Automatic following method of transfer robot

Abstract

The invention discloses an automatic following method of a transfer robot, and relates to the technical field of transfer robots. The automatic following method of the transfer robot comprises the following steps: S1, hardware upgrading, S2, movement system improvement, S3, software algorithm optimization, S4, path planning algorithm, S5, communication and control system optimization, S6, control system optimization, S7, optimization measures in practical application, and S8, man-machine interaction design. The optimized transfer robot can automatically follow the target object, manual intervention is not needed, the transfer efficiency is greatly improved, the robot can quickly respond to movement of the target object, the position and speed of the robot can be adjusted in time, it is ensured that goods can be accurately transferred to the destination in time, and meanwhile the method enhances environmental adaptability and reduces the labor intensity of workers. Through hardware upgrading and software algorithm optimization, the transfer robot can better adapt to different working environments.

Description

Technical Field

[0001] The invention relates to the technical field of transport robots, and in particular to an automatic following method for a transport robot. Background Art

[0002] In the logistics, warehousing and other industries, the application of handling robots is becoming more and more widespread. Traditional handling robots usually require manual operation or preset paths to perform handling tasks, which has some limitations. For example, manual operation is inefficient and labor-intensive, while robots with preset paths lack flexibility and cannot adapt to dynamically changing environments and task requirements.

[0003] In order to solve these problems, handling robots with automatic following functions came into being. However, existing automatic following handling robots still have some shortcomings in practical applications, such as low following accuracy, susceptibility to environmental interference, slow response speed, etc. These problems limit the performance and application scope of handling robots, and further optimization and improvement are needed. Summary of the invention

[0004] The purpose of the present invention is to solve at least one of the technical problems existing in the prior art and to provide a method for automatic following of a transport robot, which can solve the problem of low transport efficiency.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: A method for automatic following of a handling robot, comprising the following methods: S1 hardware upgrade, sensor optimization, and the use of high-precision laser radar, which can more accurately detect the position, distance and shape of the target object. The laser radar has high resolution and fast scanning speed, and can obtain three-dimensional information of the surrounding environment in real time, providing more accurate data for the robot's following;

[0006] Equipped with multiple cameras to achieve all-round visual perception. Cameras at different angles can capture different features of the target object to improve recognition accuracy. At the same time, the camera can also be used to detect obstacles and environmental changes so that the robot can adjust the following strategy in time.

[0007] The S2 motion system is improved, using high-performance motors and drivers to increase the robot's movement speed and acceleration, which allows the robot to respond more quickly to the movement of the target object and maintain close follow-up;

[0008] Install omnidirectional wheels or Mecanum wheels to enable the robot to move flexibly in any direction and adapt to different working environments and following requirements;

[0009] S3 software algorithm optimization, target recognition and tracking algorithm, using deep learning algorithm to identify and track target objects. Through a large amount of training data, the robot can learn to identify different types of target objects and accurately track targets in complex environments.

[0010] Adopt multi-sensor fusion technology to fuse the data of lidar and camera sensors to improve the accuracy and stability of target recognition and tracking;

[0011] S4 path planning algorithm, design an efficient path planning algorithm, so that the robot can follow the target object while avoiding obstacles and other interference factors. The path planning algorithm can adjust the robot's driving path in real time according to the moving speed and direction of the target object and changes in the surrounding environment;

[0012] S5 communication and control system optimization, wireless communication technology, using stable and reliable wireless communication technology to ensure real-time communication between the robot and the control center. Wireless communication allows the control center to understand the status and location of the robot at any time, and remotely control and dispatch it;

[0013] Optimize communication protocols, improve the efficiency and reliability of data transmission, reduce communication delays and data loss, and ensure that the robot can respond to control instructions in a timely manner;

[0014] S6 control system optimization, design of advanced control system, to achieve precise control of the robot, the control system can adjust the robot's movement direction and speed in real time according to the position and speed of the target object, to maintain close follow-up;

[0015] S7 optimization measures in actual application, environmental adaptability, testing and optimization of different working environments to ensure that the robot can work normally in various complex environments. For example, in a dim or smoky environment, the parameters of the sensor can be adjusted to improve its adaptability;

[0016] S8 human-machine interaction design, with a user-friendly human-machine interaction interface, allows users to easily operate and control the robot. The human-machine interaction interface can display the robot's status and location information, as well as the tracking status of the target object, allowing users to understand the robot's working status at any time;

[0017] The addition of voice recognition and voice prompt functions allows users to control the robot through voice commands, which can improve the convenience and efficiency of operation, especially in some situations that require two-handed operation.

[0018] Preferably, the S4 path planning algorithm takes into account the kinematic and dynamic characteristics of the robot to ensure the stability and safety of the robot during the following process.

[0019] Preferably, the S6 control system is optimized and a feedback control mechanism is added to enable the robot to adjust itself according to the actual following situation. For example, if the robot finds that the distance to the target object is too large or too small, it can automatically adjust the speed and direction to achieve the best following effect.

[0020] Preferably, the optimization measures in the actual application of S7 take into account the impact of different terrains and ground conditions on the robot's following, and make corresponding adjustments and optimizations. On uneven ground, the robot can adopt a more stable movement mode to ensure the accuracy of following.

[0021] Preferably, the robot's software and hardware framework is built. In terms of software framework, the innovation and technical advantage of this project lies in the research and functional integration of deep learning, transfer learning, and reinforcement learning intelligent algorithms. It mainly uses visual solutions, supplemented by multimodal fusion solutions of infrared and point cloud information, and then carries out subsequent algorithm research. Therefore, when selecting hardware, the sensor plans to choose depth plus infrared dual cameras and lidar to obtain the three types of data required.

[0022] Preferably, in terms of hardware framework, a basic model of omnidirectional mobility of a Mecanum wheel structure is designed, equipped with an independent suspension structure to enhance the flexibility and stability of the robot. The chassis control system is a main control board based on STM32, and a FreeRTOS real-time operating system is deployed to transmit the status data of the chassis to the host computer main control system. The two systems are connected to communicate through a UART bus.

[0023] Preferably, for the design of the main control embedded system, in view of the intelligent and high real-time characteristics of this project, it is planned to select a powerful edge computing product as the platform for the development of the project's main control embedded system. Edge computing refers to reducing the process of data transfer on the network, and being able to process data directly and promptly at the source, which can make up for the shortcomings of traditional cloud computing with high data transmission overhead, and bring great improvement to real-time performance. This project plans to use Nvidia Jetson nano as the main control unit to design an embedded system that can match the needs. For wireless communication, sending and receiving identification information, and mode switching between the host computer and the main control system, it is necessary to select, design, and build embedded system equipment.

[0024] Preferably, the host computer main control system is designed. This project automatically follows the host computer main control system of the intelligent logistics robot and adopts the ROS2 (Robot Operating System 2) robot operating system. ROS2 is an improved version of ROS. It achieves decentralization while maintaining the same basic concept as ROS, and introduces DDS (Data Distribution Service) data distribution service as a communication mechanism to support more message transmission modes.

[0025] Preferably, the main control system also needs to use a deep learning development framework and dependent libraries. The follower robot needs to choose a framework that supports dynamic graphs and dynamic neural networks. The Pytorch framework is the Python version of Torch, a deep neural network programming framework specifically for GPU acceleration. Torch is a classic tensor library that operates on multi-dimensional matrix data and is widely used in machine learning. Unlike the static calculation graph of Tensorflow, the calculation graph of Pytorch is dynamic and can be changed in real time according to calculation needs. Therefore, this project chooses Pytorch in combination with other neural network inference libraries to deploy deep learning models.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] 1. The method of automatic following of the transport robot improves the transport efficiency. The optimized transport robot can automatically follow the target object without human intervention, which greatly improves the transport efficiency. The robot can quickly respond to the movement of the target object and adjust its position and speed in time to ensure that the goods can be transported to the destination in a timely and accurate manner. At the same time, the method enhances environmental adaptability. Through hardware upgrades and software algorithm optimization, the transport robot can better adapt to different working environments. For example, high-precision sensors can accurately detect target objects in dimly lit or smoky environments. Omnidirectional wheels or Mecanum wheels enable the robot to move flexibly on uneven ground, improving the robot's ability to work in complex environments.

[0028] 2. The automatic following method of the handling robot improves the following accuracy and stability. By adopting deep learning algorithms and multi-sensor fusion technology, the handling robot can more accurately identify and track target objects. At the same time, the optimized path planning algorithm and feedback control mechanism can ensure the stability and safety of the robot during the following process, and reduce the occurrence of following errors and accidents.

[0029] 3. The method of automatic following of the transport robot realizes convenient human-computer interaction, a friendly human-computer interaction interface and voice recognition function, so that users can operate and control the transport robot more conveniently. Users can control the robot through voice commands, which improves the convenience and efficiency of operation, especially in some occasions that require two-handed operation.

[0030] 4. The method of automatic following of the handling robot expands the scope of application. The optimized handling robot has higher performance and reliability and can be applied to more fields and scenarios. For example, in logistics distribution, manufacturing, medical and health industries, the handling robot can automatically follow goods or personnel to achieve efficient handling and distribution tasks, saving labor costs for enterprises and improving production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention is further described below in conjunction with the accompanying drawings and embodiments:

[0032] Figure 1 It is a system design scheme diagram of the present invention; DETAILED DESCRIPTION

[0033] This section will describe in detail the specific embodiments of the present invention. The preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the accompanying drawings is to supplement the description of the text part of the specification with graphics, so that people can intuitively and vividly understand each technical feature and the overall technical solution of the present invention, but it cannot be understood as a limitation on the scope of protection of the present invention.

[0034] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0035] In the description of the present invention, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0036] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0037] The present invention provides a technical solution: a method for automatic following of a handling robot, comprising the following methods: S1 hardware upgrade, sensor optimization, and the use of a high-precision laser radar, which can more accurately detect the position, distance, and shape of a target object. The laser radar has high resolution and fast scanning speed, and can obtain three-dimensional information of the surrounding environment in real time, providing more accurate data for the robot's following;

[0038] Equipped with multiple cameras to achieve all-round visual perception. Cameras at different angles can capture different features of the target object to improve recognition accuracy. At the same time, the camera can also be used to detect obstacles and environmental changes so that the robot can adjust the following strategy in time.

[0039] The S2 motion system is improved, using high-performance motors and drivers to increase the robot's movement speed and acceleration, which allows the robot to respond more quickly to the movement of the target object and maintain close follow-up;

[0040] Install omnidirectional wheels or Mecanum wheels to enable the robot to move flexibly in any direction and adapt to different working environments and following requirements;

[0041] S3 software algorithm optimization, target recognition and tracking algorithm, using deep learning algorithm to identify and track target objects. Through a large amount of training data, the robot can learn to identify different types of target objects and accurately track targets in complex environments.

[0042] Adopt multi-sensor fusion technology to fuse data from lidar, cameras and other sensors to improve the accuracy and stability of target recognition and tracking;

[0043] S4 path planning algorithm, design an efficient path planning algorithm, so that the robot can follow the target object while avoiding obstacles and other interference factors. The path planning algorithm can adjust the robot's driving path in real time according to the moving speed and direction of the target object and changes in the surrounding environment;

[0044] Among them, the kinematic and dynamic characteristics of the robot are taken into consideration to ensure the stability and safety of the robot during the following process. For example, the robot is prevented from turning or braking suddenly when driving at high speed to avoid accidents.

[0045] S5 communication and control system optimization, wireless communication technology, using stable and reliable wireless communication technology to ensure real-time communication between the robot and the control center. Wireless communication allows the control center to understand the status and location of the robot at any time, and remotely control and dispatch it;

[0046] Optimize communication protocols, improve the efficiency and reliability of data transmission, reduce communication delays and data loss, and ensure that the robot can respond to control instructions in a timely manner;

[0047] S6 control system optimization, design of advanced control system, to achieve precise control of the robot, the control system can adjust the robot's movement direction and speed in real time according to the position and speed of the target object, to maintain close follow-up;

[0048] Among them, a feedback control mechanism is added to enable the robot to adjust itself according to the actual following situation. For example, if the robot finds that the distance to the target object is too large or too small, it can automatically adjust the speed and direction to achieve the best following effect;

[0049] S7 optimization measures in actual application, environmental adaptability, testing and optimization of different working environments to ensure that the robot can work normally in various complex environments. For example, in a dim or smoky environment, the parameters of the sensor can be adjusted to improve its adaptability;

[0050] Among them, the impact of different terrain and ground conditions on the robot's following is taken into account, and corresponding adjustments and optimizations are made. For example, on uneven ground, the robot can adopt a more stable movement mode to ensure the accuracy of following;

[0051] S8 human-machine interaction design, with a user-friendly human-machine interaction interface, allows users to easily operate and control the robot. The human-machine interaction interface can display the robot's status and location information, as well as the tracking status of the target object, allowing users to understand the robot's working status at any time;

[0052] The addition of voice recognition and voice prompt functions allows users to control the robot through voice commands, which can improve the convenience and efficiency of operation, especially in some situations that require two-handed operation.

[0053] See also Figure 1 ,Building the robot hardware and software framework,In terms of software framework, the innovation and technical advantages of this project lie in the research and functional integration of intelligent algorithms such as deep learning, transfer learning, and reinforcement learning. It mainly uses visual solutions, supplemented by multi-modal fusion solutions of infrared and point cloud information, and then conducts subsequent algorithm research. Therefore, when selecting hardware, the sensor plans to select depth plus infrared dual cameras and laser radar to obtain the three types of data required;

[0054] In terms of hardware framework, a basic model of omnidirectional mobility with Mecanum wheels is designed, equipped with an independent suspension structure to enhance the flexibility and stability of the robot. The chassis control system is a main control board based on STM32, and the FreeRTOS real-time operating system is deployed to transmit the status data of the chassis to the host computer main control system. The two systems are connected through buses such as UART for communication;

[0055] Furthermore, for the design of the main control embedded system, in view of the intelligent and high real-time characteristics of this project, it is planned to select a powerful edge computing product as the platform for the development of the project's main control embedded system. Edge computing refers to reducing the process of data transfer on the network, and being able to directly and timely process data at the source, which can make up for the shortcomings of traditional cloud computing with high data transmission overhead, and bring great improvement to real-time performance. This project plans to use Nvidia Jetson nano as the main control unit to design an embedded system that can match the needs. For wireless communication, sending and receiving identification information, mode switching, etc. between the host computer and the main control system, it is necessary to select, design and build embedded system equipment. This requires comprehensive consideration of factors such as the compatibility, transmission distance, power consumption and anti-interference ability of the wireless communication module, and select a communication module that meets the application requirements. At the same time, the model selection of the embedded processor is also very important, and it is necessary to ensure that its performance and memory are sufficient to support complex communication protocol processing and real-time requirements;

[0056] Furthermore, the host computer control system is designed. This project automatically follows the host computer control system of the intelligent logistics robot and adopts the ROS2 (Robot Operating System 2) robot operating system. ROS2 is an improved version of ROS. It realizes decentralization while maintaining the same basic concept as ROS. It introduces DDS (Data Distribution Service) data distribution service as a communication mechanism, supports more message transmission modes, such as point-to-point and propagation, and has better data transmission and real-time response capabilities to adapt to different application scenarios.

[0057] The main control system also needs to use a deep learning development framework and dependent libraries. The follower robot needs to choose a framework that supports dynamic graphs and dynamic neural networks. The Pytorch framework is the Python version of Torch, which is a deep neural network programming framework specifically for GPU acceleration. Torch is a classic tensor library that operates on multi-dimensional matrix data and is widely used in machine learning. Unlike the static calculation graph of Tensorflow, the calculation graph of Pytorch is dynamic and can be changed in real time according to calculation needs. Therefore, this project chooses Pytorch combined with other neural network inference libraries to deploy deep learning models;

[0058] Furthermore, in the research on target detection and recognition algorithms, based on the existing research foundation of target detection and recognition algorithms at home and abroad, the team deployed the lightweight and efficient YOLOv5 algorithm as a starting point, introduced the transfer learning algorithm to solve problems such as image noise, image blur, light and shadow occlusion, and improved the model training speed and generalization performance, so that the robot has the ability to adapt to the environment. After that, this project will further explore the deep learning neural network of multimodal image fusion, and plans to introduce a cross-modal learning structure based on a generative adversarial network, which can solve the problems of low efficiency and blurred details of the existing multimodal image fusion algorithm. This project constructs a fusion model through a generative adversarial network, integrates infrared and visible light image information, enhances the stability of the robot in bad weather and night conditions, and improves the robot's perception and adaptability to meet the application needs in complex and changing environments.

[0059] Furthermore, the research on environment perception and automatic following algorithms, in order to realize intelligent environment perception and automatic following in multiple environments, the robot hardware design is equipped with laser radar and depth camera to obtain dual-channel data, combined with the subsequent decision-making algorithm and motion control algorithm that adapt to the chassis structure to achieve precise position and trajectory control;

[0060] In terms of positioning algorithms, since the LiDAR SLAM algorithm transmits less information, has fast speed and high real-time performance, this project will deploy LiDAR-based SLAM real-time positioning and mapping algorithms to achieve basic functions. However, in a dynamically changing environment, a single sensor has limited action scenarios and limited environmental information obtained, which leads to problems such as poor real-time performance, path redundancy and local deadlock in the planned path. This project will further introduce deep reinforcement learning algorithms, use long and short-term memory neural networks to improve perception capabilities, use reinforcement learning to achieve interactive learning between robots and unknown environments, improve intelligent decision-making capabilities, and thus better respond to changes in obstacles, pedestrians and other dynamic objects, and achieve more accurate positioning and navigation.

[0061] Furthermore, the project will conduct systematic research and implementation of the automatic following intelligent logistics robot. The feasibility test experiment of the motion following task will be completed in the virtual simulation environment of the ROS2 system, and then the logistics handling debugging will be carried out in the real scene.

[0062] Furthermore, the method improves the handling efficiency. The optimized handling robot can automatically follow the target object without human intervention, which greatly improves the handling efficiency. The robot can quickly respond to the movement of the target object and adjust its position and speed in time to ensure that the goods can be transported to the destination in a timely and accurate manner. At the same time, the method enhances environmental adaptability. Through hardware upgrades and software algorithm optimization, the handling robot can better adapt to different working environments. For example, high-precision sensors can accurately detect target objects in dimly lit or smoky environments, and omnidirectional wheels or Mecanum wheels can enable the robot to move flexibly on uneven ground, improving the robot's ability to work in complex environments.

[0063] Furthermore, this method improves the following accuracy and stability. By using deep learning algorithms and multi-sensor fusion technology, the handling robot can more accurately identify and track the target object. At the same time, the optimized path planning algorithm and feedback control mechanism can ensure the stability and safety of the robot during the following process, reducing the occurrence of following errors and unexpected situations.

[0064] Furthermore, the method realizes convenient human-computer interaction, a friendly human-computer interaction interface and voice recognition function, so that users can operate and control the handling robot more conveniently. Users can control the robot through voice commands, which improves the convenience and efficiency of operation, especially in some occasions that require two-handed operation;

[0065] Furthermore, this method expands the scope of application. The optimized handling robot has higher performance and reliability and can be applied to more fields and scenarios. For example, in logistics distribution, manufacturing, medical and health industries, handling robots can automatically follow goods or personnel to achieve efficient handling and distribution tasks, saving labor costs for enterprises and improving production efficiency.

[0066] The embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in the technical field without departing from the purpose of the present invention.

Claims

1. A method for automatic following of a handling robot, characterized in that: The following methods are included: S1 hardware upgrade, sensor optimization, and the use of high-precision laser radar, which can more accurately detect the position, distance and shape of the target object. The laser radar has high resolution and fast scanning speed, and can obtain three-dimensional information of the surrounding environment in real time, providing more accurate data for the robot to follow; Equipped with multiple cameras to achieve all-round visual perception. Cameras at different angles can capture different features of the target object to improve recognition accuracy. At the same time, the camera can also be used to detect obstacles and environmental changes so that the robot can adjust the following strategy in time. The S2 motion system is improved, using high-performance motors and drivers to increase the robot's movement speed and acceleration, which allows the robot to respond more quickly to the movement of the target object and maintain close follow-up; Install omnidirectional wheels or Mecanum wheels to enable the robot to move flexibly in any direction and adapt to different working environments and following requirements; S3 software algorithm optimization, target recognition and tracking algorithm, using deep learning algorithm to identify and track target objects. Through a large amount of training data, the robot can learn to identify different types of target objects and accurately track targets in complex environments. Adopt multi-sensor fusion technology to fuse the data of lidar and camera sensors to improve the accuracy and stability of target recognition and tracking; S4 path planning algorithm, design an efficient path planning algorithm, so that the robot can follow the target object while avoiding obstacles and other interference factors. The path planning algorithm can adjust the robot's driving path in real time according to the moving speed and direction of the target object and changes in the surrounding environment; S5 communication and control system optimization, wireless communication technology, using stable and reliable wireless communication technology to ensure real-time communication between the robot and the control center. Wireless communication allows the control center to understand the status and location of the robot at any time, and remotely control and dispatch it; Optimize communication protocols, improve the efficiency and reliability of data transmission, reduce communication delays and data loss, and ensure that the robot can respond to control instructions in a timely manner; S6 control system optimization, design of advanced control system, to achieve precise control of the robot, the control system can adjust the robot's movement direction and speed in real time according to the position and speed of the target object, to maintain close follow-up; S7 optimization measures in actual application, environmental adaptability, testing and optimization of different working environments to ensure that the robot can work normally in various complex environments. For example, in a dim or smoky environment, the parameters of the sensor can be adjusted to improve its adaptability; S8 human-machine interaction design, with a user-friendly human-machine interaction interface, allows users to easily operate and control the robot. The human-machine interaction interface can display the robot's status and location information, as well as the tracking status of the target object, allowing users to understand the robot's working status at any time; The addition of voice recognition and voice prompt functions allows users to control the robot through voice commands, which can improve the convenience and efficiency of operation, especially in some situations that require two-handed operation.

2. The method for automatic following of a handling robot according to claim 1, characterized in that: The S4 path planning algorithm takes into account the kinematic and dynamic characteristics of the robot to ensure the stability and safety of the robot during the following process.

3. The method for automatic following of a handling robot according to claim 1, characterized in that: The S6 control system is optimized and a feedback control mechanism is added to enable the robot to adjust itself according to the actual following situation. If the robot finds that the distance to the target object is too large or too small, it can automatically adjust the speed and direction to achieve the best following effect.

4. The method for automatic following of a handling robot according to claim 1, characterized in that: The optimization measures in the actual application of S7 take into account the impact of different terrains and ground conditions on the robot's following and make corresponding adjustments and optimizations. On uneven ground, the robot can adopt a more stable movement mode to ensure the accuracy of following.

5. The method for automatic following of a handling robot according to claim 1, characterized in that: The robot's software and hardware framework is built. In terms of software framework, the innovation and technical advantages of this project lie in the research and functional integration of deep learning, transfer learning, and reinforcement learning intelligent algorithms. It is mainly based on visual solutions, supplemented by multimodal fusion solutions of infrared and point cloud information, and then conducts subsequent algorithm research. Therefore, when selecting hardware, the sensor plans to use depth plus infrared dual cameras and lidar to obtain the three types of data required.

6. The method for automatic following of a handling robot according to claim 1, characterized in that: In terms of hardware framework, a basic model of omnidirectional mobility with a Mecanum wheel structure is designed, equipped with an independent suspension structure to enhance the flexibility and stability of the robot. The chassis control system is a main control board based on STM32, and the FreeRTOS real-time operating system is deployed to transmit the status data of the chassis to the host computer main control system. The two systems are connected to communicate through the UART bus.

7. The method for automatic following of a handling robot according to claim 1, characterized in that: The design of the main control embedded system, in view of the intelligent and high real-time characteristics of this project, plans to select a powerful edge computing product as the platform for the development of the project's main control embedded system. Edge computing refers to reducing the process of data transfer on the network, and being able to process data directly and timely at the source, which can make up for the shortcomings of traditional cloud computing with high data transmission overhead, and bring great improvement to real-time performance. This project plans to use Nvidia Jetson nano as the main control unit to design an embedded system that can match the needs. For wireless communication, sending and receiving identification information, and mode switching between the host computer and the main control system, it is necessary to select, design, and build embedded system equipment.

8. The method for automatic following of a handling robot according to claim 1, characterized in that: Design of the host computer control system. This project automatically follows the host computer control system of the intelligent logistics robot and adopts the ROS2 (Robot Operating System 2) robot operating system. ROS2 is an improved version of ROS. It achieves decentralization while maintaining the same basic concept as ROS, and introduces DDS (Data Distribution Service) data distribution service as a communication mechanism to support more message transmission modes.

9. The method for automatic following of a handling robot according to claim 1, characterized in that: The main control system also needs to use a deep learning development framework and dependent libraries. The follower robot needs to choose a framework that supports dynamic graphs and dynamic neural networks. The Pytorch framework is the Python version of Torch, a deep neural network programming framework specifically for GPU acceleration. Torch is a classic tensor library that operates on multi-dimensional matrix data and is widely used in machine learning. Unlike the static calculation graph of Tensorflow, the calculation graph of Pytorch is dynamic and can be changed in real time according to calculation needs. Therefore, this project chooses Pytorch combined with other neural network inference libraries to deploy deep learning models.

Images