A remote, reversible, non-switching human-machine co-driving method and system for wheeled, fully-wire-controlled unmanned vehicles based on intent recognition

Abstract

A remote, reversible, non-switching human-machine co-driving method and system for a wheeled, fully-guided unmanned vehicle based on intent recognition, belonging to the field of human-machine co-driving technology, mainly includes the following steps: a remote control terminal collects human commands through an input device; the collected human command information is sent to the wheeled unmanned vehicle via a workshop communication network; during autonomous maneuvering, the unmanned vehicle receives the human commands from the control terminal, performs delay compensation correction on the human commands, and identifies the human intervention intent; the autonomous maneuvering planning commands are fused with the human commands to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle's drive control system. This method, by integrating human intervention correction with autonomous planning during the unmanned vehicle's autonomous maneuvering, achieves parallel and collaborative human-machine control, responds instantly to changes in human intent, improves the matching degree between the unmanned vehicle's autonomous maneuvering performance and human intent, and ensures smooth and stable vehicle control.

Description

Technical Field

[0001] This invention relates to the field of human-machine co-driving technology, and in particular to a remote reversible non-switching human-machine co-driving method and system for a wheeled, fully-guided unmanned vehicle based on intent recognition. Background Technology

[0002] Autonomous vehicles are gradually becoming a part of daily life. However, limited by the current maturity of hardware sensors and software intelligent algorithms, autonomous driving systems at this stage struggle to make accurate judgments on all complex operating conditions. They lack robustness and comprehensive handling capabilities, and are prone to errors in extracting environmental features under extreme lighting and complex road conditions. The algorithms also have a low tolerance for errors in predicting unstructured scenarios. While drivers possess flexible emergency response capabilities, long-term monitoring of the system can lead to problems such as distraction and fatigue, creating a dual dilemma of "technical blind spots + human error."

[0003] Against this backdrop, the core demand for human-machine co-driving has emerged. It is not only a key solution to the current driving dilemma, but also a necessary transitional stage towards fully autonomous driving. On the one hand, by dynamically perceiving the complexity of the environment and based on the smooth switching of control, it achieves complementary advantages in human and machine decision-making. It reduces the driving load by relying on the system's precise steady-state control capabilities and makes up for the shortcomings of the algorithm by leveraging human emergency judgment. On the other hand, it can accumulate massive amounts of interactive data in actual driving scenarios, reverse-optimize the algorithm model, and improve the system's perception capabilities, laying a solid foundation for the technological maturity and safe implementation of fully autonomous driving. It has become a core bridge that transcends the limitations of a single driving subject and connects the current stage of intelligent driving with the ultimate fully autonomous driving. Summary of the Invention

[0004] This disclosure provides a remote, reversible, non-switching human-machine co-driving method for a wheeled, fully-guided unmanned vehicle based on intent recognition. By manually intervening in the autonomous driving system, it enables autonomous maneuvering tasks in complex environments, thereby enhancing the system's environmental adaptability.

[0005] Its key solutions are as follows: ① During autonomous maneuvering, it integrates human intervention commands with autonomous maneuvering commands to achieve reversible intervention in non-switching modes while ensuring vehicle stability; ② It addresses latency issues in communication processes to ensure the synchronization of human intervention commands with autonomous driving commands; ③ The autonomous driving system adjusts its autonomous planning commands based on human intervention commands to more accurately achieve human intentions.

[0006] The remote reversible non-switching human-machine co-driving method for a wheeled, fully-guided unmanned vehicle based on intent recognition provided in this disclosure mainly includes the following steps: S1, the remote control terminal collects human commands through the input device; and sends the collected human command information to the wheeled unmanned vehicle terminal through the workshop communication network; S2: During autonomous maneuvering, the unmanned vehicle receives human commands from the control terminal, performs delay compensation and correction on the human commands, and identifies the intention of human intervention. S3 integrates autonomous maneuvering planning instructions with human instructions to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle drive control system.

[0007] Furthermore, step S1 specifically includes: S11, During the autonomous maneuver of the unmanned vehicle, the operator selects the unmanned vehicle that needs to be intervened in on the control terminal, obtains the control authority of the unmanned vehicle, sets the intervention mode to the reversible non-switching intervention mode, and sends it to the remote unmanned vehicle. S12, at the control end, the operator inputs manual intervention commands through the input device; after the control end collects the manual intervention commands, it sends them to the remote unmanned vehicle via wireless communication.

[0008] Furthermore, the specific intervention modes for the autonomous vehicle controlled by the operator include: fully autonomous mode, reversible non-switching intervention mode, reversible switching intervention mode, irreversible switching intervention mode, and remote control mode; among which: (1) Fully autonomous mode: The unmanned vehicle is in a fully autonomous maneuvering mode, and the control terminal cannot actively intervene in the execution of autonomous maneuvering commands. In this mode, unless the autonomous driving system actively requests human intervention, the unmanned vehicle refuses to parse and execute the maneuvering commands sent by the control terminal. (2) Reversible non-switching intervention mode: that is, after receiving human intervention instructions, the autonomous vehicle executes the autonomous motor decision-making and planning program in parallel. Through collaborative control, the human intervention instructions and the autonomous decision-making and planning program output instructions are integrated, and the autonomous vehicle executes based on the integrated instructions; (3) Reversible switching intervention mode: After the unmanned vehicle receives the human intervention command, the autonomous maneuvering decision-making and planning program is executed in parallel. The smooth switching module gradually transitions from the fully autonomous maneuvering command to the fully human command. After the human intervention command stops, the smooth switching module gradually transitions from the fully human command to the autonomous maneuvering command. (4) Irreversible switching intervention mode: When a remote manual intervention command is issued during the operation of the autonomous maneuvering program, the autonomous maneuvering program will immediately stop running, the unmanned vehicle will switch to remote control mode, and the manual control command will be fully responsible for subsequent control. (5) Remote control mode: The autonomous vehicle's autonomous maneuvering planning and decision-making program does not run, and the vehicle is completely controlled by the remote operator. In this mode, the maneuvering of the unmanned vehicle is completely controlled by the information collected by the remote control terminal input device.

[0009] Furthermore, in step S1, the method for collecting manual commands through an input device specifically includes: The accelerator pedal opening is collected as an acceleration command by the accelerator pedal; The brake pedal opening is collected as a braking command by the brake pedal; The steering wheel angle is collected as a steering command by the steering wheel. The upper travel of the joystick is collected as an acceleration command; the lower travel is collected as a braking command; and the left and right travels are collected as left and right steering commands.

[0010] Furthermore, in step S2, the method for delay compensation and correction of human commands by the autonomous vehicle includes: A command prediction model is constructed to predict the timing of manually corrected commands, ensuring that the commands match the real-time vehicle status. ; ; ; in, Communication delay is fed back in real time by the wireless communication system; , , For remote input of throttle, brake, and steering commands; The throttle, brake, and steering commands are adjusted in real time after compensation. , The first derivatives of the throttle, brake, and steering inputs. The second derivatives of the throttle, brake, and steering inputs are calculated using the following formulas: ; ; ; ; ; ; in, The sampling period is typically 10ms.

[0011] Furthermore, in step S2, the specific method by which the autonomous vehicle identifies the intention of human intervention based on the modified control commands includes: Input information: Adjusted throttle, brake, and steering commands ; Output information: The speed increase desired by the remote operator. Expected braking pressure The desired correction is the wheel steering angle. ; Map the throttle command to the desired increase in vehicle speed: ; in, This refers to the zero or free travel of the accelerator pedal or joystick. This refers to the maximum travel of the accelerator pedal or joystick. The maximum speed allowed for the vehicle; Map the braking command to the braking pressure to be executed: ; in, This refers to the zero position or free travel of the brake pedal or rocker arm during its downward stroke; This refers to the maximum travel of the brake pedal or rocker arm at its lowest point. The maximum permissible braking pressure for the vehicle; Map the direction command to the desired wheel steering angle correction: ; in, This refers to the zero or free travel of the steering wheel angle or the left and right travel of the joystick; This refers to the maximum travel of the steering wheel angle or the left / right movement of the joystick. This is the maximum permissible wheel angle for the vehicle.

[0012] Furthermore, step S3 specifically includes: S31, after the unmanned vehicle completes the interpretation of the human intervention intention, it sends the interpreted instructions to the unmanned vehicle's collaborative control module and autonomous driving system at the same time. S32, the unmanned vehicle-side collaborative control module, integrates human intervention commands and autonomous maneuver commands, and outputs fused control commands; S33, after receiving the fused control command, the unmanned vehicle end control execution module performs torque distribution calculation, steering control calculation and braking force calculation of the drive motor, and sends them to the motor controller, steering gear controller and brake controller for execution, and feeds back the vehicle status information to the unmanned vehicle autonomous driving system. S34, the autonomous driving system of the unmanned vehicle receives the parsed intention of human intervention and the state information of the vehicle after intervention, and adjusts its own expected speed and expected curvature planning to adapt to the human driving intention. In the S35, the autonomous vehicle transmits real-time status data to a remote control terminal via wireless communication, allowing the operator to determine whether to continue intervention.

[0013] Furthermore, in step S32, the specific method for integrating autonomous maneuvering planning instructions with manual instructions includes: Integration includes two dimensions: vertical and horizontal. In the vertical direction, a braking priority strategy is adopted. When a braking command for manual intervention is received, the speed command is reset to zero and the braking pressure is sent to the braking system. The fused speed expectation and braking pressure are processed according to the following formula: ; ; In the lateral direction, the relationship between the wheel angle direction calculated based on the manual intervention command and the wheel angle direction calculated based on the curvature output by the autonomous driving system is handled in two cases: when the wheel angle direction calculated by the manual intervention command is consistent with the wheel angle direction calculated by the autonomous driving system, a direct superposition strategy is adopted; when the wheel angle direction calculated by the manual intervention command is opposite to the wheel angle direction calculated by the autonomous driving system, the manual intervention command is taken as the final vehicle angle command, and the vehicle is completely controlled by the manual command. Assume the wheel steering angle analyzed by the autonomous driving system is The merged wheel angle command is then:

[0014] in, Based on the solution of the vehicle dynamics model Based on current vehicle speed The fusion coefficient is determined according to the following conditions: .

[0015] Furthermore, in step S34, the specific methods for adjusting the desired vehicle speed and desired curvature planning to adapt to human driving intentions include: Manual intervention commands are categorized into longitudinal intervention and lateral intervention based on their source. Accelerator and brake commands are longitudinal interventions, while directional commands are lateral interventions. (1) For longitudinal intervention, a cumulative maintenance strategy is adopted: Assuming the actual vehicle speed before human intervention was The actual vehicle speed after manual intervention was The desired vehicle speed planned by the autonomous driving system before human intervention is: If a manual intervention adjustment item is added to the desired vehicle speed after manual intervention, the adjusted desired vehicle speed will be... It is calculated using the following formula: ; in, The intervention impact coefficient, with a value range of [0,1], is dynamically adjusted based on the communication quality at the end of the intervention and is calculated using the following formula:

[0016] in, For communication delay, For network packet loss rate; Considering vehicle stability control, manual intervention commands are constrained by the current safe operating boundaries of the vehicle, as fed back by the vehicle control system. Assume the maximum permissible speed of the vehicle is... The final speed drive command sent to the drive control system is: ; (2) For lateral intervention, the expected curvature planned by the autonomous driving system when manual intervention is involved. The status is handled in two ways: When human intervention is involved, the desired curvature is... If the lane selected by the autonomous maneuver does not meet human expectations, the autonomous maneuver system will translate the planned trajectory to the current location point in the vertical direction of the original driving direction when the human intervention is required to exit. At this point, the final curvature after intervention is consistent with the desired instruction, that is... ; When human intervention is involved, the desired curvature is... This indicates that the steering curvature planned by the autonomous maneuvering system exhibits understeering or oversteering after execution, exceeding the planned desired curvature. When the manual intervention is withdrawn, the intervention factor is added back according to the following formula: ; in, The intervention impact coefficient has a value range of [0,1]. The average vehicle speed during the intervention period; Wheel angle to be corrected during intervention Curvature obtained from inverse analysis of vehicle dynamics The average; where

[0017]

[0018] Where n is the number of sampling cycles that run from the start of manual intervention to the end of manual intervention; Ultimately, the final curvature of the autonomous driving system output to the drive control system after lateral intervention is: .

[0019] A remotely reversible, non-switching human-machine co-driving system for a wheeled, fully-guided unmanned vehicle based on intent recognition, applying the above method, mainly includes: a remote control terminal and an unmanned vehicle terminal; wherein: The remote control terminal integrates multiple input devices for collecting human commands and sends the collected human command information to the wheeled unmanned vehicle terminal through the workshop communication network; at the same time, it receives vehicle status information and environmental information collected by the unmanned vehicle terminal and displays them on the human-machine interface. The unmanned vehicle terminal is used to receive human commands from the control terminal during autonomous maneuvering; perform delay compensation correction on the human commands and identify the intention of human intervention; then, it merges the autonomous maneuvering planning commands with the human commands to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle drive control system.

[0020] Compared with the prior art, the beneficial effects of this disclosure are: ① A remote reversible non-switching human-machine co-driving method for wheeled fully-controlled unmanned vehicles based on intent recognition is proposed to achieve parallel collaborative control with "autonomy as the main mode and human intervention as the auxiliary mode". It allows seamless human intervention and correction during autonomous maneuvering, ensuring smooth and stable vehicle control and timely response to changes in human intent. ② It improved the matching degree between the autonomous maneuverability of unmanned vehicles and human intentions; ③ During autonomous maneuvering, human intervention can be seamlessly initiated or withdrawn at any time to resolve the disconnect between intervention and takeover. ④ Manual intervention can correct the expected speed and expected curvature of autonomous driving planning, thereby improving the planning accuracy of the autonomous driving system; ⑤ Address the latency issues during communication to ensure the fusion and synchronization of manual intervention commands and autonomous driving commands. Attached Figure Description

[0021] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments of this disclosure taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same components.

[0022] Figure 1 This is a schematic diagram of a human-machine co-driving control method according to the present disclosure; Figure 2 This is a schematic diagram of the planned trajectory translation after manual intervention. Detailed Implementation

[0023] Preferred embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure will be thorough and complete, and will fully convey the scope of the present disclosure to those skilled in the art.

[0024] This disclosure provides a remote, reversible, non-switching human-machine co-driving method and system for a wheeled, fully-guided unmanned vehicle based on intent recognition.

[0025] In one exemplary implementation: 1. A human-machine co-driving control method / system framework based on intent recognition according to this disclosure is attached. Figure 1 As shown, the human-machine co-driving system upon which this method relies includes two parts: a remote control terminal and an unmanned vehicle terminal. The remote control terminal integrates a human-machine interface, voice input, buttons, a gear selector switch, a joystick, accelerator and brake pedals, a steering wheel, and other input devices. It is responsible for collecting human commands and sending the collected human command information to the wheeled unmanned vehicle terminal through the workshop communication network. At the same time, it receives vehicle status information and environmental information collected by the vehicle terminal and displays it on the human-machine interface. During autonomous maneuvering, the unmanned vehicle receives human commands from the control terminal. First, it performs delay compensation correction on the human commands and identifies the intention of human intervention. Then, it merges the autonomous maneuvering planning commands with the human commands to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle drive control system.

[0026] Specifically, the remote control terminal includes: an instruction acquisition module; The autonomous vehicle module includes: a communication delay compensation module, an intent understanding module, a collaborative control fusion module, and an autonomous driving adjustment module.

[0027] 2. A remote human-machine co-driving method for wheeled unmanned vehicles based on intent recognition mainly includes the following steps: Step 1: During the autonomous maneuver of the unmanned vehicle, the operator selects the unmanned vehicle that needs to be intervened in on the control terminal, obtains the control authority of the unmanned vehicle, sets the intervention mode to reversible non-switching intervention mode through the mode button, and sends it to the remote unmanned vehicle. Step 2: At the control end, the operator inputs manual intervention commands through the input device. After collecting the manual intervention commands, the control end sends them to the remote unmanned vehicle via wireless communication. Step 3: After receiving the intervention mode command sent by the control terminal, the autonomous vehicle terminal makes a judgment. If several pre-modes are reversible non-switchable intervention modes, the autonomous vehicle terminal will call the communication delay compensation module to correct the delay of the control command after receiving the manual intervention command; otherwise, it will be handled according to the requirements of other modes. Step 4: The autonomous vehicle calls the intent understanding module to perform manual intervention intent parsing on the corrected control commands, and sends the parsed commands to the autonomous vehicle's collaborative control module and autonomous driving system simultaneously. Step 5: The autonomous vehicle calls the collaborative control module to integrate human intervention commands and autonomous maneuver commands, and outputs a fused control command.

[0028] Step 6: After receiving the fusion control command, the unmanned vehicle control execution module performs torque distribution calculation, steering control calculation, and braking force calculation for the drive motor, and sends them to the motor controller, steering gear controller, and brake controller for execution, and feeds back the vehicle status information to the vehicle-side autonomous driving system.

[0029] Step 7: The autonomous driving system on the vehicle side receives the parsed human intervention intention and the vehicle's status information after intervention, and adjusts its desired speed and desired curvature planning to adapt to the human driving intention; Step 8: The autonomous vehicle sends real-time status data to the remote control terminal via wireless communication. The operator decides whether to continue intervention. If yes, return to step 2; otherwise, the operator stops inputting data, the input device returns to zero, and the vehicle enters a fully autonomous driving state.

[0030] Step 9: Finally, the operator switches the intervention mode to fully autonomous mode using the mode button on the control terminal to avoid accidental operation of the remote control terminal affecting autonomous maneuvering. At this point, the manual reversible non-switching intervention process ends.

[0031] 3. In step 1, there are 5 intervention modes: fully autonomous mode, reversible non-switching intervention mode, reversible switching intervention mode, irreversible switching intervention mode, and remote control mode. (1) Fully Autonomous Mode: The unmanned vehicle is in a fully autonomous maneuvering mode, and the control terminal cannot actively intervene in the execution of autonomous maneuvering commands. In this mode, unless the autonomous system actively requests human intervention, the unmanned vehicle refuses to parse and execute the maneuvering commands sent by the control terminal; (2) Reversible non-switching intervention mode means that after receiving human intervention instructions, the autonomous vehicle executes the autonomous motor decision-making and planning program in parallel. Through collaborative control, the human intervention instructions and the autonomous decision-making and planning program output instructions are integrated, and the autonomous vehicle executes based on the integrated instructions. (3) Reversible switching intervention mode: After the unmanned vehicle receives the human intervention command, the autonomous maneuvering decision-making and planning program is executed in parallel. The smooth switching module gradually transitions from the fully autonomous maneuvering command to the fully human command. After the human intervention command stops, the smooth switching module gradually transitions from the fully human command to the autonomous maneuvering command. (4) Irreversible switching intervention mode: When a remote manual intervention command is issued during the operation of the autonomous maneuvering program, the autonomous maneuvering program will immediately stop running, the unmanned vehicle will switch to remote control mode, and the manual control command will be fully responsible for subsequent control. (5) Remote control mode: The autonomous vehicle's autonomous maneuvering planning and decision-making program does not run, and the vehicle is completely controlled by a remote operator. In this mode, the maneuvering of the unmanned vehicle is entirely controlled by information collected by the remote control terminal input device.

[0032] 4. In step 2, the input device includes: (1) The accelerator pedal collects the accelerator pedal opening as the acceleration command; the brake pedal collects the brake pedal opening as the braking command; the steering wheel collects the steering wheel angle as the steering command. (2) The crossbar collects the upper travel as the acceleration command; the lower travel as the braking command; and the left and right travel as the left and right steering commands. 5. In step 3, the communication delay compensation module is invoked to correct the delay of the control commands, as detailed below: To address the issue of command lag caused by remote communication delays, a communication compensation module is implemented. This module uses predictive compensation to eliminate the impact of time differences and constructs a command prediction model to predict the timing of manually corrected commands, ensuring that the commands match the real-time vehicle status. ; ; ; in, Communication delay is fed back in real time by the wireless communication system; , , For remote input of throttle, brake, and steering commands; The throttle, brake, and steering commands are adjusted in real time after compensation. , The first derivatives of the throttle, brake, and steering inputs. The second derivatives of the throttle, brake, and steering inputs are calculated using the following formulas: ; ; ; ; ; ; in, The sampling period is typically 10ms.

[0033] 6. In step 4, the autonomous vehicle calls the intent understanding module to perform manual intervention intent parsing on the corrected control commands. The specific process is as follows: Input information: Adjusted throttle, brake, and steering commands ; Output information: The speed increase desired by the remote operator. Expected braking pressure The desired correction is the wheel steering angle. ; Map the throttle command to the desired increase in vehicle speed: ; in, This refers to the zero position (or free travel) of the accelerator pedal (or joystick). This refers to the maximum travel of the accelerator pedal (or joystick). The maximum speed allowed for the vehicle; Map the braking command to the braking pressure to be executed: ; in, This refers to the zero position (or free travel) of the brake pedal (or rocker arm). This refers to the maximum travel of the brake pedal (or the downward travel of the lever); The maximum permissible braking pressure for the vehicle; Map the direction command to the desired wheel steering angle correction: ; in, This refers to the zero position (or free travel) of the steering wheel angle (or the left and right travel of the joystick). This refers to the maximum travel of the steering wheel (or the left and right travel of the joystick); The maximum permissible wheel angle for the vehicle; 7. In step 4, the autonomous maneuver command is issued by the onboard autonomous system, typically indicating the desired vehicle speed. With expected curvature .

[0034] 8. In step 5, the autonomous vehicle calls the collaborative control module to integrate human intervention commands and autonomous maneuver commands, and outputs the following integrated control commands: The integration is handled in two dimensions: vertical and horizontal.

[0035] In the vertical direction, a braking priority strategy is adopted. When a braking command for manual intervention is received, the speed command is reset to zero and the braking pressure is sent to the braking system. The fused speed expectation and braking pressure are processed according to the following formula: ; ; In the lateral direction, the relationship between the wheel angle direction calculated based on the manual intervention command and the wheel angle direction calculated based on the curvature output by the autonomous driving system is handled in two ways: when the wheel angle direction calculated by the manual intervention command is consistent with that calculated by the autonomous driving system, a direct superposition strategy is adopted; when the wheel angle direction calculated by the manual intervention command is opposite to that calculated by the autonomous driving system, the manual intervention command is taken as the final vehicle angle command, and the vehicle is completely controlled by manual commands. Assume the wheel angle resolved by the autonomous driving system is... The merged wheel angle command is then: ; in, Based on the solution of the vehicle dynamics model Based on current vehicle speed The fusion coefficient is determined according to the following conditions.

[0036] ; 9. In step 7, adjust the desired vehicle speed and desired curvature planning to adapt to human driving intentions, as follows: Human intervention commands are generated when humans discover that the autonomous driving planning commands are unreasonable or do not conform to human driving intentions. In order to adapt to human intentions, the autonomous driving system needs to adjust its planned expected speed and expected curvature to avoid conflict with human intentions and go against human will.

[0037] Manual intervention commands are categorized into longitudinal intervention and lateral intervention based on their source. Accelerator and brake commands are longitudinal interventions, while directional commands are lateral interventions.

[0038] For longitudinal interventions, a cumulative maintenance strategy is employed. This assumes that before manual intervention (i.e., ...) =0 and The actual vehicle speed at (=0) is After artificial intervention (i.e.) 0 or The actual vehicle speed at 0:00 was The desired vehicle speed planned by the autonomous driving system before human intervention is: If a manual intervention adjustment item is added to the desired vehicle speed after manual intervention, the adjusted desired vehicle speed will be... It is calculated using the following formula: ; in, The intervention impact coefficient, with a value range of [0,1], is calculated at the end of the intervention (i.e., from...). 0 or The state of 0 becomes =0 and The communication quality is dynamically adjusted when the condition is 0. The following formula is used for calculation:

[0039] in, For communication delay, This refers to the network packet loss rate.

[0040] Considering vehicle stability control, manual intervention commands are constrained by the current safe operating boundaries of the vehicle, as fed back by the vehicle control system. Assume the maximum permissible speed of the vehicle is... The final speed drive command sent to the drive control system is: ; For lateral intervention, it should be handled according to manual intervention ( When the curve changes from 0 to non-zero, the expected curvature planned by the autonomous driving system The status is handled in two ways: The expected curvature of the autonomous maneuvering system when human intervention is involved. If the autonomous driving system selects a lane that does not meet human expectations (possibly due to unidentified obstacles ahead or an excessively bumpy lane), the system will disengage upon human intervention. When the value changes from non-zero to zero, the planned trajectory is translated to the current positioning point along the vertical direction of the original driving direction. For example... Figure 2 As shown.

[0041] At this point, the final curvature after intervention is consistent with the desired instruction, that is... ; When human intervention is involved, the desired curvature is... This indicates that the steering curvature planned by the autonomous maneuvering system exhibits understeering or oversteering after execution, exceeding the planned desired curvature. When the manual intervention is withdrawn, the manual intervention factor is added back according to the following formula.

[0042] ; in, The intervention impact coefficient has a value range of [0,1]. The average vehicle speed during the intervention period; Wheel angle to be corrected during intervention Curvature obtained from inverse analysis of vehicle dynamics The average; where ; ; Where n is the number of sampling cycles that run between the start and end of manual intervention.

[0043] Ultimately, the final curvature of the autonomous driving system output to the drive control system after lateral intervention is:

[0044] 10. In step 8, the vehicle entering the fully autonomous driving state means that the unmanned vehicle is completely controlled by the vehicle's autonomous driving system. That is, in the reversible non-switching intervention mode, if all input devices are in the zero position, the unmanned vehicle is also in the fully autonomous driving state.

[0045] Application Examples like Figure 1 As shown, with a 4 4. Take, for example, a fully line-controlled unmanned wheeled vehicle with Ackerman steering and a remote control seat.

[0046] 1. During the autonomous maneuver of the unmanned vehicle, the operator selects the unmanned vehicle that needs to be intervened in on the control terminal, obtains the control authority of the unmanned vehicle, sets the intervention mode to the reversible non-switching intervention mode through the mode button, and sends it to the remote unmanned vehicle. 2. At the control end, the operator inputs manual intervention commands via the steering wheel, accelerator, and brake pedals in the remote control cabin. The control end collects the accelerator pedal opening as the acceleration command. Collect the brake pedal opening as the braking command. Collect steering wheel angle as steering command , via wireless communication , , Send to the remote driverless vehicle; 3. After receiving the intervention mode command from the control terminal, the autonomous vehicle determines whether the pre-mode is a reversible, non-switchable intervention mode. If so, upon receiving the manual intervention command, the autonomous vehicle calls the communication delay compensation module to correct the delay in the control command. ; ; ; in, Communication delay is fed back in real time by the wireless communication system; , , For remote input of throttle, brake, and steering commands; The throttle, brake, and steering commands are adjusted in real time after compensation. , The first derivatives of the throttle, brake, and steering inputs. The second derivatives of the throttle, brake, and steering inputs are calculated using the following formulas: ; ; ; ; ; ; in, The sampling period is typically 10ms.

[0047] Otherwise, it should be handled according to the requirements of other modes; 4. The autonomous vehicle calls the intent understanding module to perform manual intervention intent parsing on the corrected control commands.

[0048] Map the throttle command to the desired increase in vehicle speed: ; in, This refers to the zero position (or free travel) of the accelerator pedal (or joystick). This refers to the maximum travel of the accelerator pedal (or joystick). The maximum speed allowed for the vehicle; Map the braking command to the braking pressure to be executed: ; in, This refers to the zero position (or free travel) of the brake pedal (or rocker arm). This refers to the maximum travel of the brake pedal (or the downward travel of the lever); The maximum permissible braking pressure for the vehicle; Map the direction command to the desired wheel steering angle correction: ; in, This refers to the zero position (or free travel) of the steering wheel angle (or the left and right travel of the joystick). This refers to the maximum travel of the steering wheel (or the left and right travel of the joystick); The maximum permissible wheel angle for the vehicle; The parsed instructions are simultaneously sent to the autonomous vehicle's collaborative control module and autonomous driving system. 5. The autonomous vehicle invokes the collaborative control module, integrating human intervention commands. , ) and autonomous maneuvering commands (desired vehicle speed) With expected curvature Output fusion control commands: The integration is handled in two dimensions: vertical and horizontal.

[0049] In the vertical direction, a braking priority strategy is adopted. When a braking command for manual intervention is received, the speed command is reset to zero and the braking pressure is sent to the braking system. The fused speed expectation and braking pressure are processed according to the following formula: ; ; In the lateral direction, the relationship between the wheel angle direction calculated based on the manual intervention command and the wheel angle direction calculated based on the curvature output by the autonomous driving system is handled in two ways: when the wheel angle direction calculated by the manual intervention command is consistent with that calculated by the autonomous driving system, a direct superposition strategy is adopted; when the wheel angle direction calculated by the manual intervention command is opposite to that calculated by the autonomous driving system, the manual intervention command is taken as the final vehicle angle command, and the vehicle is completely controlled by manual commands. Assume the wheel angle resolved by the autonomous driving system is... The merged wheel angle command is then: ; in, Based on the solution of the vehicle dynamics model , Based on current vehicle speed The fusion coefficient is determined according to the following conditions.

[0050] ; 6. After receiving the fusion control command, the unmanned vehicle-side drive control execution module performs torque distribution calculation, steering control calculation, and braking force calculation for the drive motor, and sends them to the motor controller, steering gear controller, and brake controller for execution, and feeds back the vehicle status information to the vehicle-side autonomous driving system.

[0051] 7. The autonomous driving system on the vehicle side receives the parsed intention of human intervention and the vehicle's state information after intervention, and adjusts its desired speed and curvature planning to adapt to the human driving intention: Human intervention commands are generated when humans discover that the autonomous driving planning commands are unreasonable or do not conform to human driving intentions. In order to adapt to human intentions, the autonomous driving system needs to adjust its planned expected speed and expected curvature to avoid conflict with human intentions and go against human will.

[0052] Manual intervention commands are categorized into longitudinal intervention and lateral intervention based on their source. Accelerator and brake commands are longitudinal interventions, while directional commands are lateral interventions.

[0053] For longitudinal interventions, a cumulative maintenance strategy is employed. This assumes that before manual intervention (i.e., ...) =0 and The actual vehicle speed at (=0) is After artificial intervention (i.e.) 0 or The actual vehicle speed at 0:00 was The desired vehicle speed planned by the autonomous driving system before human intervention is: If a manual intervention adjustment item is added to the desired vehicle speed after manual intervention, the adjusted desired vehicle speed will be... It is calculated using the following formula: ; in, The intervention impact coefficient, with a value range of [0,1], is calculated at the end of the intervention (i.e., from...). 0 or The state of 0 becomes =0 and The communication quality is dynamically adjusted when the condition is 0. The following formula is used for calculation:

[0054] in, For communication delay, This refers to the network packet loss rate.

[0055] Considering vehicle stability control, manual intervention commands are constrained by the current safe operating boundaries of the vehicle, as fed back by the vehicle control system. Assume the maximum permissible speed of the vehicle is... The final speed drive command sent to the drive control system is: ; For lateral intervention, it should be handled according to manual intervention ( When the curve changes from 0 to non-zero, the expected curvature planned by the autonomous driving system The state is handled in two ways.

[0056] When human intervention is involved, the desired curvature is... If the autonomous driving system selects a lane that does not meet human expectations (possibly due to unidentified obstacles ahead or an excessively bumpy lane), the system will disengage upon human intervention. When the value changes from non-zero to zero, the planned trajectory is translated to the current positioning point along the vertical direction of the original driving direction. For example... Figure 2 As shown.

[0057] At this point, the final curvature after intervention is consistent with the desired instruction, that is... ; When human intervention is involved, the desired curvature is... This indicates that the steering curvature planned by the autonomous maneuvering system exhibits understeering or oversteering after execution, exceeding the planned desired curvature. When the manual intervention is withdrawn, the manual intervention factor is added back according to the following formula.

[0058] ; in, The intervention impact coefficient has a value range of [0,1]. The average vehicle speed during the intervention period; Wheel angle to be corrected during intervention Curvature obtained from inverse analysis of vehicle dynamics The average; where ; ; Where n is the number of sampling periods between the start and end of manual intervention. Depend on It is calculated using the following formula: ; Where B is the track width, the lateral distance between the centers of the left and right wheels of the front axle; Ultimately, the final curvature of the autonomous driving system output to the drive control system after lateral intervention is:

[0059] 8. The autonomous vehicle sends real-time status data to the remote control terminal via wireless communication. The operator decides whether to continue intervention. If yes, return to step 2; otherwise, the operator stops inputting data, the input device returns to zero, and the vehicle enters a fully autonomous driving state.

[0060] 9. Finally, the operator switches the intervention mode to fully autonomous mode using the mode button on the control terminal to avoid accidental operation of the remote control terminal affecting autonomous maneuvering. At this point, the manual reversible non-switching intervention process ends.

[0061] The above technical solutions are merely exemplary embodiments of the present invention. For those skilled in the art, based on the application methods and principles disclosed in the present invention, it is easy to make various types of improvements or modifications, and not limited to the methods described in the specific embodiments of the present invention. Therefore, the methods described above are merely preferred and not restrictive.

Claims

1. A remote, reversible, non-switching human-machine co-driving method for a wheeled, fully-guided unmanned vehicle based on intent recognition, characterized in that: Includes the following steps: S1, the remote control terminal collects human commands through the input device; and sends the collected human command information to the wheeled unmanned vehicle terminal through the workshop communication network; S2: During autonomous maneuvering, the unmanned vehicle receives human commands from the control terminal, performs delay compensation and correction on the human commands, and identifies the intention of human intervention. S3 integrates autonomous maneuvering planning instructions with human instructions to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle drive control system.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: S11, During the autonomous maneuver of the unmanned vehicle, the operator selects the unmanned vehicle that needs to be intervened in on the control terminal, obtains the control authority of the unmanned vehicle, sets the intervention mode to the reversible non-switching intervention mode, and sends it to the remote unmanned vehicle. S12, at the control end, the operator inputs manual intervention commands through the input device; after the control end collects the manual intervention commands, it sends them to the remote unmanned vehicle via wireless communication.

3. The method according to claim 2, characterized in that, The specific intervention modes for the autonomous vehicle controlled by the operator include: fully autonomous mode, reversible non-switching intervention mode, reversible switching intervention mode, irreversible switching intervention mode, and remote control mode; among which: (1) Fully autonomous mode: The unmanned vehicle is in a fully autonomous maneuvering mode, and the control terminal cannot actively intervene in the execution of autonomous maneuvering commands. In this mode, unless the autonomous driving system actively requests human intervention, the unmanned vehicle refuses to parse and execute the maneuvering commands sent by the control terminal. (2) Reversible non-switching intervention mode: that is, after receiving human intervention instructions, the autonomous vehicle executes the autonomous motor decision-making and planning program in parallel. Through collaborative control, the human intervention instructions and the autonomous decision-making and planning program output instructions are integrated, and the autonomous vehicle executes based on the integrated instructions; (3) Reversible switching intervention mode: After the unmanned vehicle receives the human intervention command, the autonomous maneuvering decision-making and planning program is executed in parallel. The smooth switching module gradually transitions from the fully autonomous maneuvering command to the fully human command. After the human intervention command stops, the smooth switching module gradually transitions from the fully human command to the autonomous maneuvering command. (4) Irreversible switching intervention mode: When a remote manual intervention command is issued during the operation of the autonomous maneuvering program, the autonomous maneuvering program will immediately stop running, the unmanned vehicle will switch to remote control mode, and the manual control command will be fully responsible for subsequent control. (5) Remote control mode: The autonomous vehicle's autonomous maneuvering planning and decision-making program does not run, and the vehicle is completely controlled by the remote operator. In this mode, the maneuvering of the unmanned vehicle is completely controlled by the information collected by the remote control terminal input device.

4. The method according to claim 1, characterized in that, In step S1, the method for collecting manual instructions through an input device specifically includes: The accelerator pedal opening is collected as an acceleration command by the accelerator pedal; The brake pedal opening is collected as a braking command by the brake pedal; The steering wheel angle is collected as a steering command by the steering wheel. The upper travel of the joystick is collected as an acceleration command; the lower travel is collected as a braking command; and the left and right travels are collected as left and right steering commands.

5. The method according to claim 1, characterized in that, In step S2, the method for delay compensation and correction of human commands by the unmanned vehicle includes: A command prediction model is constructed to predict the timing of manually corrected commands, ensuring that the commands match the real-time vehicle status. ； ； ； in, Communication delay is fed back in real time by the wireless communication system; , , For remote input of throttle, brake, and steering commands; The throttle, brake, and steering commands are adjusted in real time after compensation. , The first derivatives of the throttle, brake, and steering inputs. The second derivatives of the throttle, brake, and steering inputs are calculated using the following formulas: ； ； ； ； ； ； in, The sampling period is typically 10ms.

6. The method according to claim 5, characterized in that, In step S2, the specific method for the autonomous vehicle to identify the intention of human intervention based on the corrected control commands includes: Input information: Adjusted throttle, brake, and steering commands ; Output information: The speed increase desired by the remote operator. Expected braking pressure The desired correction is the wheel steering angle. ; Map the throttle command to the desired increase in vehicle speed: ； in, This refers to the zero or free travel of the accelerator pedal or joystick. This refers to the maximum travel of the accelerator pedal or joystick. The maximum speed allowed for the vehicle; Map the braking command to the braking pressure to be executed: ； in, This refers to the zero position or free travel of the brake pedal or rocker arm during its downward stroke; This refers to the maximum travel of the brake pedal or rocker arm at its lowest point. The maximum permissible braking pressure for the vehicle; Map the direction command to the desired wheel steering angle correction: ； in, This refers to the zero or free travel of the steering wheel angle or the left and right travel of the joystick; This refers to the maximum travel of the steering wheel angle or the left / right movement of the joystick. This is the maximum permissible wheel angle for the vehicle.

7. The method according to any one of claims 1-6, characterized in that, Step S3 specifically includes: S31, after the unmanned vehicle completes the interpretation of the human intervention intention, it sends the interpreted instructions to the unmanned vehicle's collaborative control module and autonomous driving system at the same time. S32, the unmanned vehicle-side collaborative control module, integrates human intervention commands and autonomous maneuver commands, and outputs fused control commands; S33: After receiving the fusion control command, the unmanned vehicle end control execution module performs torque distribution calculation, steering control calculation and braking force calculation of the drive motor, and sends them to the motor controller, steering gear controller and brake controller for execution, and feeds back the vehicle status information to the unmanned vehicle autonomous driving system. S34, the autonomous driving system of the unmanned vehicle receives the parsed intention of human intervention and the state information of the vehicle after intervention, and adjusts its own expected speed and expected curvature planning to adapt to the human driving intention. In the S35, the autonomous vehicle transmits real-time status data to a remote control terminal via wireless communication, allowing the operator to determine whether to continue intervention.

8. The method according to claim 7, characterized in that, In step S32, the specific method for integrating autonomous maneuvering planning instructions with manual instructions includes: Integration includes two dimensions: vertical and horizontal. In the vertical direction, a braking priority strategy is adopted. When a braking command for manual intervention is received, the speed command is reset to zero and the braking pressure is sent to the braking system. The fused speed expectation and braking pressure are processed according to the following formula: ； ； In the lateral direction, the relationship between the wheel angle direction calculated based on the manual intervention command and the wheel angle direction calculated based on the curvature output by the autonomous driving system is handled in two cases: when the wheel angle direction calculated by the manual intervention command is consistent with the wheel angle direction calculated by the autonomous driving system, a direct superposition strategy is adopted; when the wheel angle direction calculated by the manual intervention command is opposite to the wheel angle direction calculated by the autonomous driving system, the manual intervention command is taken as the final vehicle angle command, and the vehicle is completely controlled by the manual command. Assume the wheel steering angle analyzed by the autonomous driving system is The merged wheel angle command is then: in, Based on the solution of the vehicle dynamics model Based on current vehicle speed The fusion coefficient is determined according to the following conditions: 。 9. The method according to claim 7, characterized in that, In step S34, the specific methods for adjusting the desired vehicle speed and desired curvature planning to adapt to human driving intentions include: Manual intervention commands are categorized into longitudinal intervention and lateral intervention based on their source. Accelerator and brake commands are longitudinal interventions, while directional commands are lateral interventions. (1) For longitudinal intervention, a cumulative maintenance strategy is adopted: Assuming the actual vehicle speed before human intervention was The actual vehicle speed after manual intervention was The desired vehicle speed planned by the autonomous driving system before human intervention is: If a manual intervention adjustment item is added to the desired vehicle speed after manual intervention, the adjusted desired vehicle speed will be... It is calculated using the following formula: ； in, The intervention impact coefficient, with a value range of [0,1], is dynamically adjusted based on the communication quality at the end of the intervention and is calculated using the following formula: in, For communication delay, For network packet loss rate; Considering vehicle stability control, manual intervention commands are constrained by the current safe operating boundaries of the vehicle, as fed back by the vehicle control system. Assume the maximum permissible speed of the vehicle is... The final speed drive command sent to the drive control system is: ； (2) For lateral intervention, the expected curvature planned by the autonomous driving system when manual intervention is involved. The status is handled in two ways: When human intervention is involved, the desired curvature is... If the lane selected by the autonomous maneuver does not meet human expectations, the autonomous maneuver system will translate the planned trajectory to the current location point in the vertical direction of the original driving direction when the human intervention is required to exit. At this point, the final curvature after intervention is consistent with the desired instruction, that is... ; When human intervention is involved, the desired curvature is... This indicates that the steering curvature planned by the autonomous maneuvering system results in either understeering or oversteering after execution, exceeding the planned desired curvature. When the manual intervention is withdrawn, the intervention factor is added back according to the following formula: ； in, The intervention impact coefficient has a value range of [0,1]. The average vehicle speed during the intervention period; Wheel angle to be corrected during intervention Curvature obtained from inverse analysis of vehicle dynamics The average; where Where n is the number of sampling cycles that run from the start of manual intervention to the end of manual intervention; Ultimately, the final curvature of the autonomous driving system output to the drive control system after lateral intervention is: 。 10. A remotely reversible, non-switching human-machine co-driving system for a wheeled, fully-guided unmanned vehicle based on intent recognition, using the method described in any one of claims 1-9, characterized in that... include: This includes remote control terminals and driverless vehicle terminals; among which: The remote control terminal integrates multiple input devices for collecting human commands and sends the collected human command information to the wheeled unmanned vehicle terminal through the workshop communication network; at the same time, it receives vehicle status information and environmental information collected by the unmanned vehicle terminal and displays them on the human-machine interface. The unmanned vehicle terminal is used to receive human commands from the control terminal during autonomous maneuvering; perform delay compensation correction on the human commands and identify the intention of human intervention; then, it merges the autonomous maneuvering planning commands with the human commands to obtain the final vehicle control target, which is then decomposed and executed by the unmanned vehicle drive control system.

Images