrounding objects (obstacles). Decision-making and planning technology includes behavioral decision-making, path planning, trajectory planning, and other techniques used to determine the vehicle's operating strategy. Control execution technology mainly achieves vehicle control through the execution of vehicle dynamics control techniques, supporting the realization of the vehicle's autonomous driving strategy.

[0003] Currently, remote control systems for unmanned ground vehicles include modules for sensing, control, communication, and positioning. Combined with the user interface of a remote control terminal, these systems enable remote control of the unmanned vehicle and target (vehicle, person, or other movable object) following. However, in complex outdoor environments, due to factors such as abundant vegetation, uneven terrain, and large turning radii when following targets, unmanned vehicles may fail to detect when the target disappears. This can lead to the vehicle being unable to perform the following task, forcing operators to manually intervene, increasing their workload and thus enhancing the complexity of the mission. Summary of the Invention

[0004] The main objective of this invention is to propose a perception-based reliable following system and method for special vehicles in the field, aiming to solve the technical problem that the following target may be lost by perception-based unmanned vehicles in complex field environments, and to achieve reliable following of unmanned vehicles in complex field environments.

[0005] To achieve the above objectives, this invention provides a perception-based reliable following system for special vehicles in the field, the system comprising: a vehicle operation perception system, a tracking target perception system, a tracking target monitoring system, a tracking target feature extraction system, and a target following decision system; wherein:

[0006] The vehicle operation perception system consists of positioning equipment (such as high-precision GPS and Beidou positioning system), on-board computer and other hardware and supporting software. It is used to perceive and store the location data, driving speed, driving direction and other driving information of unmanned vehicles in real time, and provide basic data support for subsequent decision-making.

[0007] The target tracking perception system consists of hardware such as lidar, millimeter-wave radar, and cameras, as well as supporting software. It is used to continuously acquire the target being followed and its attribute information (such as color, texture, shape, edges, feature points, etc.), so as to achieve real-time and accurate perception of the environment around the unmanned vehicle and the target being followed.

[0008] The target tracking monitoring system, based on data from the vehicle operation perception system and the target tracking perception system, monitors the target's speed, direction of travel, and other characteristic information, and saves the latest driving data of the target in real time. If historical data exists, it is updated to ensure the real-time nature and accuracy of the data.

[0009] The target feature extraction system analyzes the unique combination of features (such as color, texture, shape, and other attribute information) of the target being tracked through a target feature extraction model (based on deep learning algorithms), and associates these features with the task number of the unmanned vehicle, saving them in real time to the onboard database to support subsequent target recognition and decision-making.

[0010] When an autonomous vehicle detects that its target has been lost, the target-following decision system automatically controls the vehicle to move to the location where the target disappeared and triggers the target-following decision model. This model is trained based on historical data (including input data and decision results), and can quickly select the original tracking target, analyze its direction, and generate driving and following information for the autonomous vehicle, thereby achieving autonomous following. If the system is unable to make an autonomous decision, it will request human intervention to ensure vehicle operational safety.

[0011] Furthermore, to achieve the above objectives, the present invention also provides a perception-based reliable following method for special vehicles in the field, the method comprising the following steps:

[0012] S1: The unmanned vehicle performing the following task continuously acquires vehicle driving information such as location data, speed, and direction of travel using the vehicle operation perception system deployed on the unmanned vehicle.

[0013] S2: The unmanned vehicle performing the following task continuously acquires the target being followed and its attribute information (color features, texture features, shape features, edge features, feature point features, etc.) using the target perception system deployed on the unmanned vehicle.

[0014] S3: The unmanned vehicle performing the following task uses the tracking target monitoring system deployed on the unmanned vehicle. The tracking target monitoring system monitors the vehicle's speed, direction of travel, and other vehicle driving information based on the data from the vehicle operation perception system and the target perception system. The latest data on the tracking target vehicle's driving is saved in real time (and updated if there is historical data).

[0015] S4: The unmanned vehicle performing the following task analyzes the unique feature combination of the following target in real time based on the tracking target feature extraction system deployed on the unmanned vehicle and the vehicle driving information (the target recognition model analyzes and processes attribute information such as color features, texture features, shape features, edge features, feature point features, etc.), associates the unmanned vehicle's task number with the unique feature combination of the following target, and saves it to the vehicle's onboard database in real time.

[0016] S5: When the unmanned vehicle performing the following task detects that the following target has disappeared, it first controls the unmanned vehicle to the location where the following target disappeared, and then triggers the tracking target following decision system deployed by the unmanned vehicle. The tracking target following decision system selects the following target in the original following task.

[0017] S6: Execute the following task based on the target information selected by the target following decision system (the specific steps are the same as S1, S2, S3, and S4).

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. Improve tracking reliability

[0020] By using multi-sensor fusion and deep learning algorithms, this invention can achieve accurate perception and feature extraction of the target being tracked, effectively solving the problem that unmanned vehicles are prone to losing track of the target in complex outdoor environments, and significantly improving the reliability of the tracking task.

[0021] 2. No new hardware required

[0022] This invention makes full use of existing autonomous driving perception systems for unmanned vehicles (such as lidar, millimeter-wave radar, cameras, etc.), without requiring additional hardware equipment, thus reducing the complexity and cost of the system.

[0023] 3. Enhance autonomy

[0024] The target following decision system of the present invention can make a quick decision when the target is lost, re-lock the target and generate driving instructions, reducing the frequency of human intervention and improving the autonomy of unmanned vehicles.

[0025] 4. Adaptable to complex environments

[0026] This invention is specifically optimized for complex outdoor environments (such as overgrown weeds, undulating terrain, and large turning radii of targets), and can effectively cope with various complex scenarios, with broad application prospects. Attached Figure Description

[0027] Figure 1 It is a target feature extraction model.

[0028] Figure 2 It is the operating mechanism of the goal-following decision-making model.

[0029] Figure 3 It is the system process of a target-following decision-making system. Detailed Implementation

[0030] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0031] This invention proposes a perception-based reliable following system and method for special vehicles in the field. Its core lies in solving the technical problem of target loss that may occur with current perception-based unmanned vehicles in complex field environments, thus achieving reliable following of unmanned vehicles in complex environments. To achieve this goal, this invention provides a perception-based reliable following system for special vehicles in the field, such as... Figure 1 , Figure 2 and Figure 3 As shown, the system of the present invention includes: a vehicle operation perception system, a tracking target perception system, a tracking target monitoring system, a tracking target feature extraction system, and a target following decision system. The following is a detailed description of each system and its operation:

[0032] 1. Vehicle Operation Perception System

[0033] The vehicle operation perception system is the foundation of the entire following system. Its main function is to perceive and store real-time data on the unmanned vehicle's location, speed, direction, and other driving information. This system consists of positioning devices (such as high-precision GPS or BeiDou positioning systems), an onboard computer, and supporting software. The positioning devices provide accurate geographic coordinates, while the onboard computer processes and stores this data. Through the vehicle operation perception system, the unmanned vehicle can understand its own driving status in real time, providing fundamental data support for subsequent decision-making.

[0034] For example, in practical applications, the vehicle operation perception system can acquire the real-time location information of the unmanned vehicle through a high-precision GPS module, with an accuracy down to the centimeter level. Simultaneously, it obtains the vehicle's speed and direction of travel through vehicle sensors (such as speed sensors and gyroscopes). This information is transmitted in real time to the onboard computer and stored in the vehicle's onboard database for subsequent analysis and decision-making.

[0035] 2. Target tracking perception system

[0036] The target tracking perception system is a key component for enabling autonomous vehicles to follow other vehicles. Its main function is to continuously acquire information about the target the vehicle is currently following, as well as the target's attribute information. This system consists of hardware such as LiDAR, millimeter-wave radar, and cameras, along with supporting software. These sensors can perceive the vehicle's surrounding environment in real time and acquire attribute information about the target, including its color, texture, shape, edge features, and feature point characteristics.

[0037] For example, in a field environment, lidar can accurately measure the distance and relative position between an unmanned vehicle and the target it is following, millimeter-wave radar can sense the target's speed and direction of movement, and cameras can capture the target's appearance features. Through multi-sensor fusion technology, the target tracking perception system can generate a comprehensive and accurate target perception model, providing data support for subsequent target monitoring and feature extraction.

[0038] 3. Target tracking monitoring system

[0039] The target tracking monitoring system monitors the vehicle's speed, direction of travel, and other driving information based on data from the vehicle operation perception system and the target tracking perception system, and saves the latest data on the target vehicle's movement in real time. If historical data is available, it is updated. By analyzing the relative motion between the unmanned vehicle and the target, the system ensures that the unmanned vehicle can understand the dynamic changes of the target in real time.

[0040] For example, a target tracking and monitoring system can calculate the real-time operating status of the target being followed based on distance and speed data provided by lidar and millimeter-wave radar, combined with the autonomous vehicle's own driving information provided by the vehicle operation perception system. If the target suddenly accelerates or decelerates, the target tracking and monitoring system can promptly detect this change and store the updated data in the vehicle's database so that the subsequent decision-making system can make corresponding adjustments.

[0041] 4. Target Feature Extraction System

[0042] The target tracking feature extraction system is one of the core components for achieving reliable following by unmanned vehicles. Its main function is to analyze data from the vehicle operation perception system, the target tracking perception system, and the target tracking monitoring system to generate a unique combination of features for the target being followed, using a target recognition model. This system can comprehensively analyze attribute information such as color features, texture features, shape features, edge features, and feature point features to generate a unique feature combination used to uniquely identify the target. Furthermore, the system associates the current vehicle's task number with the unique feature combination of the target and saves it to the vehicle's onboard database in real time.

[0043] For example, when the target being tracked is a specific vehicle, the target feature extraction system can acquire the vehicle's color and shape features using a camera, its speed and direction of motion using millimeter-wave radar, and its three-dimensional contour features using lidar. This feature information is then input into the target recognition model and processed by a deep learning algorithm to generate a unique feature combination. This feature combination is stored in an onboard database and associated with the current task number, enabling rapid identification and re-locking of the target if it is lost.

[0044] 5. Goal-following decision-making system

[0045] The target following decision system is the core of the entire following system. Its main function is to automatically control the autonomous vehicle to move to the location where the target disappeared when the autonomous vehicle currently performing the following task detects that the target has disappeared. Then, it triggers the target following decision system to select the target from the original following task. This system operates based on a target following decision model and is trained using historical data (including data input and decision results). It can generate driving and following information for the autonomous vehicle based on the current situation, thereby achieving autonomous following.

[0046] For example, when an autonomous vehicle is driving in a wilderness environment, if it temporarily loses control of its target due to obstructed visibility or other reasons, the target following decision system will first control the vehicle to slow down and move to the location where the target disappeared. Then, the system will extract the feature combination information stored in the system based on the target's characteristics, combine it with the current location information provided by the vehicle's operation perception system, and use machine learning algorithms to quickly identify and re-lock onto the target. If the system cannot make a decision autonomously, it will request human intervention to ensure the safety of vehicle operation.

[0047] Through the coordinated operation of the aforementioned systems, this invention effectively solves the problem of unmanned vehicles losing track of targets in complex field environments, enabling reliable tracking by unmanned vehicles in such conditions. This system not only improves the autonomy and reliability of unmanned vehicles in field environments but also reduces the frequency of human intervention and lowers the complexity of the task.

[0048] Furthermore, to achieve the above objectives, the present invention also provides a perception-based reliable following method for special vehicles in the field, the method comprising the following steps:

[0049] S1: The unmanned vehicle performing the following task continuously acquires vehicle driving information such as location data, speed, and direction of travel using the vehicle operation perception system deployed on the unmanned vehicle.

[0050] S2: The unmanned vehicle performing the following task continuously acquires the target being followed and its attribute information (color features, texture features, shape features, edge features, feature point features, etc.) using the target perception system deployed on the unmanned vehicle.

[0051] S3: The unmanned vehicle performing the following task uses the tracking target monitoring system deployed on the unmanned vehicle. The tracking target monitoring system monitors the vehicle's speed, direction of travel, and other vehicle driving information based on the data from the vehicle operation perception system and the target perception system. The latest data on the tracking target vehicle's driving is saved in real time (and updated if there is historical data).

[0052] S4: The unmanned vehicle performing the following task analyzes the unique feature combination of the following target in real time based on the tracking target feature extraction system deployed on the unmanned vehicle and the vehicle driving information (the target recognition model analyzes and processes attribute information such as color features, texture features, shape features, edge features, feature point features, etc.), associates the unmanned vehicle's task number with the unique feature combination of the following target, and saves it to the vehicle's onboard database in real time.

[0053] S5: When the unmanned vehicle performing the following task detects that the following target has disappeared, it first controls the unmanned vehicle to the location where the following target disappeared, and then triggers the tracking target following decision system deployed by the unmanned vehicle. The tracking target following decision system selects the original following target.

[0054] S6: Execute the following task based on the target information selected by the target following decision system (the specific steps are the same as S1, S2, S3, and S4).

[0055] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A perception-based reliable following system for unmanned vehicles in the wild, characterized in that, The system includes: The vehicle operation perception system is used to perceive and store the vehicle's current location data, speed, and direction of travel in real time. The target tracking perception system is used to continuously acquire information about the target being followed by the vehicle and the target's attributes. The target tracking monitoring system is used to monitor and track the vehicle's driving information, including its speed and direction of travel, based on data from the vehicle operation perception system and the target perception system, and to save the latest driving data of the target vehicle in real time. The target tracking feature extraction system is used to analyze the unique feature combination of the following target based on the data from the vehicle operation perception system, the target tracking perception system, and the target tracking monitoring system, and associate the current vehicle's task number with the unique feature combination of the following target and save it to the vehicle's on-board database in real time. The target following decision system is used to automatically control the unmanned vehicle to the location where the target disappears when the unmanned vehicle currently performing the following task discovers that the target has disappeared. Then, it triggers the target following decision system to select the target from the original following task.

2. The perception-based reliable following system for unmanned vehicles in the wild according to claim 1, characterized in that, The vehicle operation perception system includes positioning, on-board computer hardware, and supporting software systems.

3. The perception-based reliable following system for unmanned vehicles in the wild according to claim 1 or 2, characterized in that, The target tracking perception system includes lidar, millimeter-wave radar, camera hardware, and supporting software system.

4. A perception-based reliable following method for unmanned vehicles in the wild, characterized in that, The method includes the following steps: S1: The unmanned vehicle performing the following task continuously acquires vehicle location data, driving speed, and driving direction information using the vehicle operation perception system; S2: The unmanned vehicle performing the following task continuously acquires information about the target being followed and its attributes using the target perception system; S3: The unmanned vehicle performing the following task monitors the target vehicle's speed and direction of travel based on the data from the vehicle operation perception system and the target perception system, and saves the latest data on the target vehicle's travel in real time. S4: The unmanned vehicle performing the following task analyzes the unique feature combination of the following target in real time based on the target feature extraction system and the vehicle driving information, associates the vehicle's task number with the unique feature combination of the following target, and saves it to the vehicle's on-board database in real time. S5: When the unmanned vehicle performing the following task detects that the following target has disappeared, it first controls the unmanned vehicle to the location where the following target disappeared, and then triggers the tracking target following decision system to select the following target in the original following task; S6: Execute the following task based on the target information selected by the target following decision system.

5. The perception-based reliable following method for unmanned vehicles in the wild according to claim 4, characterized in that, The target feature extraction system extracts and analyzes the combination information of unique features of the target through the target feature extraction model. The target following decision system selects the original target based on the target following decision model, analyzes the direction of the target, and generates the driving following information of the unmanned vehicle.

Images