Vehicle getting out of trouble guidance methods and devices

Abstract

This invention provides a vehicle traction guidance method and apparatus, relating to the technical field of vehicle control, comprising: responding to a traction guidance request, acquiring comprehensive vehicle perception information; determining the type of entrapment based on the comprehensive vehicle perception information; performing drivability analysis on terrain information in the comprehensive vehicle perception information using a real-time tire dynamics model according to the entrapment type, generating a time-seriesed traction traction operation guidance sequence; converting the traction traction operation guidance sequence into driver-oriented guidance information and outputting it; by constructing a local digital elevation model of the tire-ground contact interface and calculating the adhesion vector field, and combining the entrapment type for drivability analysis and force distribution optimization, it can provide drivers with accurate and real-time operation guidance in complex terrain, reducing reliance on driving experience and significantly improving the traction success rate and safety.

Description

Technical Field

[0001] This invention relates to the technical field of vehicle control, and in particular to a method and apparatus for guiding vehicles out of difficult situations. Background Technology

[0002] With the increasing popularity of outdoor travel, vehicles are prone to getting stuck on unpaved roads such as sand, mud, and snow due to insufficient tire traction. In such situations, getting out of trouble requires a high level of experience and skill from the driver.

[0003] Existing vehicle traction control systems primarily rely on passive control devices such as traction control, differential locks, and crawl mode, whose main function is to prevent skidding. While some advanced driver assistance systems (ADAS) possess a certain degree of automatic traction capability, their algorithmic decisions are often not optimal in complex, unstructured terrain environments. Furthermore, if an automatic attempt fails, the driver still needs to take over, at which point the driver may become more panicked due to a lack of effective guidance. Summary of the Invention

[0004] The purpose of this invention is to provide a vehicle extrication guidance method and device to alleviate the technical problem that novice drivers blindly operate the vehicle, which increases the depth of the vehicle getting stuck, causes vehicle damage, and results in a low success rate of extrication.

[0005] In a first aspect, the present invention provides a method for guiding a vehicle out of trouble, comprising: In response to a request for guidance on getting out of trouble, the system determines the type of vehicle entanglement based on comprehensive vehicle perception information. Based on the type of entrapment, a real-time tire dynamics model is used to perform a passability analysis on the comprehensive perception information of the vehicle, and a time-sequential entrapment operation guidance sequence is generated. The sequence of instructions for getting out of trouble is converted into driver-oriented guidance information and output.

[0006] In an optional implementation, the vehicle's comprehensive perception information includes surrounding terrain information; the step of generating a time-seriesd escape operation guidance sequence includes: Based on the surrounding terrain information, a local digital elevation model of the tire-ground contact interface is constructed. The terrain information represented by the local digital elevation model is input into the real-time tire dynamics model to calculate the adhesion vector field of each tire under the current terrain. Based on the type of entrapment, the real-time tire dynamics model is used to perform a passability analysis on the surrounding terrain information to determine the vehicle's passable direction and entrapment strategy. With the goal of maximizing the effective traction force of the vehicle in the passable direction, the optimal tire force distribution scheme is solved based on the adhesion vector field. The optimal tire force distribution scheme is mapped as a time-sequential command sequence of steering wheel angle, accelerator pedal opening, and brake pedal action.

[0007] In an optional implementation, when the traction control strategy includes vehicle body swaying, the step of generating a time-sequential traction control guidance sequence further includes: Determine the expected swing amplitude based on the type of entrapment; Based on the adhesion vector field, the transient ground imprint changes of each tire under the expected swing amplitude are calculated; Based on the transient grounding imprint changes, the peak adhesion window period for each tire is determined; The peak command of accelerator pedal opening in the time-sequential command sequence is aligned with the peak adhesion window period in terms of timing.

[0008] In an optional implementation, the vehicle's comprehensive perception information includes surrounding terrain information; the step of generating a time-seriesd escape operation guidance sequence further includes: Based on the surrounding terrain information, the longitudinal slope, lateral slope, and flatness of the pit where the vehicle is currently located are identified. When at least one of the longitudinal slope, transverse slope, or pit bottom flatness exceeds a preset threshold, the chassis scraping risk area generated during the escape operation is predicted. Based on the chassis scraping risk area, chassis protection operation instructions are inserted into the time-sequential escape operation guidance sequence; the chassis protection operation instructions include reducing the accelerator pedal opening to reduce the vehicle pitch angle, or adjusting the steering wheel angle to avoid the scraping risk area.

[0009] In an optional implementation, the comprehensive vehicle perception information includes tire slip ratio information and surrounding terrain information; the step of determining the type of vehicle entrapment includes: Based on the slip ratio information of each tire and the surrounding terrain information, it is determined whether the vehicle has the ability to extricate itself from a difficult situation. If available, the driver's skill in getting out of trouble, reaction speed, and operational precision in the vehicle's historical driving data will be used to determine the current driver's ability to get out of trouble. Based on the total number of operation instructions in the escape operation guidance sequence, the timing coordination accuracy between each operation instruction, and the types of operation parameters that need to be adjusted in coordination, the operation complexity required for autonomous escape is determined. The operational complexity is matched with the assessment result of the ability to get out of trouble. If the operational complexity exceeds the driver's ability to get out of trouble, then the type of being trapped is determined to be a type of being trapped that requires automatic intervention.

[0010] In an optional implementation, the method further includes: Based on historical vehicle driving data, a probabilistic model is established to characterize the deviation distribution of drivers when executing various types of operation commands; Based on the probability model and the preset minimum requirements for getting out of trouble, the operational error tolerance range of key parameters in each operation command is determined; the operational error tolerance range is the allowable range of parameters in which the driver's operational deviation will not lead to failure to get out of trouble. After outputting the guidance information, the actual deviation of the driver from the guidance information is continuously monitored; When the actual execution deviation exceeds the operation tolerance range, the current vehicle state is taken as the initial state, the key parameter with the widest operation tolerance range is selected to recalculate each operation instruction in the escape operation guidance sequence and update the guidance information.

[0011] In an optional implementation, the method further includes: The system acquires images of the vehicle's surroundings from the vehicle-mounted surround-view camera and performs semantic segmentation to identify passable areas, obstacle areas, and uncertain areas. For the uncertain region, the vehicle tires are controlled to perform a trial rotation of a preset amplitude. Based on the change in tire slip ratio during the trial rotation, the ground physical properties of the uncertain region are inferred. Based on the passable area, the obstacle area, and the uncertain area containing the ground physical properties, the surrounding terrain information in the vehicle's comprehensive perception information is determined.

[0012] In an optional implementation, the step of generating a time-sequential escape operation guidance sequence further includes: Based on the type of entrapment, determine whether the extrication operation involves the reciprocating swaying of the vehicle; If reciprocating oscillation is involved, the degree of change in ground physical properties caused by repeated tire rolling under different oscillation phases is calculated based on the real-time tire dynamics model. Based on the degree of change, the tire adhesion parameters of the subsequent swing cycle are dynamically updated, and the accelerator pedal opening and steering wheel angle in the traction operation guidance sequence are adjusted accordingly.

[0013] In an optional implementation, the step of generating a time-sequential escape operation guidance sequence further includes: Based on the type of entrapment, a parameterized strategy space is constructed with adjustable parameters such as steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing. In the parameterized strategy space, with the success rate of escape and the estimated time for escape as optimization objectives, an online optimization algorithm is used to search for the optimal parameter combination; Based on the optimal parameter combination, the time-series escape operation guidance sequence is generated.

[0014] Secondly, the present invention provides a vehicle traction guidance device, comprising: The judgment module, in response to the request for extrication guidance, determines the type of vehicle entrapment based on the vehicle's comprehensive perception information; The generation module, based on the type of entrapment, uses a real-time tire dynamics model to perform a passability analysis on the comprehensive perception information of the vehicle, and generates a time-sequential sequence of extrication operation guidance. The guidance module converts the escape operation guidance sequence into driver-oriented guidance information and outputs it.

[0015] This invention provides a vehicle traction guidance method and apparatus. First, it responds to a traction guidance request and acquires comprehensive vehicle perception information. Then, based on this comprehensive perception information, it determines the type of entrapment. Next, according to the entrapment type, it uses a real-time tire dynamics model to perform drivability analysis on the terrain information in the comprehensive vehicle perception information, generating a time-sequential traction traction operation guidance sequence. Finally, it converts this sequence into driver-oriented guidance information and outputs it. This application addresses two key points: First, it combines a real-time tire dynamics model with terrain drivability analysis, ensuring that the generation of the traction traction guidance sequence is based on a precise physical model, rather than relying solely on empirical rules. Second, it employs a time-sequential operation guidance sequence, clarifying the execution order and timing of each operation instruction, enabling the driver to complete the traction traction operation step-by-step according to the system guidance. This effectively solves the technical problem in the prior art where the driver is in a black box state and lacks real-time operation guidance in traction traction scenarios.

[0016] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a vehicle getting out of trouble guidance method provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the functional modules of a vehicle getting out of trouble guidance device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions 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, 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.

[0021] Currently, the system relies heavily on driver experience. Novice drivers lack understanding of vehicle dynamics and terrain assessment, and blind operation may lead to the vehicle getting stuck deeper, damaging the vehicle, or even causing it to roll over. In extremely complex and highly uncertain unstructured environments, the algorithmic decisions of fully automatic extrication systems may not be optimal, and if the automatic attempt fails, the driver still needs to take over, which may cause the driver to panic even more.

[0022] Based on this, the vehicle getting out of trouble guidance method and device provided in the embodiments of the present invention can transform professional getting out of trouble experience into intelligent real-time operation guidance, which significantly reduces the threshold of getting out of trouble operation and improves the success rate and safety of getting out of trouble.

[0023] To facilitate understanding of this embodiment, a detailed description of a vehicle extrication guidance method disclosed in this embodiment of the invention will be provided first.

[0024] Figure 1 This is a flowchart of a vehicle getting out of trouble guidance method provided in an embodiment of the present invention.

[0025] Reference Figure 1 The method includes the following steps: Step S102: In response to the request for extrication guidance, determine the type of vehicle entrapment based on the vehicle's comprehensive perception information; Here, for example, when the driver triggers a distress guidance request through the in-vehicle infotainment system, voice command, or physical button, the multi-sensor fusion perception module is activated to simultaneously collect the vehicle's current attitude information (including vehicle pitch angle, roll angle, yaw angle, etc.), tire slip ratio information (calculated through wheel speed sensors and vehicle speed sensors), and surrounding terrain information (images and distance data acquired through sensors such as in-vehicle surround view cameras, ultrasonic radar, and lidar). All of this information is integrated into comprehensive vehicle perception information, serving as the data foundation for subsequent assessment and decision-making.

[0026] Slip ratio refers to the degree of relative slippage between the tire and the ground during rolling, usually expressed as a percentage. Slip ratio = (wheel linear speed - vehicle speed) / wheel linear speed × 100%. When the slip ratio reaches 100%, it means that the tire is completely spinning and the vehicle cannot obtain effective driving force.

[0027] This step achieves multi-dimensional synchronous perception of the vehicle's own status and the surrounding environment through multi-sensor fusion, overcoming the shortcomings of limited perception range and insufficient accuracy of a single sensor, and providing a comprehensive and accurate data foundation for subsequent identification of the type of entrapment and generation of extrication strategies.

[0028] In a preferred embodiment, the step of acquiring comprehensive vehicle perception information may include: receiving local terrain data and ground condition information sent by nearby vehicles or roadside infrastructure via an onboard V2X communication module, and fusing the local terrain data with surrounding terrain information collected by onboard sensors. Specifically, it may receive tire slip ratio change curves, ground adhesion coefficient estimates, and pothole morphology data recorded by neighboring vehicles when they last passed through the area, and weightedly fuse these historical data with the real-time perception data of the current vehicle. The weights are dynamically adjusted based on the data timestamp, the similarity between the source vehicle and the current vehicle model, and the degree of change in environmental conditions (such as before and after rainfall). In this way, the system can use swarm intelligence to compensate for the spatiotemporal limitations of single-vehicle perception, which is particularly suitable for collaborative extrication in off-road vehicle convoy scenarios. By fusing V2X communication with multi-vehicle data, the spatiotemporal coverage of single-vehicle perception is expanded, enabling the system to obtain historical passage data of the terrain ahead in advance, avoiding repeated traps and improving the foresight and accuracy of extrication decisions.

[0029] When a vehicle becomes stuck in complex terrain such as sand or mud, each sensor simultaneously collects images of the surrounding terrain, distance information, vehicle posture, and tire slip rate data. This data is then fused to form comprehensive vehicle perception information. Based on this comprehensive perception information, the type of entrapment is determined.

[0030] Here, comprehensive vehicle perception information is analyzed to determine the current type of entrapment. The judgment criteria include: whether the slip rate of each tire exceeds a preset threshold, whether the vehicle body is in an abnormal tilted state, and whether the surrounding terrain constitutes an obstacle to passage. Types of entrapment include, but are not limited to: one tire stuck in a rut, diagonally opposite tires suspended in the air, chassis bottoming out, and all wheels slipping. Different types of entrapment correspond to different extrication strategies and operational complexities. The judgment process can employ a rule-based expert system or a machine learning-based classification model, automatically identifying the characteristic patterns of the entrapment type by learning from historical extrication data. This step achieves automated assessment and classification of the vehicle's entrapment state, avoiding misjudgments that may result from relying on the driver's subjective judgment, and providing a basis for the accurate matching of subsequent extrication strategies.

[0031] In a preferred embodiment, determining the type of vehicle entrapment further includes: constructing a digital twin model of the vehicle's current state based on comprehensive vehicle perception information, and simulating the execution effects of various entrapment operation sequences in a simulation environment using the digital twin model. The accuracy of the entrapment type determination is then verified based on the simulation results. When the simulation results deviate from the initial determination, the system automatically corrects the entrapment type determination and regenerates the entrapment operation guidance sequence. The construction of the digital twin model utilizes tire dynamics parameters, suspension characteristic parameters, and chassis geometric parameters calibrated from historical vehicle driving data to ensure a high degree of consistency between the simulation results and the actual physical response. Through digital twin simulation verification, the accuracy of entrapment type determination is improved, ineffective entrapment operations due to misjudgment are avoided, and entrapment time and vehicle wear are reduced.

[0032] Step S104: Based on the type of entrapment, use a real-time tire dynamics model to perform a drivability analysis on the terrain information in the vehicle's comprehensive perception information, and generate a time-seriesd entrapment operation guidance sequence. Here, the real-time tire dynamics model calculates the adhesion vector field of each tire under the current terrain based on terrain information from the vehicle's comprehensive perception information, including the direction and amplitude distribution of the adhesion force. Drivability analysis, based on the adhesion vector field and the type of entrapment, determines the vehicle's passable direction and traction strategy. The time-sequential traction-avoidance operation guidance sequence includes operational instructions such as steering wheel angle, accelerator pedal opening, brake pedal action, and gear selection, clearly defining the execution order and time window of each instruction. The sequence can be generated through analytical solutions based on a physical model, or by searching for the optimal combination in the parameterized strategy space using an online optimization algorithm.

[0033] Real-time tire dynamics models are mathematical calculation models based on tire mechanical properties (such as the Pacejka Magic Formula tire model and the Dugoff tire model) used to describe the force transmission relationship between the tire and the ground. This model takes tire vertical load, slip angle, slip ratio, and road adhesion coefficient as input parameters and outputs tire longitudinal force, lateral force, and self-aligning torque.

[0034] The aforementioned adhesion force vector field can be calculated using the following formula:

[0035]

[0036] in, and They represent the first The longitudinal and lateral forces of each tire. For the first The coefficient of adhesion between the tire and the ground. For the first Vertical load of each tire For the first The slip ratio of each tire. For the first The slip angle of each tire, and This refers to the nonlinear function in the tire model. It transforms the traction operation from experience-based fuzzy judgment into quantitative optimization based on a precise physical model, giving the generated guidance sequence a clear physical basis and improving the scientific rigor and reliability of the traction strategy.

[0037] In a preferred embodiment, generating a time-series-based escape operation guidance sequence further includes: acquiring crowdsourced escape data stored in a cloud server, the crowdsourced escape data including operation sequence records of successful escapes by different vehicle models under different terrain conditions and corresponding terrain feature labels; performing similarity matching between the crowdsourced escape data and the current vehicle's entrapment type and terrain information; when a match is found with crowdsourced escape data whose similarity exceeds a preset threshold, using the operation sequence corresponding to the crowdsourced escape data as an initial solution, and performing local optimization in conjunction with a real-time tire dynamics model to generate a time-series-based escape operation guidance sequence suitable for the current vehicle parameters and terrain conditions. The similarity matching is based on cosine similarity calculation of terrain feature vectors (including terrain type, slope, ground material, etc.). By introducing cloud-based crowdsourced escape data and utilizing group experience to provide a high-quality initial solution for the optimization algorithm, the convergence speed of online optimization is significantly accelerated, while simultaneously improving the generalization ability of the escape strategy under diverse terrain conditions.

[0038] Step S106: Convert the traction operation guidance sequence into driver-oriented guidance information and output it.

[0039] Here, a multimodal human-machine interface is used to transform the traction control sequence into driver-understandable guidance information in real time. This guidance information includes, but is not limited to: steering wheel rotation direction and amplitude indicators (dynamically displayed via a steering wheel icon on the instrument panel or HUD), accelerator pedal timing and depth indicators (e.g., lightly press the accelerator to 30% and hold), brake pedal input indicators, and gear shifting indicators (e.g., shifting to reverse). The guidance information is simultaneously output through multiple modal methods, including visual (graphics + text), auditory (voice prompts), and tactile (steering wheel vibration), ensuring the driver can accurately receive the instructions under any circumstances. This step transforms the professional traction control sequence into intuitive and easy-to-understand guidance information, enabling even inexperienced novice drivers to gradually complete the traction control operation under system guidance. This lowers the technical barrier to traction control and avoids secondary entrapment or accidents caused by misoperation.

[0040] In a preferred embodiment, converting the escape operation guidance sequence into driver-oriented guidance information and outputting it further includes: acquiring the driver's real-time physiological state data, including heart rate, grip strength, and gaze direction; assessing the driver's level of tension and concentration based on the physiological state data; and automatically adjusting the output method of the guidance information when the assessment results indicate that the driver is in a state of high tension or distraction, including reducing the speech rate of the voice broadcast, increasing the lead time of the operation prompts, and expanding the display area of ​​the visual prompts to include the display area in the driver's current gaze direction. For example, when it is detected that the driver's gaze is always fixed on the windshield, the system switches the steering wheel angle prompt from the instrument panel to the HUD head-up display to ensure that the prompt information is always within the driver's field of vision. By sensing the driver's physiological state and adaptively adjusting the output method of the guidance information, the efficiency of conveying guidance information in a state of driver tension is improved, avoiding missing key operation prompts due to the driver's lack of concentration, and further improving the success rate of escape from difficulties.

[0041] In a preferred embodiment of practical application, firstly, after the vehicle becomes stuck in terrain such as sand or mud, the driver triggers a distress guidance request via the touchscreen or voice command of the in-vehicle infotainment system. Upon responding to this request, the in-vehicle sensor network is activated to simultaneously acquire the vehicle's current attitude information (pitch angle, roll angle, yaw angle), tire slip ratio information (calculated via wheel speed and vehicle speed sensors), and surrounding terrain information (images and distance data acquired via surround-view cameras, ultrasonic radar, and lidar). This information is then fused into comprehensive vehicle perception information. Secondly, based on the comprehensive vehicle perception information, the type of entrapment is determined, such as identifying a single tire stuck, diagonally suspended tires, or chassis bottoming out. Then, based on the determined entrapment type, a real-time tire dynamics model is used to perform a drivability analysis on the terrain information within the comprehensive vehicle perception information. Specifically, the real-time tire dynamics model calculates the adhesion vector field of each tire under the current terrain based on terrain information. Drivability analysis, based on the adhesion vector field and the type of entrapment, determines the vehicle's passable direction and traction strategy, generating a time-sequential traction control sequence. This sequence includes operational instructions such as steering wheel angle, accelerator pedal opening, brake pedal action, and gear selection, clearly defining the execution order and time window for each instruction. Finally, the traction control sequence is transformed into driver-oriented guidance information through a multimodal human-machine interface. For example, the instrument panel displays a diagram indicating the direction and amplitude of steering wheel rotation; a voice prompt instructs the driver to lightly press the accelerator to 30% and hold for 2 seconds; an accelerator pedal depress depth progress bar is displayed on the HUD; and steering wheel vibration indicates the driver is about to perform the next operation. The guidance information is continuously updated until the vehicle successfully escapes the entrapment or the driver manually exits the entrapment.

[0042] The system receives the entrapment type assessment result and comprehensive vehicle perception information, and combines it with real-time tire dynamics models and terrain accessibility analysis to generate a time-seriesed entrapment operation guidance sequence. This sequence is then converted into visual guidance information (instrument panel display, HUD display), auditory guidance information (voice broadcast), and tactile guidance information (steering wheel vibration) via a multimodal interaction interface, and simultaneously output to the driver.

[0043] In some embodiments, the vehicle's comprehensive perception information includes surrounding terrain information; step S104, which generates a time-seriesd escape operation guidance sequence, includes: Step 2.1: Based on the surrounding terrain information, construct a local digital elevation model of the tire-ground contact interface; Here, this step extracts a local digital elevation model (DEM) of the tire-ground contact interface area from the terrain information surrounding the vehicle. First, based on the vehicle's current attitude and the positions of each tire, the spatial extent of the tire-ground contact interface is determined. Then, terrain elevation data within this range is extracted from 3D point cloud data acquired by LiDAR or binocular vision to construct the local digital elevation model. This model represents the terrain undulation features of the tire-ground contact area in a gridded form, and the grid resolution can be dynamically adjusted according to the terrain complexity.

[0044] A Digital Elevation Model (DEM) is a digital simulation of ground topography using limited terrain elevation data, representing a digital representation of the terrain's surface morphology. In this invention, a local DEM specifically refers to an elevation data grid within the tire-ground contact interface area, used to describe the microscopic undulations of the terrain in that region. By constructing a local DEM, a refined description of the terrain features in the tire-ground contact area is achieved, providing high-resolution terrain input data for the subsequent accurate calculation of the adhesion vector field.

[0045] In a preferred embodiment, constructing a local digital elevation model (DEM) of the tire-ground contact interface further includes: updating the DEM using a time-series weighted average method based on multiple terrain data collection records of the same geographical location from the vehicle's historical driving trajectory, where more recent collection records have greater weight; simultaneously, when changes in environmental conditions (such as rainfall, snowfall, or icing) are detected, the ground physical property parameters of the DEM are corrected, with correction coefficients obtained from a preset parameter mapping table based on the type and degree of environmental condition change. For example, after rainfall, a hard crust may form on the sandy surface, and the correction coefficient will reduce the ground deformability parameter of that area. By integrating historical terrain data with information on changes in environmental conditions, the accuracy and timeliness of the local DEM in dynamic environments are improved, enabling the escape strategy to adapt to real-time changes in terrain conditions.

[0046] Step 2.2: Input the terrain information represented by the local digital elevation model into the real-time tire dynamics model to calculate the adhesion vector field of each tire under the current terrain; Here, terrain information represented by a local digital elevation model is used as input parameters and imported into a real-time tire dynamics model for simulation calculations. Terrain information includes changes in ground elevation, ground material type (sand, mud, rock, ice, etc.), and the corresponding coefficient of friction. The real-time tire dynamics model, based on the Pacejka magic formula or the Dugoff tire model, calculates the adhesion vector field of each tire under the current terrain, incorporating parameters such as vertical load, sideslip angle, and slip ratio. The adhesion vector field includes the amplitude and direction distribution of adhesion force for each tire in the longitudinal and lateral directions.

[0047] Here, using the Pacejka Magic Formula tire model as an example, the formula for calculating the tire's longitudinal force is as follows:

[0048] in, For the longitudinal force of the tire, For slip ratio, For stiffness factor, For shape factor, Peak factor (with vertical load) and adhesion coefficient (related) This refers to the curvature factor. The values ​​of each factor are determined by the terrain information and ground physical properties in the local digital elevation model.

[0049] This step achieves a precise mapping from terrain geometry information to tire mechanics, overcoming the limitations of traditional methods that rely solely on slip ratio thresholds and cannot obtain the direction and amplitude distribution of adhesion force, thus providing an accurate physical basis for subsequent optimal force distribution schemes.

[0050] In a preferred embodiment, calculating the adhesion vector field of each tire under the current terrain further includes: acquiring real-time state parameters of the vehicle chassis suspension system, including shock absorber compression stroke, spring stiffness change rate, and vehicle center of gravity offset; inputting the real-time state parameters into a real-time tire dynamics model to correct the vertical load distribution of each tire, thereby updating the adhesion vector field. Specifically, when the vehicle tilts in a ditch, the vehicle's center of gravity shifts to the lower side, resulting in an increase in the vertical load on the lower tires and a decrease in the vertical load on the higher tires. The system uses suspension displacement sensors to acquire the compression of each shock absorber in real time, calculates the vertical load distribution after the center of gravity shift, and recalculates the adhesion vector field accordingly to ensure that the force distribution scheme is consistent with the actual vehicle posture. By introducing real-time suspension state parameters to correct the vertical load distribution, the calculation accuracy of the adhesion vector field in the vehicle tilt state is improved, making the traction force distribution scheme more consistent with the actual vehicle dynamics.

[0051] Step 2.3: Based on the type of entrapment, use a real-time tire dynamics model to analyze the passability of the surrounding terrain information, and determine the passable direction of the vehicle and the extrication operation strategy.

[0052] Here, this step combines the type of entrapment with the adhesion vector field to perform a drivability analysis. The core of the drivability analysis is to determine whether the vehicle can obtain sufficient total traction in each direction to overcome resistance and move. For the front wheel entrapment type, the drivable direction is usually backward; for the side-tilt entrapment type, the drivable direction is downward; for the diagonally suspended type, the drivable direction is limited to the diagonal direction with adhesion. The extrication strategy includes whether vehicle body swaying is involved and the expected swaying amplitude. For example, for a vehicle stuck in sand, the strategy might include back-and-forth swaying to gain inertia; for a vehicle stuck in mud, the strategy might include slowly releasing power to avoid digging. This step organically combines the entrapment type with terrain drivability analysis, ensuring that the determination of the drivable direction considers both the physical constraints of the terrain and the empirical rules of the entrapment type, achieving a complementarity between empirical knowledge and physical models.

[0053] In a preferred embodiment, determining the vehicle's passable direction and traction strategy further includes: acquiring the vehicle's current fuel / electricity status and remaining driving range; matching the fuel / electricity status with the estimated energy consumption of the traction strategy; if the estimated energy consumption of the traction strategy exceeds a preset energy consumption safety threshold, then prioritizing the direction with the lowest estimated energy consumption among multiple candidate traction directions, and adding energy-saving operation instructions to the traction strategy, including reducing the power of unnecessary loads such as air conditioning and prioritizing the use of low-power traction modes. The estimated energy consumption is calculated based on the simulation results of the traction operation sequence using a real-time tire dynamics model. By introducing energy consumption constraints to optimize the traction strategy, the risk of the vehicle being unable to reach a safe location due to excessive energy consumption during traction operations in remote areas is avoided, improving the practicality of the traction strategy in extreme environments.

[0054] Step 2.4: With the goal of maximizing the effective traction of the vehicle in the passable direction, solve for the optimal tire force distribution scheme based on the adhesion vector field.

[0055] Here, this step aims to maximize the effective traction of the vehicle in the passable direction, constructing a constrained optimization problem. The optimization variables are the drive / braking force distribution ratio of each tire, and the constraints include that the slip ratio of each tire does not exceed the optimal slip ratio corresponding to its peak adhesion force, the force of each tire does not exceed the upper limit of the amplitude of the adhesion vector field, and the yaw moment of the vehicle does not exceed the stability threshold. The solution method can employ numerical optimization algorithms such as sequential quadratic programming (SQP) and interior-point methods, or heuristic search methods based on genetic algorithms or particle swarm optimization.

[0056] The mathematical expression of the optimization problem is as follows:

[0057]

[0058] in, The angle for the passable direction. and The first The longitudinal and lateral forces of each tire. For the first The optimal slip ratio corresponding to the peak adhesion of each tire. For the vehicle's yaw moment, This is the stability threshold for yaw moment.

[0059] This step transforms the problem of allocating traction force into a clear mathematical optimization problem. With maximizing the effective traction force as the objective function and physical constraints as the boundary conditions, the optimal force allocation scheme obtained theoretically guarantees the efficiency and safety of the traction operation.

[0060] In a preferred embodiment, solving for the optimal tire force distribution scheme further includes: establishing a multi-objective optimization model, where the optimization objectives include maximizing effective traction, minimizing tire wear, and minimizing energy consumption; and using a weighted summation method to transform the multi-objective optimization into a single-objective optimization, wherein the weight coefficients of each objective are automatically adjusted according to the current scenario: when the vehicle is on dangerous terrain (such as the edge of a cliff), the weight of effective traction is increased to ensure safety; when the vehicle is in a normal stuck-out scenario, the weights of tire wear and energy consumption are increased to reduce the cost of getting out of trouble. The adjustment rules of the weight coefficients are based on a preset scenario-weight mapping table, which is derived from expert experience and historical escape data statistics. Through multi-objective optimization and scenario-adaptive weight adjustment, a balance is achieved between safety and economy in the force distribution scheme, avoiding extreme solutions that may be caused by single-objective optimization (such as excessive tire wear in pursuit of getting out of trouble).

[0061] Step 2.5: Map the optimal tire force distribution scheme into a time-sequential command sequence of steering wheel angle, accelerator pedal opening, and brake pedal action.

[0062] Here, this step maps the force distribution scheme to driver-executable commands. Steering wheel angle mapping is based on the vehicle's kinematics model, calculating the required front wheel angle according to the lateral force distribution scheme; accelerator pedal opening mapping is based on the engine / motor torque characteristic curve, converting the target driving force into a pedal opening percentage; brake pedal action mapping is based on the hydraulic characteristics of the braking system, converting the target braking force into brake pedal travel or braking pressure value. Each command is arranged in a time sequence, forming a time-sequential command sequence, clearly defining the execution time and duration of each command. This transforms the abstract mechanical optimization results into concrete operational commands that the driver can directly understand and execute, achieving a seamless connection from the physical model to human-machine interaction, making the traction guidance scheme practically operable.

[0063] In a preferred embodiment, the time-sequential command sequence mapped to steering wheel angle, accelerator pedal opening, and brake pedal action further includes: acquiring the vehicle's steering system response delay characteristics and throttle response delay characteristics; and setting an advance amount for each operation command in the time-sequential command sequence based on the response delay characteristics, so that when the driver executes the operation according to the command, the vehicle's actual response is exactly aligned with the expected timing of the traction-avoidance strategy. For example, if the vehicle's throttle response delay is 200 milliseconds, the execution time of the accelerator pedal command is advanced by 200 milliseconds to ensure that the driving force accurately reaches the tires when it needs to be applied. By compensating for the vehicle's response delay, the timing deviation between the command and the actual response caused by actuator lag is eliminated, improving the execution accuracy and traction-avoidance success rate of the traction-avoidance operation guidance sequence on a real vehicle.

[0064] In application, as an example, the following steps can be included: First, extract surrounding terrain information from the vehicle's comprehensive perception information. Based on the position of each tire and the vehicle's attitude, determine the spatial range of the tire-ground contact interface. Extract elevation data within this range from the 3D point cloud data acquired by LiDAR, and construct a local digital elevation model (DEM) of the tire-ground contact interface. This model describes the terrain undulation characteristics of the contact area in a gridded form. Second, input the terrain information represented by the local DEM (including ground elevation changes, ground material type, and corresponding adhesion coefficient) into the real-time tire dynamics model. The real-time tire dynamics model, based on the Pacejka magic formula, combines parameters such as the vertical load, sideslip angle, and slip ratio of each tire to calculate the adhesion force vector field of each tire under the current terrain, including the longitudinal and lateral adhesion force amplitude and direction distribution of each tire. Then, based on the type of entrapment, use the real-time tire dynamics model to perform a drivability analysis of the surrounding terrain information. For example, for front wheel ruts, the system analyzes whether the adhesion of each tire in the rearward direction is sufficient to overcome resistance, determining the passable direction as rearward. For diagonal suspension, it analyzes the diagonal direction with adhesion, determines the passable direction, and formulates an escape strategy, including whether vehicle body swaying is involved and the expected swaying amplitude. Furthermore, with the optimization objective of maximizing the vehicle's effective traction in the passable direction, a constrained optimization problem is constructed based on the adhesion vector field. The constraint condition is that the slip ratio of each tire does not exceed the optimal slip ratio corresponding to its peak adhesion. A sequential quadratic programming algorithm is used to solve for the optimal tire force distribution scheme. Finally, the optimal tire force distribution scheme is mapped to a sequence of executable operation commands for the driver. The steering wheel angle is mapped based on the vehicle kinematics model, the accelerator pedal opening is mapped based on the engine torque characteristic curve, and the brake pedal action is mapped based on the hydraulic characteristics of the braking system. Each command is arranged in a time sequence, forming a time-series command sequence, clearly defining the execution time and duration of each command.

[0065] For example, a vehicle gets stuck in sand, with each of its four tires experiencing different terrain conditions. A local digital elevation model (DEM) describes the terrain undulations of the contact area between each tire and the ground. A real-time tire dynamics model calculates the adhesion vector field for each tire based on the DEM, where the arrow length represents the adhesion amplitude and the arrow direction represents the adhesion direction. The force distribution optimization module aims to maximize the effective traction in the passable direction, solving for the optimal tire force distribution scheme, which is ultimately mapped to a time-sequenced command sequence of steering wheel angle, accelerator pedal opening, and brake pedal action.

[0066] In some embodiments, when the traction control strategy includes vehicle body swaying, the step of generating a time-sequential traction control guidance sequence in step S104 further includes: Step 3.1: When the escape strategy involves vehicle body swaying, determine the expected swaying amplitude based on the type of entrapment; Here, this step is triggered when the vehicle's swaying is involved in the extrication strategy. The expected swaying amplitude is determined based on the type of entrapment. For example, for vehicles stuck in sand, the expected swaying amplitude is typically larger (e.g., 30%-50% of the vehicle's wheelbase) to utilize inertia to overcome sand resistance; for vehicles stuck in mud, the expected swaying amplitude is typically smaller (e.g., 10%-20% of the vehicle's wheelbase) to avoid excessive digging that could deepen the entrapment. The expected swaying amplitude can be determined based on a pre-defined entrapment type-swaying amplitude mapping table, derived from real-vehicle test data and expert experience. This step determines the swaying amplitude based on the type of entrapment, avoiding ineffective operations or deepening the entrapment due to blind swaying, and providing a clear strategic basis for the swaying operation.

[0067] In a preferred embodiment, determining the expected swing amplitude further includes: acquiring terrain scan data under the vehicle chassis; identifying whether there are hard obstacles under the vehicle chassis using the terrain scan data; if hard obstacles exist, limiting the upper limit of the expected swing amplitude based on the position and height of the hard obstacles to prevent the wheels from colliding with the hard obstacles during the swing. The terrain scan data is acquired by ultrasonic sensors or lidar installed under the vehicle chassis. By scanning the terrain under the chassis and identifying obstacles, vehicle damage or failure to escape due to wheel collisions with hard obstacles during the swing is avoided, improving the safety of the swing operation.

[0068] Step 3.2: Based on the adhesion vector field, calculate the transient ground imprint changes of each tire under the expected sway amplitude; This step simulates the transient ground contact patch changes during vehicle swaying. When the vehicle sways back and forth, the pitch angle changes periodically, causing dynamic changes in the vertical load distribution of each tire, which in turn affects the contact area (ground contact patch) between the tire and the ground. Based on the adhesion vector field, combined with the vehicle kinematic model and suspension dynamic characteristics, the transient ground contact patch changes of each tire under the expected sway amplitude are calculated. The calculation results include curves showing the changes in the length, width, and area of ​​the ground contact patch over time.

[0069] Specifically, the length of the transient grounding imprint The simplified calculation is as follows:

[0070] in, The free radius of the tire. For tires at all times The vertical deformation is determined by the vertical load. and tire stiffness Decide: .

[0071] This step enables dynamic modeling of the tire-ground contact state during the oscillation process, providing accurate time-series data for determining the subsequent adhesion peak window period.

[0072] In a preferred embodiment, calculating the transient contact patch change of each tire further includes: acquiring real-time tire pressure data; and correcting the tire stiffness parameters based on the real-time tire pressure data to update the contact patch calculation results. Specifically, when the tire pressure is lower than the standard value, the tire stiffness decreases, and the contact patch area increases under the same vertical load. The system adjusts the tire stiffness coefficient according to the ratio of the real-time tire pressure to the standard tire pressure to ensure that the contact patch calculation results are consistent with the actual situation. By introducing real-time tire pressure data to correct the tire stiffness parameters, the accuracy of the contact patch calculation under tire pressure variation conditions is improved, avoiding peak window period judgment errors caused by abnormal tire pressure.

[0073] Step 3.3: Determine the peak adhesion window period for each tire based on the transient grounding imprint changes; Here, based on the transient ground contact patch variation curve, the time window for maximum adhesion of each tire is determined. The adhesion peak window period refers to the time interval where the ground contact patch area is largest and the vertical load distribution is most favorable for generating adhesion. By analyzing the ground contact patch variation curve, the time point when the ground contact patch area reaches its local maximum value is identified, and the interval where the ground contact patch area is not less than 90% of the peak value is taken as the adhesion peak window period, centered on this time point. For multi-tire cases, the peak window periods of each tire are combined to determine the globally optimal operating timing window; thus, the optimal timing for adhesion during oscillation is accurately captured, allowing the accelerator pedal operation to be applied at the moment when the tire adhesion is maximum, maximizing the effective traction output.

[0074] In a preferred embodiment, determining the peak adhesion window period for each tire further includes: acquiring the damping characteristic parameters of the vehicle suspension system; predicting the response delay of the suspension during the oscillation process based on the damping characteristic parameters, and correcting the start and end times of the peak adhesion window period accordingly. For example, when the suspension damping is large, the vehicle body attitude change lags behind the wheel movement. The system shifts the peak adhesion window period backward to compensate for the suspension response delay, ensuring that the accelerator pedal operation is aligned with the actual peak adhesion time. By compensating for the suspension response delay, the error in determining the peak window period caused by the lag in vehicle body attitude change is eliminated, making the operation timing more closely match the actual dynamic response of the vehicle.

[0075] Step 3.4: Align the peak accelerator pedal opening command in the timing command sequence with the peak adhesion window period in terms of timing; Here, the execution time of the peak accelerator pedal opening command is aligned with the peak adhesion window. When generating the timing-sequential command sequence, the peak accelerator pedal opening command value (i.e., the moment of maximum pedal depth) is set within the peak adhesion window, synchronizing the peak driving force with the peak adhesion. Simultaneously, the duration of the peak accelerator pedal command is adjusted based on the duration of the peak adhesion window, ensuring the driver continuously applies effective driving force throughout the entire peak window. This step, through timing alignment, achieves optimal matching between driving force output and tire adhesion changes, avoiding tire spin caused by excessive driving force when adhesion is insufficient, and significantly improving the efficiency of swaying and getting out of trouble.

[0076] In a preferred embodiment, aligning the peak accelerator pedal opening command with the peak adhesion window in terms of timing further includes: acquiring the driver's real-time reaction time data, which is calculated based on the average delay time from the issuance of guidance information to the driver's commencement of operation from the driver's historical operation records; and setting an advance amount for the peak accelerator pedal opening command in the time-sequential command sequence based on the real-time reaction time data, so that when the driver performs the operation according to the guidance information, the actual peak pedal pressure falls precisely within the peak adhesion window. For example, if the driver's average reaction time is 300 milliseconds, the system will issue the guidance prompt for the peak accelerator pedal opening command 300 milliseconds in advance. By compensating for the driver's reaction delay, the command execution deviation caused by human delay is eliminated, ensuring that the actual effect of the driver's operation is precisely aligned with the optimal timing planned by the system, further improving the success rate of extrication from trouble.

[0077] In application, this step can be implemented through the following example: First, when the extrication strategy involves vehicle body swaying, the expected sway amplitude is determined based on the type of entrapment. For example, for a vehicle stuck in sand, the expected sway amplitude is 40% of the vehicle's wheelbase to utilize inertia to overcome sand resistance. Second, based on the adhesion vector field, combined with the vehicle's kinematic model and suspension dynamic characteristics, the vehicle's motion process under the expected sway amplitude is simulated, and the transient ground contact patch changes of each tire are calculated, including the changes in the length, width, and area of ​​the ground contact patch over time. Then, the ground contact patch change curves are analyzed to identify the time point when the ground contact patch area reaches a local maximum value. Using this time point as the center, the interval where the ground contact patch area is not less than 90% of the peak value is taken as the adhesion peak window period. For multi-tire situations, the peak window periods of each tire are combined to determine the globally optimal operation timing window. Finally, the peak command value of the accelerator pedal opening in the timing command sequence is set at the moment when the adhesion of each tire is at its maximum, so that the peak driving force and the peak adhesion are synchronized. At the same time, the duration of the accelerator pedal peak command is adjusted according to the duration of the adhesion peak window to ensure that the driver continuously applies effective driving force throughout the peak window.

[0078] In some embodiments, the step of generating a time-sequential escape operation guidance sequence in step S104 further includes: Step 4.1: Based on the surrounding terrain information, identify the longitudinal slope, lateral slope and bottom flatness of the pit where the vehicle is currently located.

[0079] Here, this step extracts the morphological feature parameters of the pit from the surrounding terrain information. By analyzing 3D point cloud data acquired by LiDAR or binocular vision, the boundary contour of the pit where the vehicle is currently located is identified, and the longitudinal slope (the slope of the pit wall along the vehicle's direction of travel), lateral slope (the slope of the pit wall perpendicular to the vehicle's direction of travel), and bottom flatness (the surface undulation variance of the pit bottom area) are calculated. The identification method can employ point cloud-based normal vector estimation and region growing algorithms, or deep learning-based point cloud semantic segmentation networks. This achieves a quantitative description of the pit's morphological features, providing accurate geometric parameter input for subsequent prediction of chassis scratch risk.

[0080] In a preferred embodiment, identifying the longitudinal slope, lateral slope, and bottom flatness of the pit where the vehicle is currently located further includes: acquiring ultrasonic sensor data from beneath the vehicle chassis; measuring the minimum distance between the chassis and the ground using the ultrasonic sensor data; and jointly analyzing the minimum distance with the longitudinal and lateral slopes of the pit to determine whether the chassis has already contacted or is about to contact the ground. When the minimum distance is lower than a preset safety threshold, the system prioritizes triggering the chassis protection strategy, rather than waiting for the terrain recognition results to be fully processed. Through real-time monitoring and joint analysis of chassis ultrasonic sensor data, early warning of chassis scratch risks is achieved, shortening the response time of the protection strategy and avoiding chassis damage caused by data processing delays.

[0081] Step 4.2: When at least one of the longitudinal slope, lateral slope, or pit bottom flatness exceeds a preset threshold, predict the chassis scraping risk area that will occur during the extrication operation; Here, this step triggers chassis scraping risk prediction when the morphological parameters of the ditch exceed a preset threshold. Based on a three-dimensional geometric model of the vehicle chassis, and combined with extrication strategies (such as the expected vehicle trajectory and changes in vehicle posture), the system simulates the relative positional relationship between the chassis and the terrain during the extrication process in a simulation environment, predicting chassis scraping risk areas. Risk areas include the front of the chassis (insufficient approach angle), the middle of the chassis (insufficient breakover angle), the rear of the chassis (insufficient departure angle), and the locations of key components such as the oil pan and transmission housing. Through simulation prediction, the system achieves a pre-assessment of chassis scraping risks, enabling the identification of potential risks and the development of protective strategies before the extrication operation begins, thus avoiding chassis damage during the actual extrication process.

[0082] In a preferred embodiment, predicting the chassis scraping risk area generated during the escape operation further includes: acquiring a three-dimensional structural model of the vehicle chassis, the three-dimensional structural model including the precise location, size, and material information of each chassis component; based on the three-dimensional structural model, classifying the chassis scraping risk areas according to the importance of the components into high-risk areas (engine oil pan, transmission housing, etc.) and low-risk areas (chassis skid plate, exhaust pipe cover, etc.); for high-risk areas, setting stricter protection thresholds, and prioritizing the protection of high-risk areas in the chassis protection operation commands. This chassis component-based graded protection strategy makes the chassis protection operation commands more targeted, prioritizing the protection of critical components, reducing unnecessary operational restrictions while ensuring safety, and improving escape efficiency.

[0083] Step 4.3: Based on the chassis scraping risk area, insert chassis protection operation instructions into the time-sequential traction operation guidance sequence; chassis protection operation instructions include reducing the accelerator pedal opening to reduce the vehicle pitch angle, or adjusting the steering wheel angle to avoid the scraping risk area; Here, this step inserts chassis protection operation commands into the time-sequential traction control sequence based on the location and severity of the chassis scraping risk area. Specifically, when the risk area is located at the front of the chassis, a command to reduce the accelerator pedal opening is inserted during the forward phase of the traction control sequence to reduce the vehicle's pitch angle and prevent the front bumper or oil pan from contacting the ground; when the risk area is located on the side of the chassis, the system inserts a command to adjust the steering wheel angle to guide the vehicle around the scraping risk area. The insertion position and parameters of the chassis protection operation commands are determined based on the specific location of the risk area and the temporal relationship of the traction control sequence. This step seamlessly integrates the chassis protection strategy into the traction control sequence, allowing the driver to automatically receive chassis protection guidance while performing traction control operations, avoiding the risk of neglecting chassis safety due to focusing on traction control.

[0084] In a preferred embodiment, the chassis protection operation command further includes: when the chassis scraping risk area is located at the location of critical components of the vehicle chassis (such as the engine oil pan or transmission housing), adding a suspension height adjustment command to the chassis protection operation command, thereby controlling the air suspension or hydraulic suspension to raise the vehicle height to its maximum value to increase the chassis ground clearance; simultaneously, during the execution of the chassis protection operation command, the vehicle speed is limited to a preset low speed range to avoid excessive chassis impact force due to excessive vehicle speed. This combined protection strategy of suspension height adjustment and vehicle speed limitation provides maximum chassis protection for the vehicle in scenarios with high chassis scraping risk, reducing the risk of damage to critical components.

[0085] In application, this can be achieved through the following exemplary embodiments: First, extract surrounding terrain information from the vehicle's comprehensive perception information. By analyzing the 3D point cloud data acquired by LiDAR, and using point cloud-based normal vector estimation and region growing algorithms, identify the boundary contour of the pit the vehicle is currently in, and calculate the longitudinal slope, lateral slope, and bottom flatness of the pit. Second, when at least one of the longitudinal slope, lateral slope, or bottom flatness exceeds a preset threshold, trigger chassis scraping risk prediction. Based on the 3D geometric model of the vehicle chassis, combined with the escape operation strategy, simulate the relative positional relationship between the chassis and the terrain during the escape process in a simulation environment, and predict the chassis scraping risk area, including the front of the chassis (insufficient approach angle), the middle of the chassis (insufficient passing angle), and the corresponding positions of key components. Finally, based on the location and severity of the chassis scraping risk area, insert chassis protection operation instructions into the time-sequential escape operation guidance sequence. For example, when the risk area is located at the front of the chassis, a command to reduce the accelerator pedal opening to 15% is inserted during the forward phase of the escape sequence to reduce the vehicle pitch angle; when the risk area is located on the side of the chassis, a command to turn the steering wheel 15 degrees to the right is inserted to guide the vehicle around the risk area.

[0086] For example, if a vehicle gets stuck in a pit with a steep front wall, the terrain is used to identify the pit's longitudinal slope, lateral slope, and bottom flatness. When the slope exceeds a preset threshold, chassis scraping risk prediction simulates the relative position of the chassis and terrain during the escape process in a simulation environment, identifying the risk area for front chassis scraping. Chassis protection command generation inserts a protection command to reduce the accelerator pedal opening into the time-sequential escape operation guidance sequence to reduce the vehicle's pitch angle and avoid chassis scraping.

[0087] In some embodiments, the comprehensive vehicle perception information includes tire slip ratio information and surrounding terrain information; the operation of determining the vehicle's entrapment type in step S102 includes the following process: Step 5.1: Based on the slip ratio information of each tire and the surrounding terrain information, determine whether the vehicle has the ability to get out of trouble on its own; This step analyzes the slip ratio information of each tire and the surrounding terrain information to determine whether the vehicle has the capability to extricate itself from a difficult situation using its own power. It checks whether at least one tire has usable traction (i.e., the slip ratio does not exceed a preset threshold) and, based on the surrounding terrain information, determines whether the tire with usable traction is in an effective position where driving force can be applied. For example, if the slip ratio of all tires exceeds 90%, the vehicle does not have the capability to extricate itself from a difficult situation and requires external assistance or other extrication methods. This step enables a rapid assessment of the vehicle's ability to extricate itself from a difficult situation, avoiding ineffective attempts to extricate vehicles that do not possess the capability, thus saving time and energy.

[0088] In a preferred embodiment, determining whether a vehicle possesses the conditions for autonomous extrication further includes: acquiring the vehicle's four-wheel drive system status information, including the differential lock's locking status, the transfer case's gear position, and torque distribution ratio; assessing the vehicle's currently available drive force distribution capability based on the four-wheel drive system status information; if the differential lock is not locked but the vehicle has the capability to lock, adding a suggestion to lock the differential lock to the determination result, and reassessing the autonomous extrication conditions after locking. This guides the driver to fully utilize the vehicle's four-wheel drive function, improving the success rate of extrication. By comprehensively analyzing the four-wheel drive system status, not only is the autonomous extrication capability under current conditions determined, but the driver is also proactively prompted with possible vehicle function adjustments, expanding the dimensions for determining autonomous extrication conditions.

[0089] Step 5.2: If available, model the driver's skill in getting out of trouble, reaction speed and operation accuracy based on the vehicle's historical driving data to obtain the current driver's ability to get out of trouble assessment results; Here, this step assesses the driver's ability to extricate themselves from difficult situations when the vehicle is capable of doing so autonomously. The system extracts the driver's operational records in these situations from the vehicle's historical driving data, including the number of extrication maneuvers, the accuracy of each maneuver, the response delay from guidance information issuance to execution, steering wheel angle deviation, and pedal opening deviation. Based on this data, the system constructs a profile of the driver's extrication ability and uses a weighted scoring method or machine learning model to comprehensively evaluate the driver's proficiency, reaction speed, and operational precision, resulting in an assessment of their extrication ability. This step achieves a quantitative assessment of the driver's extrication ability, providing a data foundation for matching the complexity of subsequent operations with the driver's capabilities, and is a crucial step in achieving personalized extrication guidance.

[0090] In a preferred embodiment, modeling the driver's proficiency, reaction speed, and operational accuracy in escaping difficult situations further includes: acquiring the driver's driving behavior data, including following distance, lane change frequency, number of emergency brakings, and cornering speed in daily driving; and based on the driving behavior data, mapping the driver's daily driving style to the escaping ability assessment through transfer learning to establish an escaping ability prediction model. Specifically, drivers who are cautious and stable in daily driving may have slower reaction speeds but higher operational accuracy in escaping scenarios; while drivers with aggressive daily driving styles may have faster reaction speeds but lower operational accuracy. The transfer learning model is trained using the correlation between labeled escaping ability data and daily driving behavior data. By introducing daily driving behavior data into the escaping ability assessment through transfer learning, the problem of inaccurate assessments caused by novice drivers lacking historical escaping data is solved, improving the generalization and accuracy of the escaping ability assessment.

[0091] Step 5.3: Based on the total number of operation instructions in the escape operation guidance sequence, the timing coordination accuracy between each operation instruction, and the types of operation parameters that need to be adjusted in coordination, determine the operation complexity required for autonomous escape. This step quantifies the operational complexity of the traction control sequence from three dimensions. The total number of commands reflects the number of steps in the traction control operation; more steps mean more operations the driver needs to remember and execute. The timing precision between commands reflects the time requirements of the traction control operation; higher precision narrows the driver's operational window. The types of operational parameters requiring coordinated adjustment reflect the multi-task coordination requirements of the traction control operation; more parameters (such as simultaneously adjusting the steering wheel, accelerator, and brake) increase the driver's workload. The three dimensions are weighted and combined to obtain the operational complexity evaluation value. This step achieves a multi-dimensional quantitative assessment of the complexity of traction control operations, objectively measuring the driver's operational capabilities required for a specific traction control strategy, and providing comparable quantitative indicators for matching driver capabilities.

[0092] In a preferred embodiment, determining the operational complexity required for autonomous escape further includes: acquiring the environmental risk level of the escape scenario, which is assessed based on hazardous factors (such as cliffs, deep water, steep slopes, etc.) in the terrain information surrounding the vehicle; using the environmental risk level as a correction factor for operational complexity, automatically increasing the assessed value of operational complexity when the environmental risk level is high, making it more likely that automatic intervention is needed in dangerous environments, thus reducing the driver's operational stress and psychological burden in dangerous environments. By introducing the environmental risk level to correct operational complexity, a more conservative human-machine collaboration strategy can be adopted in dangerous scenarios, prioritizing the safety of the driver and vehicle.

[0093] Step 5.4: Match the operational complexity with the assessment results of the ability to get out of trouble. If the operational complexity exceeds the driver's ability to get out of trouble, then the type of entrapment is determined to be one that requires automatic intervention. Here, this step compares the complexity of the escape operation with the driver's operational capabilities. If the complexity exceeds the driver's ability, the entrapment type is determined to require automatic intervention, meaning the vehicle control system actively intervenes to perform all or part of the escape operation, rather than simply providing instructions. If the complexity is within the driver's capabilities, it is determined to be a type of entrapment that can be escaped through guidance, and the entrapment continues in guided mode. This step achieves intelligent decision-making for human-machine collaborative entrapment strategies, automatically selecting the optimal human-machine collaborative mode based on the matching result between driver capabilities and operational complexity, maximizing driver participation while ensuring safety.

[0094] In a preferred embodiment, determining whether a entrapment requires automatic intervention further includes: determining the level of automatic intervention based on the degree to which the operational complexity exceeds the driver's ability to extricate themselves; the automatic intervention levels include: Level 1, intervening only in critical operations (such as automatically controlling the steering wheel angle while the driver controls the accelerator); Level 2, intervening in all operations but retaining the driver's exit authority (the system performs all extrication operations, and the driver can take over at any time); Level 3, fully automatic intervention and locking the driver's operation (the system takes over all control, prohibiting the driver from operating until extrication is completed or the driver issues an emergency exit request). The selection of the automatic intervention level is based on the degree of difference between operational complexity and driver ability; the greater the difference, the higher the automatic intervention level. Through a multi-level automatic intervention strategy, a progressive human-machine collaborative extrication solution from guidance to full automation is achieved, ensuring safety while preserving the driver's sense of participation and control.

[0095] As an example, the following steps may be included in the application process: First, based on the tire slip ratio information and surrounding terrain information, it is determined whether the vehicle has the capability to extricate itself from a difficult situation independently. If the slip ratio of all tires exceeds 90%, the vehicle is deemed unable to extricate itself independently, and the driver is advised to seek external assistance. Second, if the vehicle has the capability to extricate itself independently, the driver's extrication operation records are extracted from the vehicle's historical driving data, including the number of extrication operations, the accuracy of step execution, response delay, steering wheel angle deviation, pedal opening deviation, etc. A weighted scoring method is used to comprehensively evaluate the driver's extrication operation proficiency, reaction speed, and operational precision, resulting in an extrication operation capability assessment result. Then, the operational complexity of the extrication operation guidance sequence is quantified from three dimensions: the total number of operation instructions reflects the number of steps; the timing coordination accuracy between operation instructions reflects the time coordination requirements; and the types of operation parameters requiring coordinated adjustment reflect the multi-task coordination burden. The three dimensions are weighted and synthesized to obtain the operational complexity evaluation value. Finally, the operational complexity is matched and compared with the driver's extrication operation capability assessment result. If the complexity of the operation exceeds the driver's ability, the type of entrapment is determined to be one that requires automatic intervention, and the vehicle control system will actively intervene to perform all or part of the entrapment operation; if the complexity of the operation is within the driver's ability, the entrapment will continue in guided mode.

[0096] In other words, based on historical vehicle driving data, the driver's proficiency, reaction speed, and operational precision in escaping difficult situations are modeled to obtain an assessment result of their escaping capability. Based on the escaping operation guidance sequence, the operational complexity is quantified from three dimensions: the total number of operation commands, the accuracy of timing coordination, and the types of operation parameters. The operational complexity is compared with the escaping capability assessment result. If the operational complexity exceeds the driver's ability, it is determined to be a type of entrapment requiring automatic intervention, and the vehicle control system actively intervenes to execute the escaping operation.

[0097] In some embodiments, the present invention further includes the following steps: Step 6.1: Based on historical vehicle driving data, establish a probabilistic model to characterize the deviation distribution of the driver when executing various types of operation commands; Here, the deviations between the actual and target parameter values ​​when the driver executes various types of operation commands are extracted from historical driving data, including steering wheel angle deviation, accelerator pedal opening deviation, and brake pedal travel deviation. Statistical analysis is performed on the deviation data to estimate the probability distribution of the deviations (e.g., normal distribution, Gaussian mixture distribution), obtaining the parameters of the probabilistic model. This model can predict the probability that the actual output value will fall within a certain deviation range when the driver executes a specific operation command. This step, through probabilistic modeling, achieves a quantitative description of the uncertainty of driver operation, providing a statistical basis for setting the subsequent operation tolerance range.

[0098] In a preferred embodiment, establishing a probabilistic model to characterize the deviation distribution of the driver when executing various types of operational commands further includes: acquiring real-time fatigue state data of the driver, which is evaluated based on facial features, eye movement features, and head posture data collected by the onboard driver monitoring system (DMS); and dynamically adjusting the deviation distribution parameters of the probabilistic model based on the real-time fatigue state data so that the model can reflect the operational deviation characteristics of the driver in the current fatigue state. For example, when the system detects that the driver is fatigued, the standard deviation of the deviation distribution is amplified by 1.5 times to more conservatively estimate the range of operational deviation. By introducing a dynamic deviation correction probabilistic model based on driver fatigue state, the setting of the fault tolerance interval more accurately reflects the driver's current actual operational ability, avoiding the failure to escape difficulties caused by using the fault tolerance interval under normal conditions when the driver is fatigued.

[0099] Step 6.2: Based on the probability model and the preset minimum requirements for getting out of trouble, determine the operational error tolerance range of key parameters in each operation command; the operational error tolerance range is the allowable range of parameters where the driver's operational deviation will not lead to failure to get out of trouble; Here, based on the deviation probability model and the minimum requirements for getting out of trouble, the operational tolerance range for each key parameter of the operation command is determined. First, the minimum requirements for getting out of trouble are determined, that is, the allowable value range of each operation parameter under the condition of successful getting out of trouble. Then, the driver deviation probability model and the minimum requirements for getting out of trouble are superimposed and analyzed to determine which parameter ranges can simultaneously satisfy the conditions for successful getting out of trouble and the driver deviation coverage probability requirement (e.g., covering 95% of the deviation probability) under driver operation deviation. This parameter range is the operational tolerance range. This step combines the conditions for successful getting out of trouble with the uncertainty of driver operation, determining the operational tolerance range that both satisfies the requirements for getting out of trouble and covers driver deviation, providing clear constraint boundaries for subsequent parameter selection.

[0100] In a preferred embodiment, determining the operational tolerance range of key parameters in each operational command further includes: acquiring current environmental condition data, including road surface slippage, visibility, and temperature; and adjusting the parameter thresholds of the minimum traction requirement based on the environmental condition data. For example, under slippery road conditions, the upper limit threshold of the accelerator pedal opening is reduced by 20% to reduce the risk of tire slippage; under low temperature conditions, the lower limit threshold of the brake pedal travel is increased by 10% to compensate for the braking response delay caused by increased brake fluid viscosity. The adjusted minimum traction requirement is used to recalculate the operational tolerance range. By adaptively adjusting the minimum traction requirement under environmental conditions, the operational tolerance range can adapt to changes in vehicle dynamics under different environmental conditions, improving the environmental adaptability of the traction strategy.

[0101] Step 6.3: After outputting the guidance information, continuously monitor the driver's actual deviation from the guidance information; Here, the driver's actual execution of the guidance information is continuously monitored during the traction guidance process. Onboard sensors collect real-time operational parameters such as steering wheel angle, accelerator pedal opening, brake pedal travel, and gear position, comparing these with the target parameter values ​​in the guidance information to calculate the actual execution deviation. The monitoring results are fed back to the traction strategy generation module in real time to determine whether subsequent operational instructions need adjustment. This step enables real-time monitoring of the driver's operational execution, providing trigger conditions and actual deviation data for replanning when deviations exceed limits.

[0102] In a preferred embodiment, continuous monitoring of the driver's actual deviation from the guidance information further includes: establishing a time-series prediction model of the driver's operational deviation; based on current and historical deviation data, predicting the deviation trend of the driver in the next few steps; when the predicted deviation trend indicates that the driver's operational deviation will continue to increase, the system triggers a replanning mechanism in advance, rather than waiting for the actual deviation to exceed the tolerance range before responding. The time-series prediction model can employ time series prediction methods such as the Autoregressive Integral Moving Average (ARIMA) model or Long Short-Term Memory (LSTM) network. By predicting deviation trends and replanning in advance, the passive response-based deviation handling is upgraded to proactive preventative deviation management, reducing escape failures caused by deviation exceeding limits and improving the robustness of escape guidance.

[0103] Step 6.4: When the actual execution deviation exceeds the operation tolerance range, take the current vehicle state as the initial state, select the key parameter with the widest operation tolerance range, recalculate each operation instruction in the escape operation guidance sequence, and update the guidance information; Here, replanning is triggered when the driver's actual deviation exceeds the operational tolerance range. Using the current vehicle state (including vehicle attitude, tire slip ratios, vehicle position, etc.) as the new initial state, the escape strategy generation module is invoked to recalculate the escape operation guidance sequence. During recalculation, the system prioritizes parameter values ​​with the widest operational tolerance range, making the newly generated sequence more tolerant of driver deviations. The updated guidance information is output to the driver through a multimodal interaction interface. This step achieves intelligent replanning after deviation exceeds the limit, handling driver deviations through follow-up replanning rather than forced error correction, ensuring the continuity of the escape strategy while avoiding negative impacts on driver confidence.

[0104] In a preferred embodiment, recalculating each operation instruction in the escape operation guidance sequence and updating the guidance information further includes: recording the operation parameters, deviation value, and reprogrammed sequence parameters of the current deviation exceeding the limit event; uploading the recorded data to a cloud server to update the parameters of the driver operation deviation probability model, enabling online learning and continuous optimization of the probability model. The cloud server aggregates deviation exceeding the limit event data from multiple vehicles, updates the global probability model through federated learning or incremental learning methods, and distributes the updated model parameters to each vehicle, forming a closed loop of model optimization driven by collective intelligence. Through cloud data aggregation and online learning, continuous optimization of the driver operation deviation probability model is achieved, enabling the model to continuously improve its prediction accuracy as data accumulates, and the escape fault tolerance strategy is also continuously improved with model optimization.

[0105] In application, the following examples can be used: First, based on historical vehicle driving data, extract the deviations between the actual and target parameter values ​​when the driver executes various types of operation commands, including steering wheel angle deviation, accelerator pedal opening deviation, and brake pedal travel deviation. Statistical analysis of the deviation data is performed to estimate the probability distribution of the deviations and establish a driver operation deviation probability model. Second, determine the minimum requirements for escaping difficulty, i.e., the allowable range of values ​​for each operation parameter under the condition of successful escaping difficulty. Overlay the driver deviation probability model with the minimum requirements for escaping difficulty to determine the parameter range that meets the condition of successful escaping difficulty while covering 95% deviation probability, serving as the operational tolerance range for key parameters of each operation command. Then, during the escaping difficulty guidance process, the driver's operation parameters are collected in real time by onboard sensors and compared with the target parameter values ​​in the guidance information to continuously monitor the actual execution deviation. Furthermore, when the driver's actual execution deviation exceeds the operational tolerance range, the current vehicle state is used as the new initial state. The escaping difficulty strategy generation module is invoked, prioritizing the parameter values ​​with the widest operational tolerance range, recalculating the escaping difficulty operation guidance sequence, and updating the guidance information through a multimodal interaction interface. Finally, the driver continues to perform the extrication operation based on the updated guidance information, continuously monitors deviations and repeats the above process until the vehicle is successfully extricated or the driver manually exits the process.

[0106] In practical applications, the following steps can be included as an example: Step 7.1: Acquire images of the vehicle's surroundings captured by the vehicle-mounted surround-view camera and perform semantic segmentation to identify passable areas, obstacle areas, and uncertain areas; Here, images of the vehicle's surroundings are acquired using an onboard surround-view camera. A deep learning semantic segmentation network is then used to perform pixel-level classification of these images, identifying passable areas (such as flat ground and hard surfaces), obstacle areas (such as rocks, trees, and walls), and uncertain areas (such as shadows, water surfaces, and vegetation-covered areas—terrain that cannot be accurately determined visually). The semantic segmentation network can employ classic network architectures such as U-Net or DeepLab, based on an encoder-decoder structure, and is trained using a large number of labeled off-road terrain images. This step achieves preliminary classification of the terrain area surrounding the vehicle through visual perception, providing target areas for subsequent active terrain detection. This allows active detection to focus on uncertain areas that cannot be determined visually, improving detection efficiency.

[0107] In a preferred embodiment, semantic segmentation to identify passable areas, obstacle areas, and uncertain areas further includes: employing a multi-frame temporal image fusion method to align and fuse consecutive frames of vehicle surrounding images, utilizing temporal information to eliminate interference factors such as shadows and occlusions in single-frame images, thereby improving the accuracy of semantic segmentation. Specifically, the pose transformation between adjacent frames is calculated through vehicle motion estimation, historical frame images are projected onto the current frame's viewpoint for fusion, and semantic segmentation is performed on the fused image, ensuring consistency of segmentation results in the temporal dimension and reducing misclassification caused by lighting changes or temporary occlusions. Through multi-frame temporal fusion, misclassifications caused by interference such as shadows and occlusions in single-frame semantic segmentation are eliminated, improving the accuracy and robustness of region identification.

[0108] Step 7.2: For the uncertain area, control the vehicle tires to perform a trial rotation of a preset amplitude. Based on the change in tire slip ratio during the trial rotation, infer the ground physical properties of the uncertain area. Here, this step performs active terrain detection on the uncertain areas identified by semantic segmentation. One or more tires of the vehicle are controlled to make small, tentative rotations (e.g., 1 / 4 turn) within the uncertain area, with the rotation amplitude preset to be sufficiently small to avoid significantly affecting the vehicle's attitude. During these tentative rotations, changes in tire slip ratio are monitored in real time, and ground physical properties are inferred based on the slip ratio response characteristics. For example, a rapid increase in slip ratio to a high value indicates soft ground (e.g., sand, mud); a stable slip ratio remaining at a low value indicates firm ground (e.g., rock, hard soil); and periodic fluctuations in slip ratio may indicate the presence of irregular bumps in the ground. This step compensates for the inherent limitations of visual perception through active tactile detection, accurately acquiring ground physical properties that cannot be determined visually, providing crucial terrain physical information for the precise formulation of subsequent escape strategies.

[0109] In a preferred embodiment, inferring the ground physical properties of an uncertain area further includes: simultaneously acquiring tire vibration signals during the trial rotation, and using spectral analysis of the vibration signals to assist in determining the ground physical properties. Specifically, tire vibration signals are acquired using accelerometers mounted on the suspension or wheel hub, and a Fast Fourier Transform (FFT) is performed on the vibration signals to extract spectral features. Different ground materials produce different vibration spectral characteristics when rolled by tires: sand produces low-frequency broadband vibrations, rocks produce high-frequency narrow-band vibrations, and mud produces mid-frequency vibrations accompanied by damping attenuation. By fusing the vibration spectral features with the slip rate change features, and using Bayesian inference or a support vector machine classifier to comprehensively determine the ground physical properties, the accuracy of the inference is improved. Through multimodal fusion of vibration signals and slip rate signals, the accuracy of ground physical property inference is improved, avoiding misjudgments that may occur under specific conditions (such as extreme temperatures or sensor malfunctions) due to a single signal source.

[0110] Step 7.3: Combine the passable area, the obstacle area, and the uncertain area containing ground physical attributes into the surrounding terrain information in the vehicle's comprehensive perception information; Here, this step fuses the results of visual perception and active detection to form complete surrounding terrain information. Passable and obstacle areas retain the semantic segmentation recognition results, while uncertain areas are updated to defined areas including ground physical attributes after active detection. The fused surrounding terrain information includes both geometric features (region type, boundaries, elevation) and physical features (ground material, adhesion coefficient), providing a comprehensive terrain data foundation for subsequent escape strategy generation. This step achieves the organic integration of visual perception and tactile detection, upgrading the surrounding terrain information from a single geometric description to a geometric-physical dual-dimensional description, significantly improving the completeness and usability of the terrain information.

[0111] In a preferred embodiment, using all areas collectively as surrounding terrain information further includes: constructing a dynamic grid map of the vehicle's surrounding terrain. The dynamic grid map stores terrain information for each grid cell, including grid type (accessible, obstacle, uncertain), ground physical properties (material type, estimated adhesion coefficient), confidence level (reliability of the data source), and timestamp (data acquisition time). As the vehicle moves during the extrication process, the dynamic grid map is updated in real-time based on newly acquired terrain data. Grids with timestamps exceeding a preset validity period are automatically downgraded to uncertain areas, triggering re-detection. Confidence level is determined based on the data source: actively detected data has higher confidence than visually perceived data, and multi-sensor fusion data has higher confidence than single-sensor data. Through real-time updates and timeliness management of the dynamic grid map, it is ensured that the surrounding terrain information always reflects the latest terrain status, avoiding the failure of the extrication strategy due to outdated terrain information.

[0112] In practical applications, the following example can be used: First, the system acquires images of the vehicle's surroundings using an onboard surround-view camera. A deep learning semantic segmentation network based on the U-Net architecture is then used to perform pixel-level classification of the images, identifying passable areas (flat ground, hard surfaces), obstacle areas (rocks, trees), and uncertain areas (shaded areas, water surfaces, vegetation areas). Second, for uncertain areas, the system controls the vehicle's tires to make a small, exploratory rotation of 1 / 4 turn, monitoring the tire slip ratio in real time. If the slip ratio rapidly rises above 80%, it is inferred to be sand or mud; if the slip ratio remains stable below 20%, it is inferred to be rock or hard soil. Then, the semantic segmentation results are fused with the ground physical attributes inferred from active detection. The passable areas, obstacle areas, and uncertain areas containing ground physical attributes are combined as the surrounding terrain information in the vehicle's comprehensive perception information. Finally, the fused surrounding terrain information is input into the escape strategy generation module for subsequent passability analysis and escape operation guidance sequence generation.

[0113] In some embodiments, the step of generating a time-sequential escape operation guidance sequence in step S104 of the foregoing embodiments further includes: Step 8.1: Determine whether the extrication operation involves the reciprocating swaying of the vehicle, based on the type of entrapment. Here, this step determines whether a reciprocating swaying strategy is needed for extrication based on the type of entrapment. For vehicles stuck in sand or mud, reciprocating swaying is a common extrication method, using inertia to overcome ground resistance through repeated forward and backward movement. For vehicles stuck in rocks or with the chassis bottomed out, reciprocating swaying is not only ineffective but may also damage the vehicle. Based on a preset entrapment type-swaying strategy mapping table, the system automatically determines whether reciprocating swaying is necessary. This step automatically determines the applicability of the reciprocating swaying strategy, avoiding wasted effort or vehicle damage caused by using swaying operations in inapplicable scenarios.

[0114] In a preferred embodiment, determining whether the extrication operation involves reciprocating oscillation of the vehicle further includes: acquiring the status parameters of the vehicle's transmission system, including transmission oil temperature, clutch temperature (for manual / dual-clutch transmissions), or torque converter oil temperature (for automatic transmissions); if any of the status parameters exceeds a preset safety threshold, the system automatically disables the reciprocating oscillation strategy, even if the type of entrapment is suitable for reciprocating oscillation, and instead adopts a single, slow extrication strategy to avoid overheating and damage to the transmission system. Simultaneously, the system issues a warning message to the driver that the transmission system temperature is too high and suggests waiting for it to cool down. Through transmission system status monitoring and temperature protection, mechanical damage caused by forcibly performing reciprocating oscillation when the transmission system is overheated is avoided, extending the service life of critical vehicle components.

[0115] Step 8.2: If reciprocating oscillation is involved, use the real-time tire dynamics model to calculate the degree of change in ground physical properties caused by repeated tire rolling under different oscillation phases; Here, this step models the time-varying effects of ground physical properties during the reciprocating oscillation process. When a vehicle oscillates back and forth, the tires repeatedly roll over the same area of ​​the ground, which may cause changes in the ground's physical properties. For example, sand may become compacted and its adhesion coefficient may increase after repeated rolling; mud may become loosened and its adhesion coefficient may decrease. The system uses a real-time tire dynamics model to simulate the tire's rolling action on the ground under different oscillation phases and calculates the degree of change in ground physical properties (such as ground density and adhesion coefficient). This step breaks the assumption of static invariance of ground properties and introduces the dynamic influence of the extrication process itself on the ground state, enabling the extrication strategy to adapt to real-time changes in ground properties and improving the timeliness and accuracy of the extrication strategy.

[0116] In a preferred embodiment, calculating the degree of change in ground physical properties due to repeated tire rolling under different oscillation phases further includes: obtaining physical property parameters of the ground material, including soil cohesion, internal friction angle, and moisture content; establishing a ground compaction-loosening mechanical model based on Terzaghi's foundation bearing capacity theory or the Bekker ground mechanics model, and calculating the change curves of ground physical properties under different rolling cycles and rolling pressures. Specifically, according to the Bekker ground mechanics model, ground subsidence... With ground pressure The relationship is:

[0117] in, It is the soil cohesive modulus. This refers to the internal friction modulus of the soil. This refers to the tire's contact patch width. The soil deformation index, This represents the amount of subsidence. Using this model, the system can calculate the change in ground subsidence after each compaction, and further deduce the changes in ground density and adhesion coefficient. By introducing a ground mechanics theory model, the calculation of changes in ground physical properties is based on classical mechanics theory, improving the scientific rigor and accuracy of the calculation results.

[0118] Step 8.3: Based on the degree of change, dynamically update the tire adhesion parameters for subsequent swing cycles, and adjust the accelerator pedal opening and steering wheel angle in the traction operation guidance sequence accordingly; Here, this step dynamically updates the strategy parameters for subsequent oscillation cycles based on the degree of change in ground physical properties. When the calculated adhesion coefficient increases after the ground has been compacted, the accelerator pedal opening is appropriately increased in subsequent oscillation cycles to utilize the increased adhesion and output greater driving force; when the adhesion coefficient decreases, the accelerator pedal opening is appropriately decreased to avoid excessive tire slippage. Simultaneously, the steering wheel angle is adjusted according to changes in ground physical properties to adapt to changes in the passable direction of the ground. This step achieves real-time adaptive adjustment of the traction strategy to changes in ground conditions, allowing each oscillation cycle to be optimized based on the latest ground conditions, avoiding the unsuitability of fixed-parameter strategies in dynamic environments.

[0119] In a preferred embodiment, dynamically updating the tire adhesion parameters for subsequent oscillation cycles further includes: after each oscillation cycle, re-collecting tire slip ratio data and ground image data using onboard sensors, comparing the actual measured values ​​with the model's predicted values; based on the comparison differences, using Kalman filtering or particle filtering methods to perform online parameter correction on the ground physical property change model, so that the model's predicted values ​​gradually approach the actual values. The corrected model is used for parameter updates in the next oscillation cycle, forming a closed-loop optimization mechanism of prediction-execution-measurement-correction. Through online measurement feedback and model parameter correction, continuous self-calibration of the ground physical property change model is achieved, improving the model's prediction accuracy and adaptability under complex and varied terrain conditions.

[0120] As an optional example, the process includes the following steps: First, determine whether a reciprocating oscillation strategy is needed for extrication based on the type of entrapment. For vehicles stuck in sand, reciprocating oscillation is determined to be necessary; for vehicles stuck in rocks, reciprocating oscillation is not required, and other extrication strategies are adopted instead. Second, if reciprocating oscillation is involved, a real-time tire dynamics model is used to simulate the tire's compaction of the ground under different oscillation phases, calculating the degree of change in ground physical properties (such as ground density and adhesion coefficient). For example, for sand, the adhesion coefficient increases by 15% after the first compaction, by a cumulative 25% after the second compaction, and tends to stabilize after the third compaction. Then, based on the degree of change in ground physical properties, the strategy parameters for subsequent oscillation cycles are dynamically updated. For example, when the adhesion coefficient increases, the accelerator pedal opening is increased from 30% to 40% in the second oscillation cycle to utilize the increased adhesion to output greater driving force. Finally, the accelerator pedal opening and steering wheel angle in the extrication operation guidance sequence are adjusted according to the updated parameters, and the guidance information is updated and output to the driver.

[0121] In some embodiments, the step of generating a time-sequential escape operation guidance sequence in step S104 of the foregoing embodiments further includes: Step 9.1: Based on the type of entrapment, construct a parameterized strategy space with adjustable parameters such as steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing. Here, this step represents the escape strategy as a parameterized space with adjustable parameters. The parameterized strategy space uses steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing as adjustable parameters, each with its own value range and discretization granularity. For example, the steering wheel angle ranges from -540° to +540°, with a discretization granularity of 5°; the accelerator pedal opening timing description includes a sequence of opening values ​​and a sequence of time intervals, with the opening value ranging from 0% to 100%. The size of the parameterized strategy space depends on the value range and discretization granularity of each parameter, and is typically very large, containing a large number of possible escape operation combinations. This step expands the discrete escape strategy library into a continuous parameterized strategy space, overcoming the limitation of incomplete coverage by the pre-defined strategy library and providing a broader and more flexible search space for online optimization algorithms.

[0122] In a preferred embodiment, constructing the parameterized strategy space further includes: pruning the parameterized strategy space based on prior knowledge of the entrapment type to narrow the search space range. For example, for the front wheel entrapment type, the steering wheel angle is limited to ±30° (because large steering may cause vehicle sideslip); for the sand entrapment type, the accelerator pedal opening is limited to below 60% (because excessive opening will cause the tires to dig into the crater). The constraint pruning rules are based on an entrapment type-parameter constraint mapping table, which is derived from expert experience and historical escape data. The pruned parameterized strategy space is significantly reduced in size, enabling the online optimization algorithm to converge to the optimal solution in a shorter time. Through constraint pruning based on prior knowledge, the search range is significantly reduced while maintaining the integrity of the search space, significantly improving the search efficiency and convergence speed of the online optimization algorithm.

[0123] Step 9.2: In the parameterized strategy space, with the success rate of getting out of trouble and the estimated time for getting out of trouble as the optimization objectives, the online optimization algorithm is used to search for the optimal parameter combination. During the search process, the online optimization algorithm uses a real-time tire dynamics model to simulate and deduce each candidate parameter combination, and iteratively updates the search direction according to the deduction results until convergence. Here, this step employs an online optimization algorithm to search for the optimal parameter combination in the parameterized policy space. The optimization objective is a multi-objective optimization of the success probability of escaping entrapment and the estimated escaping time. During the search process, the online optimization algorithm uses a real-time tire dynamics model to simulate and extrapolate each candidate parameter combination, predicting the escaping process under that parameter combination, evaluating the success probability and estimated time, and iteratively updating the search direction based on the simulation results until convergence to the optimal parameter combination. Commonly used online optimization algorithms include Bayesian optimization, genetic algorithms, particle swarm optimization, and simulated annealing.

[0124] The weighted objective function for multi-objective optimization is as follows:

[0125] in, For parameter combination vectors, The probability of successfully escaping (0-1). The estimated time required to escape the predicament To the maximum allowable time, and Let be the weighting coefficient, satisfying The optimization objective is to minimize .

[0126] This step upgrades the generation of escape strategies from discrete selection to continuous space search. By combining online optimization algorithms with model simulation, it achieves the global optimal solution for escape strategies, avoiding the limitations of incomplete coverage and local optima in the preset strategy library.

[0127] In a preferred embodiment, the online optimization algorithm for searching the optimal parameter combination further includes: introducing a diversity preservation mechanism during the search process. This mechanism includes: randomly generating exploratory candidate combinations with a certain probability (e.g., 10%) during each iteration, rather than generating them entirely based on the current optimal direction; and maintaining a taboo list of already searched regions to prevent the search algorithm from repeatedly searching already evaluated parameter combination regions. When the search gets stuck in a local optimum for more than a preset number of iterations, the system automatically triggers a restart mechanism, randomly selecting a new initial point from the parameterized policy space to restart the search. The diversity preservation and restart mechanisms effectively prevent the online optimization algorithm from prematurely converging to a local optimum. Through these mechanisms, the global search capability of the online optimization algorithm is enhanced, avoiding local optimum traps caused by improper initial point selection or fixed search paths, and increasing the probability of finding the global optimum.

[0128] Step 9.3: Generate a time-seriesed escape operation guidance sequence based on the optimal parameter combination.

[0129] Here, this step transforms the optimal parameter combination obtained from the online optimization algorithm into a specific time-sequential sequence of traction-avoidance operation guidance. The steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing from the optimal parameter combination are arranged chronologically to form a complete sequence containing each operation command, execution time, and duration. This sequence is the final time-sequential traction-avoidance operation guidance sequence output to the traction-avoidance guidance module. This step realizes the transformation from optimized parameters to executable commands, enabling the driver to understand and execute the abstract parameter combination obtained from the online optimization algorithm.

[0130] In a preferred embodiment, generating a time-seriesed escape operation guidance sequence based on the optimal parameter combination further includes: robustness verification of the optimal parameter combination; applying a ±5% random perturbation to each parameter based on the optimal parameter combination to generate a set of perturbed parameter combinations; simulating each perturbed parameter combination using a real-time tire dynamics model to evaluate the success probability of escape after perturbation; if the success probability of escape for all perturbed parameter combinations is higher than a preset threshold (e.g., 90%), the robustness of the optimal parameter combination is confirmed to meet the requirements, and this parameter combination is used to generate the escape operation guidance sequence; otherwise, the system adds a robustness penalty term to the optimization objective and re-searches for a parameter combination with better robustness. The robustness penalty term is defined as the variance of the success probability of escape for the perturbed parameter combination. Through robustness verification and penalty term optimization, it is ensured that the generated optimal parameter combination not only performs optimally under the current simulation conditions but also maintains a high success rate of escape even with small parameter fluctuations, improving the reliability of the escape strategy in practical applications.

[0131] In application, this step can be implemented through the following example: First, based on the type of entrapment, a parameterized strategy space is constructed with adjustable parameters such as steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing. Each parameter is encoded according to a preset value range and discretization granularity to form a search space. Second, with the success rate of escaping the entrapment and the estimated time to escape as optimization objectives, a Bayesian optimization algorithm is used to search for the optimal parameter combination in the parameterized strategy space. During the search process, the Bayesian optimization algorithm uses a Gaussian process to construct a surrogate model of the objective function, selects the next candidate parameter combination by collecting a function (such as the desired improvement in EI), and uses a real-time tire dynamics model to simulate and extrapolate the candidate parameter combination, evaluating the success rate of escaping the entrapment and the estimated time to escape. The surrogate model is updated based on the extrapolation results, and the search direction is iteratively updated until convergence to the optimal parameter combination. Then, the robustness of the optimal parameter combination obtained by the search is verified. Based on the optimal parameter combination, a set of perturbed parameter combinations is generated by applying ±5% random perturbation to each parameter, and the real-time tire dynamics model is used to simulate and extrapolate each perturbed parameter combination. If the success rate of escape from the obstacle is higher than 90% for all combinations of perturbation parameters, then the robustness requirement is confirmed to be met. Finally, based on the optimal parameter combination, a complete time-series escape operation guidance sequence containing each operation command, execution time, and duration is generated and output to the multimodal interaction interface.

[0132] In some embodiments, such as Figure 2 As shown, an embodiment of the present invention provides a vehicle traction guidance device, comprising: The judgment module, in response to the request for extrication guidance, determines the type of vehicle entrapment based on the vehicle's comprehensive perception information; The generation module, based on the type of entrapment, uses a real-time tire dynamics model to perform a passability analysis on the comprehensive perception information of the vehicle, and generates a time-sequential guide sequence for getting out of trouble. The guidance module converts the escape operation guidance sequence into driver-oriented guidance information and outputs it.

[0133] It is understood that the specific operation methods of each functional module in this embodiment can be referred to the detailed description of the corresponding steps in the above method embodiment, and will not be repeated here.

[0134] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and apparatus described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0135] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0136] In the description of this invention, it should be noted that the terms center, up, down, left, right, vertical, horizontal, inner, and outer, indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms first, second, and third are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0137] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the scope of the technology disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention.

Claims

1. A method for guiding vehicles out of difficult situations, characterized in that, include: In response to a request for guidance on getting out of trouble, the vehicle's type of entrapment is determined based on comprehensive vehicle perception information; Based on the type of entrapment, a real-time tire dynamics model is used to perform a passability analysis on the comprehensive perception information of the vehicle, and a time-sequential entrapment operation guidance sequence is generated. The sequence of instructions for getting out of trouble is converted into driver-oriented guidance information and output.

2. The method according to claim 1, characterized in that, The comprehensive vehicle perception information includes surrounding terrain information; The steps for generating a time-series guide sequence for escape operations include: Based on the surrounding terrain information, a local digital elevation model of the tire-ground contact interface is constructed. The terrain information represented by the local digital elevation model is input into the real-time tire dynamics model to calculate the adhesion vector field of each tire under the current terrain. Based on the type of entrapment, the real-time tire dynamics model is used to perform a passability analysis on the surrounding terrain information to determine the vehicle's passable direction and entrapment strategy. With the goal of maximizing the effective traction force of the vehicle in the passable direction, the optimal tire force distribution scheme is solved based on the adhesion vector field. The optimal tire force distribution scheme is mapped as a time-sequential command sequence of steering wheel angle, accelerator pedal opening, and brake pedal action.

3. The method according to claim 2, characterized in that, When the traction control strategy includes vehicle body swaying, the step of generating a time-sequential traction control guidance sequence further includes: Determine the expected swing amplitude based on the type of entrapment; Based on the adhesion vector field, the transient ground imprint changes of each tire under the expected swing amplitude are calculated; Based on the transient grounding imprint changes, the peak adhesion window period for each tire is determined; The peak command of accelerator pedal opening in the time-sequential command sequence is aligned with the peak adhesion window period in terms of timing.

4. The method according to claim 1, characterized in that, The comprehensive vehicle perception information includes surrounding terrain information; The steps for generating a time-series guide sequence for escape operations also include: Based on the surrounding terrain information, the longitudinal slope, lateral slope, and flatness of the pit where the vehicle is currently located are identified. When at least one of the longitudinal slope, transverse slope, or pit bottom flatness exceeds a preset threshold, the chassis scraping risk area generated during the escape operation is predicted. Based on the chassis scraping risk area, chassis protection operation instructions are inserted into the time-sequential escape operation guidance sequence; the chassis protection operation instructions include reducing the accelerator pedal opening to reduce the vehicle pitch angle, or adjusting the steering wheel angle to avoid the scraping risk area.

5. The method according to claim 1, characterized in that, The comprehensive vehicle perception information includes tire slip rate information and surrounding terrain information; the steps for determining the type of vehicle entrapment include: Based on the slip ratio information of each tire and the surrounding terrain information, it is determined whether the vehicle has the ability to extricate itself from a difficult situation. If available, the driver's skill in getting out of trouble, reaction speed, and operational precision in the vehicle's historical driving data will be used to determine the current driver's ability to get out of trouble. Based on the total number of operation instructions in the escape operation guidance sequence, the timing coordination accuracy between each operation instruction, and the types of operation parameters that need to be adjusted in coordination, the operation complexity required for autonomous escape is determined. The operational complexity is matched with the assessment result of the ability to get out of trouble. If the operational complexity exceeds the driver's ability to get out of trouble, then the type of being trapped is determined to be a type of being trapped that requires automatic intervention.

6. The method according to claim 1, characterized in that, The method further includes: Based on historical vehicle driving data, a probabilistic model is established to characterize the deviation distribution of drivers when executing various types of operation commands; Based on the probability model and the preset minimum requirements for getting out of trouble, the operational error tolerance range of key parameters in each operation command is determined; the operational error tolerance range is the allowable range of parameters in which the driver's operational deviation will not lead to failure to get out of trouble. After outputting the guidance information, the actual deviation of the driver from the guidance information is continuously monitored; When the actual execution deviation exceeds the operation tolerance range, the current vehicle state is taken as the initial state, the key parameter with the widest operation tolerance range is selected to recalculate each operation instruction in the escape operation guidance sequence and update the guidance information.

7. The method according to claim 1, characterized in that, The method further includes: The system acquires images of the vehicle's surroundings from the vehicle-mounted surround-view camera and performs semantic segmentation to identify passable areas, obstacle areas, and uncertain areas. For the uncertain region, the vehicle tires are controlled to perform a trial rotation of a preset amplitude. Based on the change in tire slip ratio during the trial rotation, the ground physical properties of the uncertain region are inferred. Based on the passable area, the obstacle area, and the uncertain area containing the ground physical properties, the surrounding terrain information in the vehicle's comprehensive perception information is determined.

8. The method according to claim 1, characterized in that, The steps for generating a time-series guide sequence for escape operations also include: Based on the type of entrapment, determine whether the extrication operation involves the reciprocating swaying of the vehicle; If reciprocating oscillation is involved, the degree of change in ground physical properties caused by repeated tire rolling under different oscillation phases is calculated based on the real-time tire dynamics model. Based on the degree of change, the tire adhesion parameters of the subsequent swing cycle are dynamically updated, and the accelerator pedal opening and steering wheel angle in the traction operation guidance sequence are adjusted accordingly.

9. The method according to claim 1, characterized in that, The steps for generating a time-series guide sequence for escape operations also include: Based on the type of entrapment, a parameterized strategy space is constructed with adjustable parameters such as steering wheel angle, accelerator pedal opening timing, brake pedal action timing, and gear shifting timing. In the parameterized strategy space, with the success rate of escape and the estimated time for escape as optimization objectives, an online optimization algorithm is used to search for the optimal parameter combination; Based on the optimal parameter combination, the time-series escape operation guidance sequence is generated.

10. A vehicle traction guidance device, characterized in that, include: The judgment module, in response to the request for extrication guidance, determines the type of vehicle entrapment based on the vehicle's comprehensive perception information; The generation module, based on the type of entrapment, uses a real-time tire dynamics model to perform a passability analysis on the comprehensive perception information of the vehicle, and generates a time-sequential guide sequence for getting out of trouble. The guidance module converts the escape operation guidance sequence into driver-oriented guidance information and outputs it.

Images