Adaptive following method for rideable automatic following cart in complex environment

Abstract

The application is a kind of self-adaptive following method for a rideable automatic following cart in a complex environment, and relates to the technical field of computer processing, which comprises the following steps: S01, obtaining environment perception data and target positioning data through a multi-source perception module, wherein the environment perception data comprises obstacle information and environment characteristic information around the cart, and the target positioning data comprises position information and motion state information of the followed target; S02, performing signal quality evaluation on the target positioning data through a signal interference detection module, identifying and marking the interfered positioning signal, performing filtering processing and / or signal compensation processing on the interfered positioning signal according to the signal quality evaluation result, and generating anti-interference positioning data; S03, calculating the current environment complexity level according to the environment perception data through an environment complexity evaluation module. The application significantly improves the accuracy and reliability of the positioning data in a complex electromagnetic environment, and ensures that the cart can still realize stable following in an environment with serious signal interference.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to an adaptive following method for a rideable, automatically following cargo vehicle in complex environments. Background Technology

[0002] With development, outdoor camping vehicles are also moving towards intelligence, with automatic following, riding, remote control, or multi-functionality becoming the mainstream, in order to meet the diverse needs of users in different scenarios.

[0003] Currently, existing automated guided vehicles primarily rely on single or combined technologies such as visual images or LiDAR to locate and follow targets. Combined technologies are the most common, as exemplified by Chinese Patent No. CN115937810B, published on February 10, 2026, which discloses a sensor fusion method based on binocular camera guidance. This method belongs to the field of intelligent vehicle environmental perception technology. The invention employs a binocular vision-guided fusion method, which can establish the motion trajectory of a target using visual information, thereby predicting its position in three-dimensional space. Furthermore, the invention uses a data-layer fusion method, utilizing visual prediction results to assist LiDAR in defining the detection range, significantly reducing the amount of computational data and lowering the performance requirements of the computing platform. Based on the target detection results obtained from binocular vision, the invention can accurately obtain target depth, expand target detection information, and enhance the robustness of target detection.

[0004] In the prior art, including the aforementioned patents, when vision and lidar are fused, optical interference (such as image overexposure caused by strong light and noise from laser beams intersecting) and spatiotemporal asynchrony of heterogeneous data often lead to problems such as feature misalignment and confidence decay. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention provides an adaptive following method for a rideable, automatically following cargo vehicle in complex environments. The method includes the following steps:

[0006] S01. Obtain environmental perception data and target positioning data through a multi-source perception module. The environmental perception data includes information on obstacles around the cargo vehicle and environmental features. The target positioning data includes the position information and motion state information of the target being followed.

[0007] S02. The signal quality of the target positioning data is evaluated by the signal interference detection module, the interference of positioning signals is identified and marked, and the interference of positioning signals is filtered and / or compensated according to the signal quality evaluation results to generate anti-interference positioning data.

[0008] S03. The environmental complexity assessment module calculates the current environmental complexity level based on the environmental perception data.

[0009] S04. The adaptive following control module generates an adaptive following control command based on the anti-interference positioning data, the current environmental complexity level, and the current working mode of the cargo vehicle. The current working mode includes riding mode and following mode.

[0010] S05. The execution module controls the drive unit and steering unit of the cargo vehicle to perform a following action according to the adaptive following control command;

[0011] Furthermore, the adaptive following function of the cargo vehicle is achieved through the interaction and coordination of the above steps via data or control information.

[0012] Preferably, step S02 includes performing signal-to-noise ratio detection on the target positioning data to determine whether the signal-to-noise ratio of each positioning signal source is lower than a preset threshold.

[0013] When the signal-to-noise ratio is lower than a preset threshold, the corresponding positioning signal source is marked as an interfered signal source;

[0014] Based on the type of interference from the source of the interference signal, a corresponding filtering algorithm is selected to filter the interference location signal.

[0015] The step of selecting the corresponding filtering algorithm based on the interference type of the interfered signal source includes:

[0016] When the interference type is non-line-of-sight propagation interference, the extended Kalman filter algorithm is used to filter the interfered positioning signal;

[0017] When the interference type is interference from wireless signals in the same frequency band, a filtering method combining bandpass filtering and moving average filtering is used to filter the interfered positioning signal.

[0018] When the interference type is multipath interference, the strong tracking extended Kalman filter algorithm is used to filter the interfered positioning signal.

[0019] Preferably, step S02 further includes assigning confidence values ​​to the multi-source positioning signals during the signal quality assessment process. The multi-source positioning signals include ultra-wideband positioning signals, visual positioning signals, and / or inertial navigation positioning signals.

[0020] The fusion weights are determined based on the confidence levels of each positioning signal, and the multi-source positioning signals are weighted and fused to generate fused positioning data.

[0021] When the confidence level of a certain information source is lower than the preset confidence threshold, the fusion weight of that information source is reduced or the information source is removed from the fusion calculation.

[0022] Preferably, step S03 includes extracting environmental feature parameters based on the environmental perception data, the environmental feature parameters including obstacle density, personnel flow density, light intensity and / or signal obstruction degree; inputting the environmental feature parameters into a preset environmental complexity assessment model to calculate the current environmental complexity level, the environmental complexity level including at least low complexity level, medium complexity level and high complexity level.

[0023] Preferably, step S04, which generates adaptive following control instructions based on the environmental complexity level, includes:

[0024] When the environmental complexity level is low, the first following strategy is adopted, using the main positioning information source in the anti-interference positioning data as the following basis, and controlling the cargo vehicle to follow the target at a first following distance and a first following speed.

[0025] When the environmental complexity level is medium complexity level, a second following strategy is adopted, using the multi-source fusion positioning result in the anti-interference positioning data as the following basis, and controlling the cargo vehicle to follow the target at a second following distance and a second following speed, wherein the second following distance is greater than the first following distance;

[0026] When the environmental complexity level is high, a third following strategy is adopted. The multi-source fusion positioning result in the anti-interference positioning data is combined with the obstacle avoidance planning path as the following basis. The cargo vehicle is controlled to follow the target at a third following distance and a third following speed. The third following distance is greater than the second following distance, and the third following speed is less than the second following speed.

[0027] Preferably, step S04, which generates adaptive following control commands based on the current operating mode of the cargo vehicle, includes:

[0028] When the current working mode is riding mode, the system detects whether there is a rider on the cargo vehicle. If there is a rider, the automatic following function is paused and the system switches to manual driving control mode.

[0029] When the current working mode is the follow mode, the position of the target being followed is continuously tracked according to the anti-interference positioning data, and the cargo vehicle is controlled to automatically follow the target being followed.

[0030] It also includes automatically switching from the riding mode to the following mode when it is detected that there is no rider on the cargo vehicle and a follow start command is received in the riding mode;

[0031] In the follow mode, when a rider is detected on the cargo vehicle and a riding start command is received, the system automatically switches from the follow mode to the riding mode.

[0032] Preferably, during the execution of step S05, the obstacle avoidance module is used to detect obstacle information on the vehicle's travel path in real time.

[0033] When an obstacle is detected in the travel path, an obstacle avoidance and following path is planned based on the location information of the obstacle and the location information of the target being followed.

[0034] Control the cargo vehicle to travel along the obstacle avoidance and following path, avoiding obstacles while maintaining the following of the target;

[0035] The planned obstacle avoidance and following path includes:

[0036] Based on the real-time position of the target being followed, predict the target's predicted motion trajectory within a future preset time window;

[0037] A local obstacle map is constructed based on the position and movement of the obstacles;

[0038] In the local obstacle map, the obstacle avoidance and following path is generated by a path planning algorithm that combines the A* algorithm and the artificial potential field method, with the current position of the cargo vehicle as the starting point and the corresponding point on the predicted motion trajectory as the ending point.

[0039] Preferably, the step S01 of acquiring target positioning data through the multi-source sensing module includes:

[0040] The first positioning data of the target is obtained by receiving the positioning signal emitted by the ultra-wideband positioning tag worn on the target by a miniature ultra-wideband positioning base station located on the autonomous following vehicle.

[0041] Image data containing the target being followed is acquired by a visual sensor, and the image data is used for target detection and recognition to obtain the second positioning data of the target;

[0042] The inertial measurement unit detects the motion state data of the cargo vehicle itself, and combines the first positioning data and / or the second positioning data to calculate the trajectory and obtain the third positioning data of the target.

[0043] Preferably, the signal interference detection module in step S02 identifies the interfered positioning signal in the following manner:

[0044] Perform time-domain and frequency-domain analysis on the target positioning signal to extract signal feature parameters;

[0045] The signal feature parameters are compared with a preset normal signal feature template to calculate the degree of signal distortion.

[0046] When the degree of signal distortion exceeds a preset distortion threshold, the corresponding positioning signal is determined to be an interfered signal.

[0047] As a preferred option, it also includes:

[0048] Before step S01, the multi-source sensing module is initialized and verified to check the communication status and data validity of each sensor.

[0049] During the execution of step S05, the following error of the cargo vehicle is monitored in real time. When the following error exceeds a preset error threshold, a following abnormality alarm is triggered and the cargo vehicle is controlled to decelerate and stop.

[0050] The present invention has at least the following beneficial effects:

[0051] 1. This invention employs a signal interference detection module to assess signal quality and perform anti-interference processing on target positioning data. This effectively identifies and suppresses various external signal interferences, including non-line-of-sight propagation interference, co-frequency wireless signal interference, and multipath interference. By utilizing multiple filtering algorithms, such as extended Kalman filtering, a combination of bandpass filtering and moving average filtering, and strong-tracking extended Kalman filtering, it addresses different interference types. Furthermore, by incorporating confidence assessment and weighted fusion mechanisms for multi-source positioning signals, it significantly improves the accuracy and reliability of positioning data in complex electromagnetic environments, ensuring stable following of the cargo vehicle even in environments with severe signal interference.

[0052] 2. This invention calculates the current environmental complexity level through an environmental complexity assessment module and adaptively switches the following strategy according to different complexity levels. In low-complexity environments, a simplified single-source following strategy is used to improve response speed; in medium-complexity environments, a multi-source fusion positioning and following strategy is used to balance accuracy and efficiency; and in high-complexity environments, a fusion obstacle avoidance planning following strategy is used to ensure safety. This hierarchical adaptive mechanism enables the vehicle to maintain good following performance in environments of varying complexity, from open fields to densely populated areas, effectively avoiding the target loss or mis-following problems that easily occur with single following strategies in complex environments.

[0053] 3. This invention achieves intelligent detection and automatic switching between riding and following modes. By detecting whether a rider is on the vehicle and whether a mode switching command has been received, the system can automatically and seamlessly switch between riding and following modes. This satisfies the user's need for cycling as a mode of transportation while also enabling unmanned following transport, greatly expanding the application scenarios and ease of use of the vehicle. Simultaneously, during the following process, the obstacle avoidance module detects obstacles in real time and plans obstacle-avoidance following paths. A path planning algorithm combining the A* algorithm and the artificial potential field method is used, enabling the vehicle to continuously follow the target while avoiding obstacles, further improving the safety and reliability of following in complex environments. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart of an adaptive following method for a rideable, automatically following cargo vehicle in complex environments, provided in Embodiment 1 of the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0058] Example 1

[0059] This embodiment provides an adaptive following method for a rideable, automatically following cargo vehicle in complex environments. The method includes the following steps: Figure 1 As shown:

[0060] S01. Obtain environmental perception data and target positioning data through the multi-source perception module. The environmental perception data includes information on obstacles around the cargo vehicle and environmental features. The target positioning data includes the position information and motion status information of the target being followed.

[0061] Specifically, the system constructs a multi-source sensing module consisting of various types of sensors, including ultra-wideband (UWB) positioning base stations, visual sensors (such as depth cameras or RGB cameras), lidar (LiDAR), and inertial measurement units (IMU), to simultaneously capture two types of complementary key data information. This lays a comprehensive and accurate data foundation for subsequent signal anti-interference processing, environmental complexity assessment, and adaptive following control.

[0062] The first category is environmental perception data, which is primarily collected using LiDAR point cloud data and depth camera visual image data. This type of data includes not only the outlines, distances, and relative speeds of dynamic obstacles (such as pedestrians and other mobile devices) and static obstacles (such as walls, shelves, and trees) within a preset range around the vehicle (e.g., 180° forward and 3-5 meters to the sides), but also a wide range of environmental features reflecting the essential attributes of the current work scenario, such as the degree of openness of the environment, real-time changes in light intensity, the percentage of wireless signal obstruction by buildings or vegetation, and the density of pedestrian traffic in the surrounding area. These multi-dimensional environmental feature parameters will be quantified and input into the subsequent environmental complexity assessment model to accurately calculate whether the current scene belongs to the low, medium, or high complexity level, and also provide raw coordinate data support for the obstacle avoidance module to build a local obstacle map.

[0063] The second category is target localization data, which fully leverages the complementary advantages of multiple sensor sources. On one hand, the system receives pulse signals from the positioning tag worn by the target through an ultra-wideband (UWB) positioning base station, and calculates the target's approximate three-dimensional spatial coordinates using Time Difference of Arrival (TDOA) or Angle of Arrival (AOA) algorithms. On the other hand, the system acquires continuous frame images containing the target through a visual sensor, and uses a pre-trained target detection and recognition model (such as YOLO or DeepSort algorithms) to accurately select and track the target, thereby obtaining the target's precise pixel position in the image coordinate system. Based on this, the system aligns and fuses the visual recognition results with the UWB ranging results in spatial coordinates, thus accurately capturing the real-time three-dimensional spatial position (X, Y, Z axis coordinates) of the target. Meanwhile, the system utilizes the differential changes in position data over continuous time series, combined with the vehicle's own motion state detected by the inertial measurement unit (IMU) (such as vehicle acceleration and angular velocity), to inversely calculate and compensate for the target's relative displacement drift caused by vehicle motion. This allows for accurate acquisition of the target's motion state information, including speed, acceleration, direction angle, and rate of change of heading. This method of acquiring target information, encompassing both absolute spatial position and dynamic motion trends, not only accurately pinpoints the target's current location but also effectively predicts its movement trend within a very short future time window. This provides highly reliable and low-latency data input for the subsequent adaptive following control module to adjust the following distance, optimize the following speed, and achieve intelligent and seamless switching between riding and following modes based on environmental complexity.

[0064] S02. The signal quality of the target positioning data is evaluated by the signal interference detection module, the interference of positioning signals is identified and marked, and the interference of positioning signals is filtered and / or compensated according to the signal quality evaluation results to generate anti-interference positioning data.

[0065] Furthermore, in the above embodiments, the signal-to-noise ratio of the target positioning data is detected to determine whether the signal-to-noise ratio of each positioning signal source is lower than a preset threshold.

[0066] When the signal-to-noise ratio is lower than a preset threshold, the corresponding positioning signal source is marked as an interfered signal source;

[0067] Based on the type of interference from the source of the interference signal, select the corresponding filtering algorithm to filter the interference-affected positioning signal;

[0068] Selecting the appropriate filtering algorithm based on the type of interference from the affected signal source includes:

[0069] When the interference type is non-line-of-sight propagation interference, the extended Kalman filter algorithm is used to filter the interfered positioning signal;

[0070] When the interference type is the same frequency band wireless signal interference, a filtering method combining bandpass filtering and moving average filtering is used to filter the interfered positioning signal;

[0071] When the interference type is multipath interference, the strong tracking extended Kalman filter algorithm is used to filter the interfered positioning signal.

[0072] Specifically, the module first performs a comprehensive signal quality assessment on the target positioning data acquired by the multi-source sensing module, extracts signal feature parameters through time domain analysis and frequency domain analysis, compares these feature parameters with preset normal signal feature templates, calculates the degree of signal distortion, and identifies and marks the interfered positioning signal accordingly.

[0073] Specifically, the signal quality assessment process begins by detecting the signal-to-noise ratio (SNR) of each location signal source. SNR is a core indicator of signal quality; when the SNR of a signal source falls below a preset threshold, it indicates significant interference, and the system automatically marks the source as an affected signal. After marking, the system further identifies the type of interference to determine its specific nature. Different types of signal interference have different statistical characteristics and propagation features, thus requiring the use of appropriate filtering algorithms.

[0074] Regarding interference type identification and filtering algorithm selection, this invention designs differentiated processing schemes for the three most common types of interference in practical applications. Firstly, when the interference type is identified as non-line-of-sight (NLS) propagation interference, meaning there are buildings, walls, or other physical obstacles obstructing the vehicle and the target being followed, the positioning signal cannot propagate in a straight line but reaches the receiver through diffraction or reflection. In this case, the positioning error usually exhibits a biased estimate with certain statistical regularities. To address this situation, this invention employs the Extended Kalman Filter (EKF) algorithm to filter the interfered positioning signal. EKF can optimally estimate the state in a nonlinear system, predicting and updating the target position through the system state equation and observation equation, effectively suppressing positioning errors caused by NLS propagation.

[0075] Secondly, when the interference type is identified as interference from wireless signals operating in the same frequency band, meaning there are electromagnetic signals emitted by other wireless devices (such as Wi-Fi devices, Bluetooth devices, other UWB devices, etc.) operating in the same or adjacent frequency band as the positioning system in the environment, these signals overlap with the positioning signal in the spectrum, causing waveform distortion of the positioning signal. To address this, this invention employs a filtering method combining bandpass filtering and moving average filtering. First, a bandpass filter preserves the effective frequency band components of the positioning signal while filtering out out-of-band interference components. Then, a moving average filter smooths the bandpass-filtered signal, further suppressing residual random noise and interference within the band, thereby restoring the true waveform of the positioning signal.

[0076] Third, when the interference type is identified as multipath interference, meaning that the positioning signal is reflected multiple times by the ground, walls, or other reflectors during propagation, signals from multiple different propagation paths arrive at the receiver simultaneously. These signals superimpose at the receiver, causing interference and resulting in significant oscillations in the positioning measurement near the actual location. Multipath interference is characterized by its non-Gaussian nature and strong time-varying characteristics, making it difficult for conventional filtering algorithms to effectively address. This invention employs a strong tracking extended Kalman filter algorithm to handle this situation. This algorithm introduces a time-varying fading factor on top of the standard extended Kalman filter, enabling real-time adjustment of the filter gain to enhance the tracking ability for abrupt changes and model mismatches. This effectively suppresses positioning fluctuations caused by multipath effects and improves the stability and reliability of positioning data in complex propagation environments.

[0077] S03. The environmental complexity assessment module calculates the current environmental complexity level based on the environmental perception data and extracts environmental feature parameters based on the environmental perception data. The environmental feature parameters include obstacle density, personnel flow density, light intensity and / or signal obstruction degree.

[0078] The environmental characteristic parameters are input into the preset environmental complexity assessment model to calculate the current environmental complexity level. The environmental complexity level includes at least low complexity level, medium complexity level and high complexity level.

[0079] Specifically, the environmental complexity assessment module first extracts environmental feature parameters from the environmental perception data that can reflect the complexity of the environment in multiple dimensions. These environmental characteristic parameters preferably include, but are not limited to, the following four key indicators: First, obstacle density, which refers to the number and density of static obstacles (such as walls, shelves, trees, etc.) and dynamic obstacles (such as other mobile devices, pedestrians, etc.) within a preset range around the vehicle (e.g., a circular area with a radius of 5 to 15 meters centered on the vehicle). This parameter directly reflects the complexity of the physical constraints that the vehicle needs to avoid during the following process. Second, personnel flow density, which refers to the number of people moving within a preset range around the vehicle and their activity level. This parameter reflects the intensity of random interference from dynamic non-cooperative targets on the following path. Third, light intensity, which refers to the visible light intensity value in the environment. This parameter directly affects the imaging quality of the visual sensor, and thus the reliability of visual positioning and target recognition. Fourth, signal obstruction degree, which refers to whether there are buildings, metal structures, or other obstructions on the line-of-sight path between the vehicle and the target being followed, and the degree of attenuation of wireless positioning signals such as ultra-wideband signals and Bluetooth signals by the obstructions. This parameter directly reflects the risk of wireless positioning signals being affected by non-line-of-sight propagation.

[0080] After extracting the aforementioned environmental feature parameters, the environmental complexity assessment module inputs these multidimensional feature parameters as vectors into a pre-built environmental complexity assessment model. This assessment model can be a machine learning classification model (such as support vector machines, random forests, or lightweight gradient boosting trees) pre-trained based on a large amount of environmental data from different scenarios, or it can be a multi-factor weighted evaluation model built based on fuzzy logic reasoning or analytic hierarchy process. Its core function is to establish a mapping relationship between multidimensional environmental feature parameters and environmental complexity levels. After calculation by the model, the environmental complexity level corresponding to the current environment is output. In a preferred embodiment of the present invention, the environmental complexity level is divided into at least three progressive levels: low complexity level, indicating that the current environment is open, with few obstacles, sparse personnel flow, good lighting conditions, and slight signal obstruction; at this level, the vehicle can adopt a simplified following strategy to pursue higher following efficiency. Medium complexity level, indicating that the current environment has a certain number of obstacles and personnel flow, fluctuating lighting conditions, or moderate signal obstruction; at this level, the vehicle needs to adopt a more robust multi-source fusion positioning and following strategy. High complexity level, indicating that the current environment has dense obstacles, congested personnel flow, poor lighting conditions, or severe signal obstruction; at this level, the vehicle needs to adopt the most conservative following strategy based on fusion obstacle avoidance planning to ensure safety. Through the quantitative assessment of the above environmental complexity levels, the system can perceive environmental changes in real time and adaptively adjust the following control parameters accordingly, thereby ensuring that the vehicle can achieve stable, safe, and efficient automatic following in operating environments of different complexity levels.

[0081] S04. The adaptive following control module generates adaptive following control commands based on anti-interference positioning data, current environmental complexity level, and the current working mode of the cargo vehicle. The current working mode includes riding mode and following mode.

[0082] Furthermore, the above embodiments generate adaptive following control commands based on the environmental complexity level, including:

[0083] When the environmental complexity level is low, the first following strategy is adopted, using the main positioning information source in the anti-interference positioning data as the following basis, and controlling the cargo vehicle to follow the target at the first following distance and the first following speed.

[0084] When the environmental complexity level is medium, the second following strategy is adopted. The multi-source fusion positioning result in the anti-interference positioning data is used as the following basis, and the cargo vehicle is controlled to follow the target at the second following distance and the second following speed. The second following distance is greater than the first following distance.

[0085] When the environmental complexity level is high, the third following strategy is adopted. The multi-source fusion positioning results in the anti-interference positioning data are combined with the obstacle avoidance planning path as the following basis. The cargo vehicle is controlled to follow the target at the third following distance and the third following speed. The third following distance is greater than the second following distance and the third following speed is less than the second following speed.

[0086] Specifically, the system pre-classifies the current environment into low-complexity, medium-complexity, and high-complexity levels. When the current environment is determined to be of low complexity—for example, in an open square, a sparsely populated park road, or a flat area without significant obstacles—the system adopts a first following strategy, which uses the primary positioning source (preferably an ultra-wideband positioning signal) in the generated anti-interference positioning data as the main following basis. Under this strategy, the cargo vehicle follows the target at a preset first following distance (preferably 1.0–1.5 meters) and a first following speed (preferably 2.0–3.0 meters per second), prioritizing the response speed and maneuverability of the following. When the current environment is determined to be of medium complexity—for example, in a campus or community road environment with moderate population density and some static obstacles—the system activates a second following strategy. This strategy uses the multi-source fusion positioning results from the anti-interference positioning data as the following basis (e.g., weighted fusion of ultra-wideband positioning data with visual positioning data and inertial navigation data). Under this strategy, the vehicle follows the target at a second following distance (preferably 1.8–2.5 meters) and a second following speed (preferably 1.2–1.8 meters per second). The second following distance is greater than the first following distance, and the second following speed is less than the first following speed. Higher positioning reliability and following safety are achieved by appropriately increasing the following latency and decreasing the following speed. When the current environment is determined to be of high complexity—for example, in densely populated areas, complex indoor environments with dynamic moving obstacles, or urban canyon environments with dense buildings and severe signal obstruction—the system adopts a third following strategy. This strategy uses the multi-source fusion positioning results from the anti-interference positioning data as the basis, combined with the obstacle avoidance path planned in real time by the obstacle avoidance module as the comprehensive following basis. Under this strategy, the cargo vehicle follows the target at a third following distance (preferably 2.8 to 3.5 meters) and a third following speed (preferably 0.5 to 1.0 meters per second), wherein the third following distance is greater than the second following distance and the third following speed is less than the second following speed. By further increasing the following delay and decreasing the following speed, and in combination with obstacle avoidance path planning, it is ensured that the cargo vehicle can maintain stable following of the target and effectively avoid various obstacles in complex environments.

[0087] Regarding the smooth transition of strategy switching, preferably, when the environmental complexity level changes between adjacent levels, the system adopts a gradual control method based on fuzzy logic—that is, the following distance and following speed are gradually adjusted between the set values ​​corresponding to each strategy according to a preset gradual slope, rather than abruptly switching, so as to prevent the cargo vehicle from jerking or shaking due to the instantaneous change of control parameters, thereby improving the smoothness of the following process and the user experience.

[0088] Preferably, the aforementioned following speed refers to the relative speed of the vehicle to the target in the direction of the target's travel, in order to avoid collisions caused by the vehicle's continuous movement when the target is stationary. The specific values ​​of the following distance and following speed corresponding to each level can be customized by the user through the human-computer interaction interface according to the actual scenario, and the system can self-learn and dynamically optimize the optimal parameters for each level based on historical following data.

[0089] Meanwhile, regarding the adaptation of the cargo vehicle's current operating mode, the system determines whether it is in riding mode or follow mode by detecting the presence of a rider on the vehicle in real time (e.g., through seat pressure sensors or infrared sensors). When the system recognizes that it is in riding mode and detects the presence of a rider, it pauses the automatic follow function and switches to manual driving control mode to ensure the rider has complete control over the cargo vehicle. When the system recognizes that it is in follow mode, it automatically activates the aforementioned adaptive follow control logic based on the environmental complexity level and continuously tracks the target being followed according to the corresponding follow strategy. In addition, the system also supports an automatic switching mechanism between riding mode and follow mode—in riding mode, when the system detects that there is no rider on the cargo vehicle and simultaneously receives a follow start command from the user, it automatically switches from riding mode to follow mode; in follow mode, when the system detects that there is a rider on the cargo vehicle and simultaneously receives a riding start command from the user, it automatically switches from follow mode to riding mode, thus achieving seamless and intelligent connection between the two operating modes and meeting the user's flexible switching needs in diverse usage scenarios.

[0090] S05. The execution module controls the drive and steering devices of the cargo vehicle to perform following actions according to the adaptive following control command.

[0091] Furthermore, the adaptive following function of the cargo vehicle is achieved through the interaction and coordination of the above steps via data or control information.

[0092] Furthermore, the above embodiments also include, during the execution of step S05, real-time detection of obstacle information on the travel path of the cargo vehicle by the obstacle avoidance module;

[0093] When an obstacle is detected in the travel path, an obstacle avoidance and following path is planned based on the location information of the obstacle and the location information of the target being followed.

[0094] Control the cargo vehicle to travel along the obstacle avoidance and following path, avoiding obstacles while maintaining follow of the target being followed;

[0095] The planned obstacle avoidance and following path includes:

[0096] Predict the target's trajectory within a preset time window based on the target's real-time position.

[0097] Construct a local obstacle map based on the location and movement of obstacles;

[0098] In the local obstacle map, an obstacle avoidance and following path is generated using a path planning algorithm that combines the A* algorithm and the artificial potential field method, with the current position of the cargo vehicle as the starting point and the corresponding point on the predicted motion trajectory as the ending point.

[0099] Specifically, during the execution of step S05, an obstacle avoidance module also detects obstacle information on the vehicle's travel path in real time. This obstacle avoidance module can use lidar, ultrasonic sensors, millimeter-wave radar, depth cameras, or a combination of these sensors to continuously scan the travel area in front of the vehicle and acquire information such as the position, size, direction of movement, and speed of obstacles in real time. The detection range of the obstacle avoidance module can be adaptively adjusted according to the current travel speed of the vehicle. When the vehicle speed is high, the detection range is expanded accordingly to allow the system more sufficient obstacle avoidance response time; when the vehicle speed is low, the detection range can be appropriately reduced to reduce the computational burden on the system.

[0100] When the obstacle avoidance module detects an obstacle in the current path of the cargo vehicle, the system enters obstacle avoidance and following mode. In this mode, the system does not simply control the cargo vehicle to stop or go around the obstacle and then resume following. Instead, it comprehensively considers the positional information of the obstacle and the real-time positional information of the target being followed, and dynamically plans an obstacle avoidance and following path that can safely avoid obstacles and continuously maintain the following of the target. Specifically, the system first predicts the target's trajectory within a preset time window (e.g., 3 to 5 seconds) based on the target's real-time position and motion state (including speed, direction, acceleration, etc.). This trajectory prediction can be implemented using algorithms such as uniform motion models, uniform acceleration motion models, or Kalman filter prediction models to accurately estimate the target's positional change trend over a period of time. At the same time, the system constructs a local obstacle map in a local area around the cargo vehicle based on the obstacle information detected by the obstacle avoidance module (including the obstacle's static / dynamic attributes, geometric dimensions, current coordinates, and motion vectors). This local obstacle map is centered on the cargo vehicle and covers a preset area in front (e.g., 5 to 15 meters in front and 3 to 5 meters to the left and right). It represents the obstacle's projection position, occupied area, and movement trend on a two-dimensional plane in a rasterized or vectorized map.

[0101] Based on the completion of target trajectory prediction and local obstacle map construction, the system plans an obstacle avoidance and following path within the local obstacle map. This path planning starts at the current position of the vehicle and ends at the corresponding point on the predicted target trajectory (i.e., the following position the vehicle should reach under ideal obstacle-free conditions). A path planning algorithm combining the A algorithm and the artificial potential field method is used to generate the obstacle avoidance and following path. The A algorithm searches for the globally optimal feasible path from the starting point to the ending point in the local obstacle map. This algorithm evaluates the cost of each candidate path node through a heuristic function, enabling it to quickly find a collision-free feasible path in environments with complex obstacle distributions. The artificial potential field method is used to locally optimize and smooth the path generated by the A algorithm. Its basic principle is to construct a gravitational field at the target position (attracting the vehicle towards the target) and a repulsive field at the obstacle positions (pushing the vehicle away to avoid collision). A smooth local motion direction is generated through the vector superposition of the gravitational and repulsive forces. The combination of the two algorithms leverages the robustness and completeness of Algorithm A in global path search, while utilizing the real-time and smoothness advantages of the artificial potential field method in local obstacle avoidance. This generates a safe, smooth, and follow-along path that meets the following requirements. Ultimately, the system controls the vehicle to travel along this path, ensuring it safely avoids various obstacles while maintaining continuous tracking of the target. Throughout the obstacle avoidance process, the system continuously updates the target position and obstacle information and replans the path in real time to cope with the dynamically changing and complex environment.

[0102] Example 2

[0103] Based on the above embodiment one, in this example, step S02 further includes assigning confidence values ​​to the multi-source positioning signals during the signal quality assessment process. The multi-source positioning signals include ultra-wideband positioning signals, visual positioning signals, and / or inertial navigation positioning signals.

[0104] The fusion weights are determined based on the confidence levels of each positioning signal, and the multi-source positioning signals are weighted and fused to generate fused positioning data.

[0105] When the confidence level of a certain information source is lower than the preset confidence threshold, the fusion weight of that information source is reduced or the information source is removed from the fusion calculation.

[0106] Specifically, in the implementation process, the confidence level is assigned based on the extraction and comprehensive analysis of real-time characteristic parameters of each positioning signal source. For ultra-wideband positioning signals, the system comprehensively evaluates the ranging accuracy and positioning reliability of the source based on parameters such as signal arrival time difference residual, signal-to-noise ratio (SNR), received signal strength indication (RSSI), and signal angle of arrival, combined with the data consistency verification results between the current and historical times. For visual positioning signals, the system determines its positioning reliability based on the detection box confidence score, feature point matching quantity, reprojection error, and target tracking continuity index output by the target detection algorithm. For inertial navigation positioning signals, the system judges the reliability of its short-term attitude and relative displacement based on the zero-velocity correction state, the real-time drift rate of the accelerometer and gyroscope, and the consistency between the integral displacement and the historical position estimation. After normalization, the above characteristic parameters are mapped to confidence values ​​in the range of 0, 10, and 1 through a preset confidence evaluation model. The closer the confidence level is to 1, the higher the accuracy and reliability of the positioning data provided by the source at the current time.

[0107] After assigning confidence levels to each information source, the system employs a dynamic weighted fusion strategy to generate the final fused positioning data. Specifically, the fusion weight of each positioning signal is positively correlated with its current confidence level, typically calculated using a normalized weighted average. This ensures that high-confidence sources dominate the final fused positioning result, while the contribution of low-confidence sources is correspondingly weakened. This dynamic weighted fusion mechanism fully leverages the complementary advantages of multiple sensor sources—ultra-wideband provides high accuracy in open-line-of-sight environments, vision provides intuitive positioning when the target is clearly visible, and inertial navigation provides smooth and continuous displacement information in a short time—allowing the system to adaptively adjust its dependence on each information source under different operating scenarios, thereby significantly improving the accuracy and smoothness of the positioning results.

[0108] Meanwhile, the system also features a strict source protection and isolation mechanism to cope with extremely harsh signal conditions. When the real-time confidence level of a source falls below a preset confidence threshold, it means that the source has essentially lost its effective positioning capability in the current environment. If it is forcibly introduced into the fusion calculation at this time, the large error it carries will severely contaminate the fusion positioning result, leading to a decrease in overall positioning accuracy or even loss of control. Therefore, the system will automatically reduce the fusion weight of the source. When its confidence level remains below the threshold, the source will be temporarily removed from the fusion calculation, achieving rapid isolation of the faulty or heavily disturbed source. This removal and weight reduction protection mechanism ensures that the fused positioning data is always calculated based on the optimal combination of available signals in the current environment, effectively avoiding the negative impact of a single sensor failure on the overall system performance.

[0109] Example 3

[0110] Based on the above embodiment one, in this example, generating adaptive following control instructions according to the current working mode of the cargo vehicle in step S04 includes:

[0111] When the current working mode is riding mode, it detects whether there is a rider on the cargo vehicle. If there is a rider, it pauses the automatic following function and switches to manual driving control mode.

[0112] When the current working mode is follow mode, the position of the target being followed is continuously tracked based on anti-interference positioning data, and the cargo vehicle is controlled to automatically follow the target being followed.

[0113] It also includes automatically switching from riding mode to follow mode when no rider is detected on the cargo vehicle and a follow start command is received in riding mode;

[0114] In follow mode, when a rider is detected on the cargo vehicle and a riding start command is received, the system automatically switches from follow mode to riding mode.

[0115] Specifically, when the system determines that the current operating mode is riding mode, it first uses detection devices such as pressure sensors, infrared sensors, or image recognition modules located on the vehicle's seat or pedals to detect whether a rider is present on the vehicle. If a rider is detected, it means the vehicle is currently being used by a user. At this time, the system automatically pauses the automatic following function and switches control of the vehicle to manual driving mode. In manual driving mode, the vehicle's direction and speed are entirely controlled by the rider using traditional riding controls such as handlebars, brakes, and throttle. The automatic following function is paused or dormant to avoid conflicts between automatic driving commands and manual operation, ensuring riding safety.

[0116] In riding mode, the system continuously monitors the presence of the rider on the delivery vehicle. When the system detects that there is no rider on the delivery vehicle (i.e., the rider has dismounted) and receives a follow start command from the user, the system automatically switches from riding mode to follow mode. The follow start command can be issued by the user via a worn remote control, smartphone app, voice control, or gesture recognition. After switching to follow mode, the system immediately activates the automatic follow function, starting to track the location of the target being followed based on anti-interference positioning data, and controlling the delivery vehicle to automatically follow the user. This function allows the user to walk without any manual operation of the delivery vehicle, greatly improving ease of use.

[0117] Similarly, in follow mode, the system continuously monitors for the presence of a rider on the vehicle. When the system detects a rider and receives a start command from the user, it automatically switches from follow mode to ride mode. The start command can also be issued via remote control, app, voice, or gestures. Upon switching to ride mode, the system immediately pauses the automatic follow function and transfers control to the manual driving control system, allowing the rider to drive the vehicle normally. This feature allows users to simply sit on the vehicle and issue a start command when they need to ride, and the system automatically switches modes without requiring cumbersome settings or adjustments.

[0118] Example 4

[0119] Based on the above embodiment one, the step S01 in this example, which involves acquiring target positioning data through a multi-source sensing module, includes:

[0120] The first positioning data of the target is obtained by receiving the positioning signal emitted by the ultra-wideband positioning tag worn on the target by a miniature ultra-wideband positioning base station located on the autonomous following vehicle.

[0121] The system acquires image data containing the target being followed using a visual sensor, performs target detection and recognition on the image data, and obtains the target's second location data.

[0122] The inertial measurement unit detects the motion state data of the cargo vehicle itself, and combines the first positioning data and / or the second positioning data to calculate the trajectory and obtain the third positioning data of the target.

[0123] Specifically, target positioning data is acquired through a multi-source sensing module, which includes three independent yet complementary positioning methods: ultra-wideband positioning, visual positioning, and inertial navigation positioning, in order to achieve redundant acquisition of target location information and improve reliability.

[0124] First, the initial positioning data of the target is obtained through ultra-wideband (UWB) positioning technology. Specifically, a miniature UWB positioning base station (or UWB positioning base station) is deployed on the automated guided vehicle. This base station communicates with the UWB positioning tag worn on the target being followed. The UWB positioning tag on the target periodically transmits UWB positioning signals at a certain frequency (e.g., 10Hz to 50Hz). After receiving the signal, the UWB positioning base station on the vehicle calculates the distance and orientation of the target relative to the vehicle by measuring the time of arrival (TOA), time difference of arrival (TDOA), or angle of arrival (AOA), thereby obtaining the initial positioning data of the target. Preferably, multiple UWB positioning base stations (e.g., four, respectively located at the four apex corners of the vehicle) can be deployed on the vehicle. By jointly processing the same tag signal from multiple base stations, the positioning accuracy and robustness are further improved.

[0125] Second, secondary localization data of the target is acquired through a visual sensor. Specifically, a visual sensor (e.g., a monocular camera, a binocular camera, or a depth camera) is deployed on the vehicle. This visual sensor acquires image data containing the target being followed at a certain frame rate (e.g., 30fps). After acquiring the image data, the system first uses a target detection algorithm (e.g., YOLO series algorithms, SSD algorithms, or Transformer-based target detection networks) to detect the target in the image, identifying and locating the target's pixel position in the image. Then, combining the camera's intrinsic and extrinsic parameter calibration data, the pixel coordinates of the target in the image are converted into the target's three-dimensional spatial coordinates relative to the vehicle, obtaining the target's secondary localization data. Furthermore, to improve the accuracy and robustness of target recognition, a deep learning model can be used to extract and match the target's appearance features, enabling the vehicle to accurately distinguish the target from other people or objects in a multi-person environment.

[0126] Third, the system uses an inertial measurement unit (IMU) to combine the first two types of positioning data to perform trajectory estimation and obtain the target's third positioning data. Specifically, the IMU (including a three-axis accelerometer and a three-axis gyroscope) is installed on the vehicle to detect its own acceleration and angular velocity in real time. Based on the vehicle's motion data detected by the IMU and its initial position information, the system uses a dead reckoning algorithm to calculate the vehicle's real-time position and attitude. On this basis, the system further combines the first positioning data (UWB positioning data) and / or the second positioning data (visual positioning data) with the vehicle's own pose information to reverse-engineer the target's third positioning data relative to the world coordinate system or relative to the vehicle. The main function of the third positioning data is that when the UWB signal is temporarily lost due to obstruction or interference, or when visual positioning fails due to changes in lighting or target obstruction, the IMU track estimation data can provide reliable positioning supplementation in a short period of time, filling the positioning gap during the failure of a single signal source, thereby ensuring the continuity and stability of positioning data.

[0127] Example 5

[0128] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps:

[0129] The vehicle acquires environmental perception data and target positioning data through a multi-source perception module. The environmental perception data includes information on obstacles around the vehicle and environmental features, while the target positioning data includes the position and motion status information of the target being followed.

[0130] The signal interference detection module assesses the signal quality of the target positioning data, identifies and marks the interfered positioning signals, and performs filtering and / or signal compensation processing on the interfered positioning signals based on the signal quality assessment results to generate anti-interference positioning data.

[0131] The environmental complexity assessment module calculates the current environmental complexity level based on environmental perception data;

[0132] The adaptive following control module generates adaptive following control commands based on anti-interference positioning data, the current environmental complexity level, and the current working mode of the cargo vehicle. The current working mode includes riding mode and following mode.

[0133] The execution module controls the drive and steering devices of the cargo vehicle to perform following actions according to the adaptive following control instructions;

[0134] Furthermore, the adaptive following function of the cargo vehicle is achieved through the interaction and coordination of the above steps via data or control information.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0136] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0137] Example 6

[0138] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps:

[0139] The vehicle acquires environmental perception data and target positioning data through a multi-source perception module. The environmental perception data includes information on obstacles around the vehicle and environmental features, while the target positioning data includes the position and motion status information of the target being followed.

[0140] The signal interference detection module assesses the signal quality of the target positioning data, identifies and marks the interfered positioning signals, and performs filtering and / or signal compensation processing on the interfered positioning signals based on the signal quality assessment results to generate anti-interference positioning data.

[0141] The environmental complexity assessment module calculates the current environmental complexity level based on environmental perception data;

[0142] The adaptive following control module generates adaptive following control commands based on anti-interference positioning data, the current environmental complexity level, and the current working mode of the cargo vehicle. The current working mode includes riding mode and following mode.

[0143] The execution module controls the drive and steering devices of the cargo vehicle to perform following actions according to the adaptive following control instructions;

[0144] Furthermore, the adaptive following function of the cargo vehicle is achieved through the interaction and coordination of the above steps via data or control information.

[0145] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An adaptive following method for a rideable, automatically following cargo vehicle in complex environments, characterized in that, The method includes the following steps: S01. Obtain environmental perception data and target positioning data through a multi-source perception module. The environmental perception data includes information on obstacles around the cargo vehicle and environmental features. The target positioning data includes the position information and motion state information of the target being followed. S02. The signal quality of the target positioning data is evaluated by the signal interference detection module, the interference of positioning signals is identified and marked, and the interference of positioning signals is filtered and / or compensated according to the signal quality evaluation results to generate anti-interference positioning data. S03. The environmental complexity assessment module calculates the current environmental complexity level based on the environmental perception data. S04. The adaptive following control module generates an adaptive following control command based on the anti-interference positioning data, the current environmental complexity level, and the current working mode of the cargo vehicle. The current working mode includes riding mode and following mode. S05. The execution module controls the drive unit and steering unit of the cargo vehicle to perform a following action according to the adaptive following control command; Furthermore, the adaptive following function of the cargo vehicle is achieved through the interaction and coordination of the above steps via data or control information.

2. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, Step S02 includes performing signal-to-noise ratio detection on the target positioning data to determine whether the signal-to-noise ratio of each positioning signal source is lower than a preset threshold. When the signal-to-noise ratio is lower than a preset threshold, the corresponding positioning signal source is marked as an interfered signal source; Based on the type of interference from the source of the interference signal, a corresponding filtering algorithm is selected to filter the interference location signal. The step of selecting the corresponding filtering algorithm based on the interference type of the interfered signal source includes: When the interference type is non-line-of-sight propagation interference, the extended Kalman filter algorithm is used to filter the interfered positioning signal; When the interference type is interference from wireless signals in the same frequency band, a filtering method combining bandpass filtering and moving average filtering is used to filter the interfered positioning signal. When the interference type is multipath interference, the strong tracking extended Kalman filter algorithm is used to filter the interfered positioning signal.

3. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, Step S02 further includes assigning confidence values ​​to the multi-source positioning signals during the signal quality assessment process. The multi-source positioning signals include ultra-wideband positioning signals, visual positioning signals, and / or inertial navigation positioning signals. The fusion weights are determined based on the confidence levels of each positioning signal, and the multi-source positioning signals are weighted and fused to generate fused positioning data. When the confidence level of a certain information source is lower than the preset confidence threshold, the fusion weight of that information source is reduced or the information source is removed from the fusion calculation.

4. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, Step S03 includes extracting environmental feature parameters based on the environmental perception data. The environmental feature parameters include obstacle density, personnel flow density, light intensity, and / or signal obstruction degree. The environmental feature parameters are input into a preset environmental complexity assessment model to calculate the current environmental complexity level. The environmental complexity level includes at least a low complexity level, a medium complexity level, and a high complexity level.

5. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 4, characterized in that, Step S04, which generates adaptive following control instructions based on the environmental complexity level, includes: When the environmental complexity level is low, the first following strategy is adopted, using the main positioning information source in the anti-interference positioning data as the following basis, and controlling the cargo vehicle to follow the target at a first following distance and a first following speed. When the environmental complexity level is medium complexity level, a second following strategy is adopted, using the multi-source fusion positioning result in the anti-interference positioning data as the following basis, and controlling the cargo vehicle to follow the target at a second following distance and a second following speed, wherein the second following distance is greater than the first following distance; When the environmental complexity level is high, a third following strategy is adopted. The multi-source fusion positioning result in the anti-interference positioning data is combined with the obstacle avoidance planning path as the following basis. The cargo vehicle is controlled to follow the target at a third following distance and a third following speed. The third following distance is greater than the second following distance, and the third following speed is less than the second following speed.

6. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, The step S04, which generates adaptive following control commands based on the current operating mode of the cargo vehicle, includes: When the current working mode is riding mode, the system detects whether there is a rider on the cargo vehicle. If there is a rider, the automatic following function is paused and the system switches to manual driving control mode. When the current working mode is the follow mode, the position of the target being followed is continuously tracked according to the anti-interference positioning data, and the cargo vehicle is controlled to automatically follow the target being followed. It also includes automatically switching from the riding mode to the following mode when it is detected that there is no rider on the cargo vehicle and a follow start command is received in the riding mode; In the follow mode, when a rider is detected on the cargo vehicle and a riding start command is received, the system automatically switches from the follow mode to the riding mode.

7. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, It also includes, during the execution of step S05, real-time detection of obstacle information on the vehicle's path by the obstacle avoidance module; When an obstacle is detected in the travel path, an obstacle avoidance and following path is planned based on the location information of the obstacle and the location information of the target being followed. Control the cargo vehicle to travel along the obstacle avoidance and following path, avoiding obstacles while maintaining the following of the target; The planned obstacle avoidance and following path includes: Based on the real-time position of the target being followed, predict the target's predicted motion trajectory within a future preset time window; A local obstacle map is constructed based on the position and movement of the obstacles; In the local obstacle map, the obstacle avoidance and following path is generated by a path planning algorithm that combines the A* algorithm and the artificial potential field method, with the current position of the cargo vehicle as the starting point and the corresponding point on the predicted motion trajectory as the ending point.

8. The adaptive following method for a rideable, automatically following cargo vehicle in complex environments according to claim 1, characterized in that, The step S01, which involves acquiring target localization data through a multi-source sensing module, includes: The first positioning data of the target is obtained by receiving the positioning signal emitted by the ultra-wideband positioning tag worn on the target by a miniature ultra-wideband positioning base station located on the autonomous following vehicle. Image data containing the target being followed is acquired by a visual sensor, and the image data is used for target detection and recognition to obtain the second positioning data of the target; The inertial measurement unit detects the motion state data of the cargo vehicle itself, and combines the first positioning data and / or the second positioning data to calculate the trajectory and obtain the third positioning data of the target.

9. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the steps of the adaptive following method for a rideable, automatically following cargo vehicle in complex environments as described in any one of claims 1-8.

10. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the steps of the adaptive following method for a rideable, automatically following cargo vehicle in complex environments as described in any one of claims 1-8.

Images