Remote man-machine co-driving method, device and system and storage medium

Abstract

The invention discloses a remote man-machine co-driving method, device and system, and a storage medium, and the method comprises the steps: achieving the dynamic connection management of multiple vehicles through a reverse proxy technology, carrying out the adaptive scheduling of video and audio stream bandwidth through the combination of an operation focus, and carrying out the remote control of variable-weight man-machine co-driving through the fusion of an artificial potential field auxiliary decision and a tactile feedback mechanism. According to the technical scheme, the problems that an existing remote driving system is difficult in multi-vehicle management, low in bandwidth utilization rate and poor in takeover safety in a weak network environment are solved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicles and remote driving technology, specifically relating to a remote human-machine co-driving method, device, system, and storage medium. Background Technology

[0002] With the advancement of 5G and autonomous driving technologies, remote driving has become an important auxiliary means in scenarios such as mining areas, ports, and chemical industrial parks. Operators can simultaneously monitor and take over multiple unmanned vehicles in emergencies through a cloud-based cockpit. However, this model, which tightly connects people, vehicles, and the cloud, also brings new challenges: First, managing large-scale connections is difficult. Most existing systems rely on fixed network addresses (such as static IPs) for "point-to-point" connections, which is not flexible enough. Once the number of vehicles increases or the network environment changes, the configuration and maintenance costs are high, and vehicles cannot achieve automatic access. Second, bandwidth resources are prone to conflict. When multiple vehicles upload high-definition video at the same time, network congestion is very likely to occur. This situation can lead to video stuttering and even loss of control signals, seriously affecting driving safety. Finally, the takeover process poses safety hazards. Traditional takeover methods usually involve a direct switch between "fully automatic" and "fully manual" modes. In the event of network latency, instability, or encountering sudden obstacles, if there is no mechanism to smoothly transition based on latency and risk levels, vehicles are prone to instability.

[0003] There has been some research on the communication and control issues related to remote driving. The paper [Real-Life Implementation and Evaluation of Coupled Congestion Control for WebRTC Media and Data Flows] implements the FSEv2 algorithm at the browser level, enabling RTP and SCTP congestion control to work together, thus protecting the quality and low latency of real-time video communication while ensuring data transmission. However, it neglects the bandwidth allocation problem under multi-channel concurrency. The paper [Challenges of Remote Driving on Public Roads Using 5G Public Networks] mentions that while traditional fixed-IP-based vehicle management systems are stable, they cannot adapt to the dynamically changing 5G intranet environment. The paper [Improved Perceptual Quality of Traffic Signs and Lights for the Teleoperation of Autonomous Vehicle Remote Driving via Multi-CategoryRegion of Interest Video Compression] proposes an improved video compression method that significantly improves the visual quality of traffic signs and lights, crucial for remote drivers, under limited network bandwidth and low latency requirements. However, it only considers the network-side state and does not incorporate the operator's subjective focus (attention), resulting in low resource allocation efficiency. The paper [Remote Human-Machine Cooperative Driving Control Method for...] IntelligentVehicles proposed a remote human-machine cooperative driving control method based on variable structure control, which realizes the switching between remote driver and autonomous driving controller. However, its switching logic is essentially based on the switching of safety boundaries, lacks a dynamic weight smooth transition mechanism based on real-time risk or driver intention, and does not consider the tactile interaction feedback of the driver during the takeover process, resulting in insufficient on-site perception ability of the driver. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides a remote human-machine co-driving method, device, system, and storage medium. It designs a remote co-driving platform with strong scalability, high resource utilization, and latency robustness. By combining Frp reverse proxy technology with a Redis database, it solves the problem of dynamic multi-vehicle addressing in environments without public IP addresses. Through a differentiated bandwidth scheduling strategy based on operational focus, it addresses network congestion in multi-vehicle concurrent scenarios. Finally, by integrating the Artificial Potential Field (APF) algorithm with force feedback technology, it ensures driving safety in weak network environments.

[0005] To achieve the above objectives, the present invention provides the following solution: A remote human-machine co-driving method includes: Step S1: Construct a multi-intelligent vehicle remote control system architecture; wherein, the architecture includes: vehicle terminal, cloud, control terminal, downlink low-latency control link and uplink immersive video feedback link; Step S2: Design a dynamic connection model based on the architecture of the multi-intelligent vehicle remote control system; Step S3: Based on the dynamic connection model, design an attention-based bandwidth adaptive scheduling strategy; wherein, the strategy automatically adjusts the transmission quality according to the operator's current focus of attention; Step S4: Execute the co-driving control mechanism based on the architecture of the multi-intelligent vehicle remote control system, the dynamic connection model, and the bandwidth adaptive scheduling strategy.

[0006] As a preferred option, in step S2, based on the architecture of the multi-intelligent vehicle remote control system, a dynamic service discovery mechanism based on Frp is designed using reverse proxy technology and dynamic data mapping technology.

[0007] As a preferred option, in step S4, an artificial potential field algorithm and force feedback technology are introduced to implement a closed-loop safety control mechanism that includes time delay measurement, obstacle avoidance assistance, force interaction, and dynamic weight allocation.

[0008] The present invention also provides a remote human-machine co-driving device, comprising: The first processing module is used to construct a remote control system architecture for multiple intelligent vehicles; wherein, the architecture includes: vehicle terminal, cloud terminal, control terminal, downlink low-latency control link and uplink immersive video feedback link; The second processing module is used to design a dynamic connection model based on the architecture of the multi-intelligent vehicle remote control system. The third processing module is used to design an attention-based bandwidth adaptive scheduling strategy according to the dynamic connection model; wherein the strategy automatically adjusts the transmission quality according to the operator's current focus of attention; The fourth processing module is used to execute the co-driving control mechanism based on the architecture of the multi-intelligent vehicle remote management and control system, the dynamic connection model, and the bandwidth adaptive scheduling strategy.

[0009] As a preferred option, the second processing module, based on the architecture of the multi-intelligent vehicle remote control system, adopts reverse proxy technology and dynamic data mapping technology to design a dynamic service discovery mechanism based on Frp.

[0010] As a preferred option, the fourth processing module introduces an artificial potential field algorithm and force feedback technology to execute a closed-loop safety control mechanism that includes time delay measurement, obstacle avoidance assistance, force interaction, and dynamic weight allocation.

[0011] The present invention also provides a remote human-machine co-driving system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a remote human-machine co-driving method when executed by the processor.

[0012] The present invention also provides a storage medium storing a computer program, which executes a remote human-machine co-driving method when running.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. Lightweight and highly scalable architecture: It abandons the complex VPN networking and uses Frp dynamic port mapping to open up a channel for direct access from the cloud to the vehicle's internal network, allowing the system to automatically identify and obtain the vehicle's connection address.

[0014] 2. High bandwidth resource utilization: A dual-mode scheduling strategy is established to concentrate the limited 5G uplink bandwidth on the vehicles that the operator is most concerned about, thus taking into account both overall monitoring and precise control.

[0015] 3. High security in weak network environments: A variable-weight co-driving mechanism based on an artificial potential field was constructed. This method can dynamically adjust the permission ratio between manual and automatic driving according to the real-time network latency and convert the repulsive force of obstacles into force feedback from the steering wheel, allowing the operator to perceive environmental risks through touch even in environments with video lag. This effectively solves the problem of operational errors caused by the asynchrony between visual images and control signals in remote driving. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the 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.

[0017] Figure 1 This is a flowchart of the remote human-machine co-driving method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the data link for a single vehicle. Figure 3 This is a flowchart of adaptive bandwidth scheduling for audio and video streams based on operational focus. Figure 4 Weighting of remote human-machine co-driving A graph showing how network latency changes. Detailed Implementation

[0018] 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.

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0020] Example 1 like Figure 1 As shown, this invention provides a remote human-machine co-driving method, progressing step-by-step in the order of "establishing connection—optimizing perception—ensuring safety." First, a three-layer system hardware architecture is constructed, including the vehicle-mounted terminal, the cloud, and the control terminal. Second, reverse proxy technology and dynamic data mapping technology are used to overcome communication barriers between the 5G internal network and the public network. This mechanism enables dynamic vehicle registration, effectively solving the problem of difficult remote access. Next, based on the established connection, to address the bandwidth shortage issue that may result from simultaneous transmission by multiple vehicles, an audio / video stream scheduling strategy based on operator attention is introduced. This strategy automatically adjusts the transmission quality according to the operator's current focus, prioritizing high-definition, low-latency visibility for the target vehicle, thus solving the problem of video transmission stuttering. Finally, to address the control safety challenges in weak network environments, a method integrating the Artificial Potential Field (AFS) algorithm and force feedback technology is introduced to convert obstacle repulsion into steering wheel force feedback, thereby ensuring driving safety. Specifically, it includes: Step 1: Establish a multi-intelligent vehicle remote control system architecture To meet the requirements of multi-vehicle collaborative control in this invention, a three-layer system architecture of "vehicle-cloud-device" is first established.

[0021] Vehicle-mounted terminal (controlled end): The vehicle is an intelligent inspection vehicle with a drive-by-wire chassis. It is equipped with environmental perception sensors (cameras, radar) and a 5G communication module (CPE). It has a built-in reverse proxy client (Frpc) for establishing communication tunnels, and a local autonomous control module for executing underlying security policies.

[0022] Cloud server (transfer layer): Deployed with reverse proxy server (Frps) and in-memory database (Redis). Frps is used to provide intranet penetration service, and Redis is used to maintain dynamically generated "vehicle ID-port" mapping table.

[0023] Remote management platform (control terminal): Includes a visual interactive interface and a driving simulator. The platform embeds a "bandwidth scheduling controller" and a "human-machine co-driving arbitrator," which are used for audio and video stream management and control command generation, respectively.

[0024] like Figure 2 As shown, for remote driving of a single vehicle within a multi-intelligent vehicle, the system architecture designed in this invention includes two core data links: a downlink low-latency control link and an uplink immersive video feedback link.

[0025] Downlink Control Link: Designed to achieve real-time, second-level control of remote vehicles. This process begins with the main control unit's interactive subsystem. The system uses the USB HID protocol to collect physical control signals generated by the driver's force feedback from the steering wheel and accelerator pedal, encapsulates these signals into high-priority network data packets, and sends them to the public network mapping port of the cloud relay server. After identifying the corresponding intranet penetration tunnel, the cloud-based Frps server transparently forwards the control commands to the vehicle-mounted controlled terminal located in the 5G intranet environment via a TCP / UDP long connection. Upon receiving the data, the vehicle-mounted terminal uses a local input mapping module to restore the network commands to standard operating system-level control signals. Finally, the controller reads these signals and converts them into control information containing linear and angular velocities. This control information is then input to the vehicle chassis via the ( / cmd_vel) topic to achieve control, thus completing the closed-loop control from remote input to physical execution.

[0026] Uplink video feedback link: Responsible for providing the driver with a real-time driving view. The vehicle's environmental perception camera captures the view in front of the vehicle in real time and publishes the data through ROS topics. The in-vehicle streaming service (StreamingService) captures this frame and calls the GPU hardware encoder (NVENC / VCE) of the onboard computing unit to perform efficient H.264 / H.265 compression, generating RTP / RTSP video streams. The packaged video data is actively pushed to the cloud via a reverse tunnel established by Frpc, and then forwarded to the main control terminal interaction subsystem by Frps. The main control terminal uses the hardware decoding capability of the local terminal to restore the video data and render the low-latency real-time road condition image of 1080P / 60fps onto the driver's screen, realizing a closed-loop feedback from in-vehicle environmental perception to remote visual monitoring.

[0027] Step 2: Design a dynamic service discovery mechanism based on Frp To address the challenges of dynamic changes in vehicle IP addresses and large-scale cluster management in 5G non-public network environments, this invention establishes a dynamic connection model based on a three-stage process of "registration-heartbeat-mapping".

[0028] Vehicle registration: After the onboard terminal starts up, it initiates a registration request to the cloud management service via HTTPS. Upon receiving the request, the cloud server verifies the legitimacy of the registration. If the verification is successful, it queries the available port pool in the Redis database and allocates an idle public network port. (e.g., 7001), and generate the corresponding Frp proxy configuration file. The cloud updates the Frps service configuration through the API hot-reload mechanism, activating the TCP / UDP pass-through function of this port.

[0029] State Storage and Mapping: A vehicle metadata table with a hash structure is established in the cloud-based Redis database and named Vehicle:{ID}:Info. The stored content includes: Proxy_Port (mapping port), Inner_IP (internal IP), and Login_Time (login time). This design eliminates the need for the monitoring platform to pre-define fixed connection addresses in its code. The platform only needs to query the Redis database to obtain the current access point for any vehicle in real time. This mechanism avoids the tedious process of manually configuring connection parameters; once a vehicle connects to the network, the system automatically identifies it and establishes a connection.

[0030] Heartbeat Keep-Alive and Abnormal Circuit Breaker: Establish a WebSocket long connection between the vehicle terminal and the cloud, and set the heartbeat period. The heartbeat packet carries real-time telemetry data: {CPU_Load, Mem_Usage, 5G_Signal_Strength}. A watchdog timer is set in the cloud. If continuous... If no heartbeat is received from a vehicle within 3 seconds, the system automatically triggers the circuit breaker mechanism: immediately releasing the mapped port occupied by that vehicle. The system also marks the vehicle as offline in Redis and grays out the vehicle icon on the front-end interface to prevent operators from accidentally connecting.

[0031] Step 3: Design an attention-based bandwidth adaptive scheduling strategy To simultaneously monitor all vehicles and precisely control specific vehicles within the limited 5G upload bandwidth, this system is designed with two modes: "global data-saving monitoring" (background mode) and "directional high-definition takeover" (focus mode). For example... Figure 3 As shown, the system dynamically adjusts the underlying video transmission configuration through application layer control commands to achieve switching between different modes.

[0032] Define operator for vehicle Attention status , where 0 represents "global provincial flow monitoring" and 1 represents "targeted high-definition takeover".

[0033] Video stream encoding parameters Defined as a quintuple: .

[0034] Scheduling policy function The design is as follows: 1. Global data saving monitoring The system automatically enters this mode when a vehicle is not selected by the operator but is only under monitoring in a list or map. The global data saving monitoring mode expression is as follows: The vehicle-mounted device invokes the GStreamer framework to start a low-power streaming pipeline. The encoder is configured to MJPEG format, resolution is set to 320×240 (QVGA), frame rate is limited to 10fps, and target bitrate is controlled within 300Kbps. In this mode, data packets are extremely small, occupying only a tiny amount of bandwidth, ensuring that the network backbone does not become congested when multiple vehicles connect concurrently.

[0035] 2. Targeted HD takeover Triggered when an operator is detected to perform a "double-click" or "maximize" operation on the vehicle icon. The expression for the directional HD takeover mode is as follows: The front-end sends a control signal {"Action":"Focus","Target":"Vehicle_01"} to the cloud via WebSocket. Upon receiving the signal, the vehicle-mounted unit seamlessly switches to a high-performance encoding link without interrupting the connection, utilizing the input-selector component. The encoding format is switched to H.265 (HEVC), zero-latency tuning is enabled, and the frame rate is adjusted to reduce encoding and decoding time. The resolution is increased to 1920×1080 (FHD), the frame rate is increased to 30fps or 60fps, and the dynamic bitrate (VBR) target is set to 4Mbps-8Mbps. The cloud identifies the traffic on this port and marks its packets as high priority (DSCP EF), ensuring they are prioritized for forwarding in the network transmission queue.

[0036] Step 4: Construct a variable-weight human-machine co-driving control mechanism based on an artificial potential field To address the issues of lag in control and mismatch between human and vehicle perception caused by reliance on visual feedback in weak network environments, this invention introduces the Artificial Potential Field (APF) algorithm and force feedback technology to construct a closed-loop safety control system that includes "time delay measurement, obstacle avoidance assistance, force interaction, and dynamic weight allocation".

[0037] 4.1 Constructing an auxiliary decision-making mechanism based on a potential field model: First, a virtual force field model of the driving environment is established, mapping the remote driver's operational intentions to a "target gravitational field" and mapping obstacles sensed by the vehicle-mounted lidar to an "obstacle repulsive field"; when the vehicle enters the obstacle's safe radius... At that time, a repulsive force inversely proportional to the square of the distance is generated. Simultaneously, gravity is generated based on the driver's desired path. .

[0038] The onboard computing unit calculates the resultant force vector of the two forces in real time. The direction of the resultant force represents the optimal path under the current environment, merging the target path with vehicle safety considerations. Subsequently, the system uses the PurePursuit algorithm to convert the resultant force vector into a steering angle. , This refers to the automatic obstacle avoidance steering angle calculated by the vehicle-mounted terminal based on the artificial potential field, providing a benchmark for subsequent human-machine co-driving integration.

[0039] 4.2 Variable weight fusion control based on dual perception of execution latency and risk: To address the control lag issue in weak network environments, the system establishes a high-precision end-to-end delay measurement mechanism. NTP time synchronization and a sliding window filter are used to calculate the smoothed one-way network delay in real time. and network round-trip time (RTT) This allows for the dynamic adjustment of the driver's and vehicle's authority levels. Ultimately, control commands are issued to the chassis. The chassis control commands are: This refers to the desired steering angle input by the remote driver via the cockpit, representing the intention of manual driving. Weighting coefficient. Defined as the percentage of permissions granted for remote human driving. For example... Figure 4 As shown, the system uses a Sigmoid nonlinear function to establish the mapping relationship between time delay and weight, and is constrained by environmental risk factors. In the low-delay interval (… ), The system is in remote driving mode, accurately responding to remote commands; in high latency intervals ( ), The degradation rapidly approaches zero, and the system smoothly switches to autonomous driving mode to prevent blindly executing delayed commands. Environmental risk factors... This determines the upper limit of the artificial weight. For example... Figure 4 As shown, in a security scenario ( Under low latency It can reach 1.0; however, in high-risk scenarios (such as...) Even with good network communication, the system will force the human weight to be limited to a low level, compelling the vehicle to adopt more local obstacle avoidance commands based on APF, thereby achieving autonomous safety protection.

[0040] 4.3 Establish a closed loop for tactile feedback and abnormal circuit breaker tripping. To achieve two-way perception capabilities between humans and vehicles and ensure driving safety, the system incorporates force-sensing interaction and a circuit breaker mechanism. The total repulsive force field gradient calculated on the vehicle side is transmitted back to the cloud via a 5G uplink. The remote driving terminal generates a reverse damping torque on the steering wheel based on this repulsive force value. When the vehicle approaches an obstacle, the driver can intuitively feel the repulsive force on the steering wheel, thus instinctively taking evasive action in situations where vision may be impaired. If a network interruption or latency exceeding a safety threshold for two consecutive seconds is detected, the system immediately triggers the circuit breaker mechanism, forcibly resetting the weights. Set to 0 and apply maximum braking force brake lock until the network is restored or manual intervention is required.

[0041] The innovative aspects of this invention are as follows: 1. A dynamic service discovery method based on the combination of reverse proxy and in-memory database is proposed, which solves the problem of addressing and connection management of multiple mobile terminals in the absence of public IP address.

[0042] 2. A differentiated bandwidth scheduling strategy based on operational attention was designed. Unlike the traditional approach where all vehicles compete for bandwidth, the system switches the video stream between "data-saving mode" and "high-definition mode" according to the driver's operation.

[0043] 3. A tactile human-machine co-driving system based on "artificial potential field + force feedback" has been established. Unlike the limitations of traditional remote driving that relies on pure visual perception, this system introduces a virtual force field algorithm to transform the repulsive force of obstacles into force feedback from the steering wheel.

[0044] Example 2 The present invention also provides a remote human-machine co-driving device, comprising: The first processing module is used to construct a remote control system architecture for multiple intelligent vehicles; wherein, the architecture includes: vehicle terminal, cloud terminal, control terminal, downlink low-latency control link and uplink immersive video feedback link; The second processing module is used to design a dynamic connection model based on the architecture of the multi-intelligent vehicle remote control system. The third processing module is used to design an attention-based bandwidth adaptive scheduling strategy according to the dynamic connection model; wherein the strategy automatically adjusts the transmission quality according to the operator's current focus of attention; The fourth processing module is used to execute the co-driving control mechanism based on the architecture of the multi-intelligent vehicle remote management and control system, the dynamic connection model, and the bandwidth adaptive scheduling strategy.

[0045] As one embodiment of the present invention, the second processing module, based on the architecture of the multi-intelligent vehicle remote control system, adopts reverse proxy technology and dynamic data mapping technology to design a dynamic service discovery mechanism based on Frp.

[0046] As one embodiment of the present invention, the fourth processing module introduces an artificial potential field algorithm and force feedback technology to execute a closed-loop safety control mechanism that includes time delay measurement, obstacle avoidance assistance, force interaction, and dynamic weight allocation.

[0047] Example 3 The present invention also provides a remote human-machine co-driving system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program executes a remote human-machine co-driving method when executed by the processor.

[0048] Example 4 The present invention also provides a storage medium storing a computer program, which executes a remote human-machine co-driving method when running.

[0049] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A remote human-machine co-driving method, characterized in that, include: Step S1: Construct a multi-intelligent vehicle remote control system architecture; wherein, the architecture includes: vehicle terminal, cloud, control terminal, downlink low-latency control link and uplink immersive video feedback link; Step S2: Design a dynamic connection model based on the architecture of the multi-intelligent vehicle remote control system; Step S3: Based on the dynamic connection model, design an attention-based bandwidth adaptive scheduling strategy; wherein, the strategy automatically adjusts the transmission quality according to the operator's current focus of attention; Step S4: Execute the co-driving control mechanism based on the architecture of the multi-intelligent vehicle remote control system, the dynamic connection model, and the bandwidth adaptive scheduling strategy.

2. The remote human-machine co-driving method as described in claim 1, characterized in that, In step S2, based on the architecture of the multi-intelligent vehicle remote control system, a dynamic service discovery mechanism based on Frp is designed using reverse proxy technology and dynamic data mapping technology.

3. The remote human-machine co-driving method as described in claim 2, characterized in that, In step S4, an artificial potential field algorithm and force feedback technology are introduced to execute a closed-loop safety control mechanism that includes time delay measurement, obstacle avoidance assistance, force interaction, and dynamic weight allocation.

4. A remote human-machine co-driving device, characterized in that, include: The first processing module is used to construct a remote control system architecture for multiple intelligent vehicles; wherein, the architecture includes: vehicle terminal, cloud terminal, control terminal, downlink low-latency control link and uplink immersive video feedback link; The second processing module is used to design a dynamic connection model based on the architecture of the multi-intelligent vehicle remote control system. The third processing module is used to design an attention-based bandwidth adaptive scheduling strategy according to the dynamic connection model; wherein the strategy automatically adjusts the transmission quality according to the operator's current focus of attention; The fourth processing module is used to execute the co-driving control mechanism based on the architecture of the multi-intelligent vehicle remote management and control system, the dynamic connection model, and the bandwidth adaptive scheduling strategy.

5. The remote human-machine co-driving device as described in claim 4, characterized in that, Based on the architecture of the multi-intelligent vehicle remote control system, the second processing module adopts reverse proxy technology and dynamic data mapping technology to design a dynamic service discovery mechanism based on Frp.

6. The remote human-machine co-driving device as described in claim 5, characterized in that, The fourth processing module introduces artificial potential field algorithm and force feedback technology to execute a closed-loop safety control mechanism that includes time delay measurement, obstacle avoidance assistance, force interaction and dynamic weight allocation.

7. A remote human-machine co-driving system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program performing the remote human-machine co-driving method as described in any one of claims 1-3 when executed by the processor.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed, performs the remote human-machine co-driving method as described in any one of claims 1-3.

Images