Method and system for controlling a transfer case of an autonomous vehicle

Abstract

The application discloses a kind of splitter control method and control system of unmanned vehicle, belong to unmanned vehicle control technical field.It includes: obtaining wheel speed sensor, inertial measurement unit and the multi-source sensor data of radar, respectively extract features and calculate the original probability corresponding to different road conditions;The original probability is corrected based on the real-time signal-to-noise ratio of each sensor, and the road condition recognition result is obtained by using D-S evidence theory fusion;When the fusion confidence is lower than threshold value, switch to adaptive weighted fusion based on historical accuracy and current data quality;According to the road condition recognition result and the preset road condition-mode mapping relationship, generate shift decision;After meeting the shift check condition, execute shift, and use vehicle longitudinal dynamics model to carry out closed-loop feedback check.The application has perfect fault tolerance capability, realizes the safe, reliable, intelligent automatic switching of splitter mode.

Description

Technical Field

[0001] This invention relates to the field of autonomous vehicle control technology, specifically to a transfer case control method and control system for an autonomous vehicle. Background Technology

[0002] Four-wheel drive systems, with their advantages of simple structure, high transmission efficiency, and strong off-road reliability, are widely used in hardcore off-road vehicles and some special vehicles. A typical full-time four-wheel drive system uses a transfer case to switch between modes such as H (high-speed two-wheel / four-wheel drive), HL (high-speed four-wheel drive lock), N (neutral), L (low-speed four-wheel drive), and LL (low-speed four-wheel drive lock). However, traditional part-time four-wheel drive systems require the driver to judge road conditions based on experience and manually operate the transfer case switch to switch modes, which demands a high level of professional knowledge and operational proficiency. For drivers unfamiliar with the system, incorrect gear selection may lead to wear and tear on the transmission system, difficulty in steering, or even the risk of rollover; and in complex road conditions requiring four-wheel drive intervention, failure to switch in time may result in the vehicle getting stuck or reduced off-road capability.

[0003] To address the aforementioned issues, existing technologies have proposed automatic or assisted gear shifting solutions. For example, there is an existing solution called an all-terrain system-based gear shifting system and vehicles equipped with it. This solution acquires gear requests, all-terrain system (ATS) modes, vehicle signals, and road condition signals through an input module. The ATS module then comprehensively judges these signals and sends a shift mode signal to the transfer case controller to execute the shift, or sends a prompt signal to the instrument panel to guide the driver. This technical solution can automatically select or prompt gears when transitioning between specific road conditions (such as bumpy roads and paved roads), reducing the driver's operational difficulty.

[0004] However, the above-mentioned solutions still have shortcomings in practical applications. Their road condition recognition mechanisms are relatively simple, for example, only roughly classifying road conditions into "bumpy road conditions," "paved road conditions," and "stuck road conditions," mainly relying on discrete, macroscopic qualitative judgments. This simplified road condition recognition method has particularly prominent limitations when facing complex, dynamic, and unstructured driving scenarios such as autonomous driving or remote control. First, autonomous vehicles may face multiple tasks such as delivery, exploration, and rescue, needing to cope with continuous changes from paved urban highways to low-traction or unpaved surfaces such as mining areas, forest roads, snowfields, and sandy areas. A single qualitative judgment cannot finely characterize key features such as road surface adhesion coefficient and undulation, resulting in coarse transfer case switching decisions that may not provide optimal passability or fuel economy. Second, existing solutions do not disclose how to effectively fuse data from multiple heterogeneous sensors such as wheel speed sensors, inertial measurement units, and radar / cameras to overcome the problem of decreased data reliability from a single sensor in complex environments (such as wheel speed sensors covered in mud or IMU signal noise caused by vibration). This results in insufficient robustness and confidence in road condition recognition results under harsh environments, potentially leading to misjudgments or inability to make decisions. Furthermore, its gear shifting decisions lack reverse verification and fault diagnosis mechanisms based on dynamic models. When sensors malfunction or the road condition recognition model fails, the system lacks fault tolerance and degradation processing strategies, which is unacceptable for autonomous driving applications with extremely high safety requirements.

[0005] Therefore, how to achieve refined and robust identification of vehicle driving conditions in autonomous driving or remote control mode, and how to safely and reliably control the transfer case to switch modes based on the identification results, while also having fault tolerance capability, is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] The present invention aims to solve at least one technical problem existing in the above-mentioned background art, and provides a transfer case control method and control system for unmanned vehicles.

[0007] The technical solution of the present invention is: a transfer case control method for an unmanned vehicle, applied to an unmanned vehicle with a transfer case, the method comprising: In response to the autonomous driving enable signal being valid and the transfer case switch position being consistent with the current gear, the autonomous driving mode is entered. Acquire multi-source sensor data of the vehicle and calculate the original probability of each sensor data corresponding to different road condition types; use DS evidence theory and fuse the original probabilities based on the real-time signal-to-noise ratio of each sensor to identify the current driving road condition and obtain the road condition identification result. Based on the road condition recognition results and the preset road condition-mode mapping relationship, a transfer case shifting decision is generated; Under the condition that the preset shift verification conditions are met, the transfer case shift decision is executed, and the transfer case is controlled to switch to the target gear.

[0008] According to the present invention, a transfer case control method for an unmanned vehicle includes multi-source sensor data such as wheel speed sensor data, inertial measurement unit data, and radar data; the method fuses the multi-source sensor data to identify the current road conditions, specifically including: Feature extraction was performed on wheel speed sensor data, inertial measurement unit data, and radar data respectively, and the original probability of each sensor data corresponding to different road condition types was calculated. By employing the DS evidence theory, the original probabilities of data from each sensor are fused to obtain the fused road condition recognition result.

[0009] The transfer case control method for an unmanned vehicle provided by the present invention further includes: Obtain the real-time signal-to-noise ratio of each sensor; Based on the real-time signal-to-noise ratio of each sensor, the original probability of each sensor data is corrected to obtain the basic probability allocation value of each sensor for different road condition subsets. The basic probability assignment value is positively correlated with the signal-to-noise ratio of the sensor.

[0010] According to the transfer case control method of an unmanned vehicle provided by the present invention, the method for fusing the basic probability assignment values ​​using DS evidence theory includes: The dynamic conflict factor is calculated based on the basic probability allocation values ​​of each sensor and the dynamic weights calculated according to the real-time signal-to-noise ratio of each sensor. Using the dynamic conflict factor, the basic probability assignment values ​​of multiple sensors are calculated by DS synthesis to obtain the fusion confidence of each road condition proposition; If the fusion confidence level is higher than the preset first confidence level threshold, the road condition recognition result is directly output.

[0011] According to a transfer case control method for an autonomous vehicle provided by the present invention, if the fusion confidence level is not higher than the first confidence level threshold, the method further includes: Entering the adaptive weighted fusion stage: Based on the historical recognition accuracy of each sensor and the current data quality, calculate the dynamic fusion weight of each sensor; The original probabilities of each sensor are weighted and averaged based on the dynamic fusion weights to obtain the final road condition recognition result.

[0012] According to the transfer case control method for an unmanned vehicle provided by the present invention, the road condition-mode mapping relationship includes: When the road condition identification result is a paved road surface and the confidence level is higher than the second confidence threshold, the transfer case shifts to H mode. When the road condition identification result is a low-adhesion road surface and the confidence level is higher than the third confidence level threshold, the transfer case shift decision is to switch to HLock mode; When the road condition recognition result is an off-road surface and the confidence level is higher than the fourth confidence level threshold, the transfer case shift decision is to switch to LLock mode; When the confidence level of the road condition recognition result is lower than any of the thresholds, the transfer case shifts to maintain the current gear.

[0013] According to the present invention, a transfer case control method for an unmanned vehicle includes shift verification conditions including: the current transmission gear, the current vehicle speed, the matching relationship between the braking signal and the target gear; the matching relationship is determined according to a preset shift condition table, which specifies the allowed range of current transmission gears, the upper limit of vehicle speed, and whether the braking signal needs to be valid for different target gears.

[0014] According to the present invention, a transfer case control method for an unmanned vehicle is provided, the method further includes: after the transfer case performs a gear shift, using a vehicle longitudinal dynamics model to back-calculate the actual adhesion coefficient of the current road surface, and comparing the actual adhesion coefficient with the expected adhesion coefficient corresponding to the road condition recognition result; if the deviation exceeds a preset threshold, the sensor is determined to be abnormal, and the basic probability allocation value or fusion weight of the sensor determined to be abnormal in the subsequent road condition recognition fusion process is reduced.

[0015] According to a transfer case control method for an autonomous vehicle provided by the present invention, the method further includes exiting the autonomous driving mode in response to any of the following conditions: Received a signal indicating that autonomous driving is disabled; Driver operation of transfer case switch detected; A system fault flag bit has been detected being set. When exiting the autonomous driving mode, if the current gear is inconsistent with the gear indicated by the switch signal, the transfer case will be controlled to switch to the gear indicated by the switch signal when the shift verification condition is met.

[0016] This invention also relates to a transfer case control system for an unmanned vehicle, comprising: The mode management module is used to control the vehicle to enter the autonomous driving mode in response to the autonomous driving enable signal being valid and the transfer case switch position being consistent with the current gear. The road condition recognition module is used to acquire multi-source sensor data of the vehicle, calculate the original probability of each sensor data corresponding to different road condition types, and fuse the original probabilities based on the DS evidence theory and the real-time signal-to-noise ratio of each sensor to identify the current driving road condition and obtain the road condition recognition result. The shift decision module is used to generate a transfer case shift decision based on the road condition recognition result and the preset road condition-mode mapping relationship; The shift execution module is used to execute the transfer case shift decision and control the transfer case to switch to the target gear when the preset shift verification conditions are met. The exit management module is used to control the vehicle to exit the autonomous driving mode in response to exit conditions.

[0017] The advantages of this invention are: 1. High accuracy and robustness in road condition recognition. The control method of this invention integrates three sensors with vastly different physical characteristics: wheel speed sensors, inertial measurement units (IMUs), and radar. It extracts multi-dimensional features such as slip ratio, wheel speed fluctuation entropy, vertical vibration intensity, road surface slope, and smoothness. By synthesizing evidence from different sources using DS evidence theory, it effectively handles information uncertainty and conflict. Especially when a single sensor is interfered with (e.g., the wheel speed sensor is covered in mud, or the radar lens is dirty), information from other sensors can still support accurate judgment, significantly improving the robustness of road condition identification in harsh off-road environments.

[0018] 2. Adaptive switching of fusion strategy The control method of this invention proposes a DSAW-Fusion architecture. In normal mode, it employs high-precision DS evidence theory fusion; when the evidence is highly conflicting or the fusion confidence is insufficient, it automatically switches to adaptive weighted average fusion. The latter comprehensively considers the historical recognition accuracy of the sensors and the current data quality (signal-to-noise ratio). Even in extreme cases where some sensors fail or the signal quality deteriorates significantly, the system can still output stable and continuous road condition recognition results, preventing the entire function from failing due to a single sensor failure. This greatly improves the availability and fault tolerance of the autonomous driving system.

[0019] 3. Safe and reliable gear shifting decision-making, with comprehensive mechanical protection. The control method of this invention establishes a differentiated confidence threshold decision mechanism (0.8 for paved roads, 0.7 for low-adjacent roads, and 0.75 for off-road roads). Higher confidence levels are required for high-risk gear shifts (such as entering LLock low-speed lock-up mode) to avoid dangerous operations due to misjudgment. Simultaneously, before gear shifting, a preset shift condition table must be used for verification, strictly limiting vehicle speed, transmission gear, and braking signals to ensure that the gear shifting action is completed within the mechanically permissible safety boundaries, effectively preventing hardware damage such as gear grinding and wear.

[0020] 4. Closed-loop feedback verification, with self-diagnosis and fault-tolerant repair capabilities. The control method of this invention introduces a vehicle longitudinal dynamics model, uses actual driving force and acceleration to infer the road surface adhesion coefficient, and performs a closed-loop comparison with the road condition identification results. When the deviation exceeds a threshold, the system can automatically identify sensor anomalies, reduce the fusion weight of faulty sensors, and achieve self-diagnosis and fault-tolerant repair. This mechanism enables the system to operate stably for a long time and adapt to practical engineering problems such as sensor aging and sudden environmental changes.

[0021] 5. The mode switching process is complete, balancing safety and user-friendliness. The control method of this invention incorporates a comprehensive mechanism for entering and exiting unmanned mode. Before entering, the system verifies the consistency between the transfer case switch position and the current gear to ensure a stable initial state. Upon exiting, it differentiates between three scenarios: disabled signals, driver intervention, and system malfunction, prioritizing driver input to achieve a smooth transfer of control. If the target gear does not meet the safety conditions for shifting, the system maintains the current gear until the conditions are met, avoiding the risks associated with forced shifting and fully embodying the safety-first design philosophy. Attached Figure Description

[0022] Figure 1 : A flowchart of the transfer case control method for the unmanned vehicle of the present invention. Detailed Implementation

[0023] Embodiments of the present invention are described in detail below, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0024] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this 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. Therefore, they should not be construed as limitations on this invention.

[0025] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0026] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0027] The present invention will now be clearly and completely described in conjunction with the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] This invention provides a transfer case control method and control system for autonomous vehicles, applicable to 4×4 autonomous vehicles equipped with a gear-type transfer case. The transfer case has the following gears: H (high-speed two-wheel drive / four-wheel drive), HLock (high-speed four-wheel drive lock), L (low-speed four-wheel drive), LLock (low-speed four-wheel drive lock), and N (neutral, generally used for towing). This invention aims to solve the problems of existing all-wheel drive systems in autonomous driving scenarios, such as coarse road condition recognition, poor robustness of multi-sensor fusion, and lack of fault tolerance mechanisms, achieving refined and highly reliable automatic transfer case shifting control.

[0029] Example 1: This example provides a basic flow of a transfer case control method for an unmanned vehicle, such as... Figure 1 As shown, it includes the following steps: Step 101: In response to a valid autonomous driving enable signal and the transfer case switch position matching the current gear, the system enters autonomous driving mode. Specifically, the system monitors the autonomous driving enable signal (usually issued by the autonomous driving domain controller) in real time. When the signal is detected to be valid (e.g., high level or rising edge trigger), the system simultaneously reads the current physical position signal of the transfer case switch. The system checks the transfer case switch status, and only allows entry into autonomous driving mode when the switch is fault-free and the gear indicated by its position signal matches the current actual gear of the transfer case. If the switch is faulty, entry is prohibited or a preset safety mode is used. This consistency check ensures that the transfer case is in a known and stable state when entering autonomous driving mode.

[0030] Step 102: Acquire multi-source sensor data of the vehicle and perform fusion processing on the multi-source sensor data to identify the current driving road conditions and obtain road condition identification results. The multi-source sensors include at least wheel speed sensors, inertial measurement units (IMUs), and radar / lidar.

[0031] Step 103: Based on the road condition recognition results and the preset road condition-mode mapping relationship, generate a transfer case shift decision. The mapping relationship defines the optimal transfer case operating mode (e.g., H, HLock, LLock) for different road conditions (e.g., paved roads, low-adhesion roads, off-road roads).

[0032] Step 104: Under the condition that the preset shift verification conditions are met, execute the transfer case shift decision and control the transfer case to switch to the target gear. The shift verification conditions include safety conditions such as vehicle speed threshold, transmission gear status, and valid braking signal.

[0033] The H mode corresponds to the transfer case being in the H position and the transfer case not locked; the HLock mode corresponds to the transfer case being in the H position and the transfer case locked; the LLock mode corresponds to the transfer case being in the L position and the transfer case locked.

[0034] This embodiment transforms the traditional manual gear-shifting process, which relies on the driver's subjective judgment, into a closed-loop automated process consisting of automatic road condition recognition, automatic mode decision-making, and automatic safety execution. The consistency verification between the switch position and the current gear before entering autonomous mode ensures the deterministic state of the system upon startup. The autonomous driving system autonomously determines the required driving mode based on information from the perception layer and automatically completes the transfer case gear shift, requiring no manual intervention throughout the entire process.

[0035] This embodiment achieves full automation of transfer case control, eliminating reliance on manual operation in autonomous driving scenarios. Through pre-emptive gear consistency and shift safety condition checks, it ensures a clear system startup status and safe, controllable shifting actions, avoiding the risk of transfer case damage or vehicle instability due to incorrect shifting timing, thus improving system safety and reliability.

[0036] Example 2: This example, based on Example 1, further specifies the concrete implementation method of road condition fusion recognition in step 102. The multi-source sensor data includes wheel speed sensor data, inertial measurement unit data, and radar data. The fusion processing of the multi-source sensor data to identify the current driving road conditions specifically includes the following sub-steps: Step 201: Extract features from wheel speed sensor data, inertial measurement unit data, and radar data respectively, and calculate the original probability of each sensor data corresponding to different road condition types.

[0037] ① Wheel speed sensor feature extraction: Calculate real-time slip ratio l : ,in oh For wheel speed, R The equivalent radius of the tire. The reference speed is calculated based on the average speed of the non-braking wheels.

[0038] Calculate wheel speed fluctuation entropy H Within a 2-second sliding time window, the wheel speed value is discretized into 20 equally spaced intervals, and the probability of the wheel speed falling into each interval is calculated. Then calculate .

[0039] Calculating the probability of low-adhesion road surfaces based on slip ratio: ,in This is the lower limit of the slip ratio threshold (e.g., 0.05; however, this value is not limited in actual applications and can be set according to actual needs). This is the upper limit of the slip ratio threshold (e.g., 0.30; however, this value is not limited in actual applications and can be set according to actual needs).

[0040] Calculating the probability of off-road surfaces based on wheel speed fluctuation entropy: in The reference vibration entropy threshold for paved roads (e.g., 1.0; however, this value is not limited in actual applications and can be set according to actual needs). The reference vibration entropy threshold for off-road surfaces (e.g., 3; this value is not limited in actual applications and can be set according to actual needs) needs to be calibrated on a real vehicle.

[0041] Probability of paved road surface: .

[0042] ②IMU feature extraction: Calculate the standard deviation of vertical vibration intensity First, calculate the mean vertical acceleration within a 100ms window (removing the gravitational acceleration component), then calculate... .

[0043] Probability of off-road terrain: , in The standard deviation of the vertical vibration intensity of the paved road surface (e.g., 0.3; this value is not limited in actual application and can be set according to actual needs). This is the standard deviation of the vertical vibration intensity on the off-road surface (e.g., 1.5~2.0; in actual applications, this range is not limited and can be set according to actual needs).

[0044] Probability of paved road surface: .

[0045] ③ Radar feature extraction: Extract the slope ahead i and road surface smoothness Ground points are segmented from the point cloud using RANSAC or GPF algorithms, and a plane is fitted for each road segment (5-10 meters ahead). Calculate longitudinal slope Calculate flatness based on ground point cloud elevation residuals .

[0046] Probability of off-road terrain: (Assuming a 30° slope).

[0047] Probability of paved road surface: .

[0048] Step 202: Using the DS evidence theory, the raw probabilities of the data from each sensor are fused to obtain the fused road condition recognition result. Construct the recognition framework. I ={paved road, low-adhesion road, off-road road}, and use the raw probabilities calculated by the above sensors as input to the Basic Probability Assignment (BPA) to execute the DS synthesis rule.

[0049] This embodiment makes extensive use of sensors with different physical phenomena to perceive road conditions: wheel speed sensors reflect the dynamic friction between the tires and the ground (the degree of slippage), IMUs reflect the intensity of vibrations of the vehicle body caused by road surface excitation, and radar directly measures the geometric features of the terrain ahead (slope, smoothness). DS evidence theory can effectively handle the uncertainties and conflicts between different sensor information, combining multiple pieces of evidence to form a more comprehensive and credible judgment.

[0050] Compared to single-sensor or simple weighted fusion, this embodiment significantly improves the accuracy and robustness of road condition recognition. For example, on muddy roads, the wheel speed sensor detects a high slip ratio (indicating low adhesion), the IMU detects severe vibration (indicating off-road), and the radar detects uneven terrain (indicating off-road). The three corroborate each other, making the confidence level for off-road or low-adhesion roads much higher than relying on only one sensor.

[0051] Example 3: This example, based on Example 2, modifies the original probabilities before fusion using DS evidence theory. Specifically, the method further includes the following steps: Step 301: Obtain the real-time signal-to-noise ratio (SNR) of each sensor. The SNR is calculated using techniques known in the art: for wheel speed sensors, the ratio of signal power (mean square value of effective wheel speed signal) to high-frequency noise power (obtained through high-pass filtering) is used as the SNR; for IMUs, the current SNR is estimated using their built-in temperature compensation model and noise variance; for radars, the SNR is calculated based on the ratio of echo signal strength to background noise strength.

[0052] Step 302: Based on the real-time signal-to-noise ratio of each sensor, correct the raw probabilities of each sensor's data to obtain the Basic Probability Assignment (BPA) for each sensor to different road condition subsets. The correction formula is as follows: in, For sensors i Road condition proposition The basic probability allocation value, For sensors i Real-time signal-to-noise ratio, N Total number of sensors (in this embodiment) N =3), For sensors i Road conditions The original probability estimate.

[0053] The basic probability allocation value is positively correlated with the signal-to-noise ratio (SNR) of the sensor; that is, the higher the SNR of the sensor, the greater the proportion of its BPA value to the total BPA.

[0054] In complex and ever-changing off-road environments, sensor signal quality can fluctuate dramatically. For example, wheel speed sensors may be covered in mud, resulting in weak signals (low SNR), while IMUs on smooth surfaces produce clear signals (high SNR). The core idea of ​​this embodiment is to give sensors with good signal quality (high SNR) a greater voice in fusion decision-making, while automatically reducing the contribution of sensors with poor signal quality (low SNR). This correction is dynamic and adaptive.

[0055] This embodiment effectively solves the shortcomings of traditional DS methods in terms of equal or fixed sensor weights, enabling the system to adapt to different environments: when rain or snow degrades radar signals, the system automatically places greater trust in wheel speed and IMU; when high-speed bumps cause high noise in the IMU signal, the system places greater trust in radar and wheel speed. This greatly improves the system's reliability in real, harsh environments.

[0056] Example 4: This example, based on Example 3, further refines the specific method for fusing the basic probability assignment values ​​using DS evidence theory. Specifically, it includes the following steps: Step 401: Calculate the dynamic conflict factor K(t) Conflict factors in traditional DS theory K This is the sum of the products of all propositions whose intersection is empty. This embodiment uses an improved dynamic conflict factor, calculated as follows: in, For dynamic weights calculated based on signal-to-noise ratio: .

[0057] When calculating collisions, this formula multiplies the BPA of each sensor by its signal-to-noise ratio weight, so that sensors with low signal-to-noise ratios have a lower weight in collision calculations, thus avoiding the collapse of the entire fusion system due to a single faulty sensor.

[0058] Step 402: Using the dynamic conflict factor, perform DS synthesis calculation on the basic probability assignment values ​​of multiple sensors to obtain the fusion confidence of each road condition proposition. The synthesis formula is as follows: in, To integrate road condition questions Confidence level, denominator This is the normalization factor.

[0059] Step 403: If the fusion confidence level is higher than the preset first confidence level threshold (set to 0.8 in this embodiment), the road condition recognition result is directly output, that is, the road condition proposition with the highest confidence level is selected as the final recognition result.

[0060] The core of the DS synthesis rule is the consensus principle, which states that the confidence level of a proposition will only significantly increase when multiple pieces of evidence jointly support it. Dynamic conflict factor. K(t) The degree of inconsistency between pieces of evidence is quantified. This embodiment introduces a signal-to-noise ratio weight into the conflict factor, ensuring that when a sensor malfunctions or is severely interfered with, its conflict with other sensors is not excessively amplified, thereby avoiding calculation anomalies or misjudgments caused by an excessively small normalized denominator.

[0061] This embodiment significantly improves the numerical stability and anti-interference capability of the fusion algorithm. The system has the ability to continue working even when some sensors fail or signal quality is severely degraded, and the entire road condition recognition function will not fail due to a brief loss of signal or a surge in noise from a single sensor. This is crucial for a safety-first autonomous driving system.

[0062] Example 5: This example supplements Example 4, addressing the scenario where the DS fusion confidence is insufficient. If the fusion confidence is not higher than the first confidence threshold (i.e. If the data shows a significant conflict or lack of confidence among the multi-sensor evidence, then the method further includes an adaptive weighted fusion phase, which specifically includes the following steps: Step 501: Calculate the dynamic fusion weights of each sensor based on their historical recognition accuracy and current data quality. The calculation formula is as follows: in, Let be the historical recognition accuracy of sensor i within the previous time window. The specific calculation method is as follows: within a sliding time window (e.g., 60 seconds), statistical analysis of sensor data is performed. i The percentage of times that the road condition recognition results are consistent with the vehicle dynamics back-inference results; For sensors i The current data quality can be directly assessed using its signal-to-noise ratio. ; α and β The adjustment coefficient satisfies α+β =1, in this embodiment α =0.4, β =0.6 indicates that more confidence is placed in the current data quality than in the historical accuracy.

[0063] Step 502: Calculate a weighted average of the original probabilities of each sensor based on the dynamic fusion weights to obtain the final road condition recognition result. The calculation formula is as follows: Then, the road condition with the highest probability is selected as the final recognition result: .

[0064] When DS fusion fails due to highly conflicting evidence, it essentially means that expert review has failed. At this point, the system degenerates into a more conservative weighted voting mechanism. This mechanism comprehensively considers each sensor's historical reputation (historical accuracy) and current performance (data quality), deriving a compromised and stable output through a weighted average.

[0065] This embodiment constructs a complete DSAW-Fusion (DS Evidence Theory and Adaptive Weighted Fusion) architecture, realizing adaptive switching of the fusion strategy. Under normal circumstances, high-precision DS fusion is used; under abnormal or conflicting conditions, it seamlessly switches to highly robust weighted average fusion. This ensures that the road condition recognition system can output a continuous and uninterrupted result under any circumstances, which is a key technology for building a highly available autonomous driving system.

[0066] Example 6: This example details the specific logic for generating a shift decision based on the road condition recognition result in step 103. The road condition-mode mapping relationship is as follows: When the road condition identification result is a paved road surface and its confidence level (from the fusion result of Example 4 or Example 5) is higher than the second confidence level threshold (set to 0.8 in this example, but this value is not limited in actual application and can be set according to actual needs), the transfer case shift decision is to switch to H mode. H mode is an intelligent four-wheel drive mode that can automatically distribute torque between the front and rear axles as needed while ensuring fuel economy.

[0067] When the road condition identification result is a low-adhesion road surface (such as icy, snowy, or waterlogged road surface) and its confidence level is higher than the third confidence level threshold (set to 0.7 in this embodiment, but this value is not limited in actual applications and can be set according to actual needs), the transfer case shift decision is to switch to HLock mode. In HLock mode, the transfer case is locked, and the torque of the front and rear axles is distributed in a fixed ratio (such as 50:50) to provide stable driving force and prevent slippage.

[0068] When the road condition identification result is an off-road surface (such as gravel, mud, sand, or potholes) and its confidence level is higher than the fourth confidence threshold (set to 0.75 in this embodiment, but this value is not limited in actual applications and can be set according to actual needs), the transfer case shifts to LLock mode. In LLock mode, the transfer case not only locks but also provides torque amplification through the planetary gear set (such as amplification by 2.5 times, but this value is not limited in actual applications and can be set according to actual needs) to provide ultimate climbing and getting-out-of-trouble capabilities.

[0069] When the confidence level of the road condition recognition result is lower than any of the aforementioned thresholds, the transfer case shifts to maintain the current gear to avoid risky shifting operations due to uncertainty in recognition.

[0070] This embodiment decouples the continuous road condition recognition confidence level from the discrete shift decision. Different road conditions have different risk tolerances for shift decisions. For example, accidentally entering LLock mode (off-road low-speed lock mode) on paved roads can lead to steering difficulties, transmission system wear, and even rollover risks. Therefore, LLock mode requires a higher confidence threshold to trigger; while H mode is relatively safe, and the threshold can be appropriately lowered.

[0071] This embodiment provides a safe and flexible gear shifting decision mechanism. By setting differentiated confidence thresholds, it effectively avoids hesitation or blind gear shifting when the road condition recognition boundary is ambiguous, especially maintaining necessary caution when switching to high-risk off-road modes, significantly improving the system's decision-making safety.

[0072] Example 7: This example details the specific content of the preset shift verification conditions in step 104. The shift verification conditions include: the matching relationship between the current transmission gear, the current vehicle speed, the braking signal, and the target gear. The matching relationship is determined according to a preset shift condition table, which specifies the allowed range of current transmission gears, the upper limit of vehicle speed, and whether a valid braking signal is required for different target gears.

[0073] In this embodiment, the exemplary shift condition table is as follows: Note: The above thresholds can be adjusted based on the actual vehicle calibration results.

[0074] Transfer cases typically employ gear or gear sleeve structures internally. Shifting requires reducing or eliminating the speed difference and load between the driving and driven gears. Therefore, the safest condition for shifting is when the vehicle is nearly stationary (extremely low speed), the transmission is in neutral to cut off power input, and under certain conditions, braking is also necessary to prevent rolling. This embodiment quantifies these physical constraints into specific judgment logic and uses a lookup table method for efficient verification.

[0075] This embodiment ensures that every gear shift is performed within the mechanically permissible safety boundaries, maximizing the protection of core components such as transfer case gears, synchronizers, and engagement sleeves. It avoids serious malfunctions such as gear grinding and damage caused by shifting at high speeds or under load, thus extending the service life of the transfer case.

[0076] Example 8: This example adds closed-loop verification and fault diagnosis capabilities to the system. The method further includes the following steps: Step 801: After the transfer case shifts gears, the actual coefficient of adhesion on the current road surface is calculated using the vehicle's longitudinal dynamics model. The basic formula is as follows: in, The driving force (which can be calculated from the engine output torque, transmission ratio, final drive ratio, transmission efficiency, and wheel radius) m For the overall vehicle quality, a This is the longitudinal acceleration (measured by the IMU). The drag force (including rolling resistance, air resistance, etc.) can be estimated using a calibrated model. This allows for the estimation of the actual coefficient of adhesion. ,in g This represents gravitational acceleration. For brake wheels, a similar method can be used to calculate the coefficient of adhesion based on the braking force.

[0077] Step 802: Calculate the actual adhesion coefficient obtained by reverse calculation. The expected adhesion coefficient range corresponding to the road condition identification result is compared. For example, the expected adhesion coefficient of paved road surfaces. m >0.7, expected coefficient of adhesion for icy and snowy roads m <0.3, the expected coefficient of adhesion for off-road surfaces is between 0.2 and 0.6.

[0078] Step 803: If the deviation exceeds a preset threshold (for example, If the absolute difference from the theoretical center value is greater than 0.4, the sensor is determined to be abnormal, and the basic probability allocation value or fusion weight of the sensor determined to be abnormal is reduced in the subsequent road condition recognition fusion process. Specifically, the system can reduce the historical accuracy of sensors determined to be suspicious. Alternatively, its signal-to-noise ratio (SNR) participation can be directly reduced, decreasing its influence in subsequent fusion. Simultaneously, the system records fault codes and sends alarm information to the upper-level controller via the vehicle network.

[0079] Road condition recognition is essentially based on inference from indirect features, while vehicle dynamic response (the relationship between actual acceleration and driving force) is a direct physical manifestation of road surface adhesion characteristics. This embodiment uses the execution results to reverse-verify the decision-making basis, forming a closed loop. If there is a significant contradiction between the decision-making basis (road condition recognition results) and the execution results (vehicle dynamic response), it indicates that there may be a problem at the perception level of the system.

[0080] This embodiment endows the system with self-diagnostic and fault-tolerant repair capabilities. It can identify whether sensors are aging, malfunctioning, or obstructed (such as wheel speed sensors being covered in mud or radar lenses being dirty), and automatically adjust the fusion strategy to isolate the impact of faults, preventing the system from continuously making incorrect decisions due to erroneous input. This is a key aspect of improving the long-term operational reliability of autonomous driving systems.

[0081] Example 9: This example describes how the system safely exits autonomous driving mode. The method further includes exiting the autonomous driving mode in response to any of the following conditions: 1. Received driverless mode disabled signal: When the system receives the falling edge or set signal of the transfer case driverless mode disabled signal, the exit process is triggered.

[0082] 2. Driver operation of transfer case switch detected: When the system detects that the driver has actively operated the transfer case switch (the switch position has changed), it immediately exits the unmanned mode.

[0083] 3. System fault flag bit is detected: When the system self-test or feedback verification in Example 8 detects a serious fault (such as gear shift actuator jamming, position sensor signal loss, multiple sensors failing at the same time, etc.) and sets the fault flag bit to 1, the unmanned mode is forcibly exited.

[0084] When exiting the autonomous driving mode, the system reads the current physical position signal of the transfer case switch. If the current actual gear position is inconsistent with the gear position indicated by the switch signal, the system controls the transfer case to switch to the gear position indicated by the switch signal when the shift verification condition is met. If the current condition does not meet the shift safety requirements, the system maintains the current gear position until the condition is met, or handles the situation according to the preset safety strategy (such as requesting the vehicle to slow down and stop).

[0085] Exiting autonomous driving mode must be designed to be both smooth and safe. This embodiment distinguishes between different scenarios, including planned exit (enable signal disappearance), human takeover (driver intervention), and emergency exit (system failure), and designs differentiated processing logic for each scenario. Upon exiting, the transfer case is preferentially switched to the position indicated by the switch signal to achieve a smooth transfer of control; if the conditions are not met, the switch is delayed or safety measures are taken to avoid danger caused by forced gear shifting.

[0086] This embodiment provides clear and comprehensive mode switching boundary conditions. Driver intervention is prioritized to ensure smooth control handover during human-machine co-driving; a strict fault exit mechanism ensures the system will not operate while malfunctioning; and the strategy of maintaining the gear position when conditions are not met prevents loss of vehicle control caused by forcibly resetting the gear position under unsafe conditions. This fully embodies the design principle of safety first.

[0087] Specifically, such as Figure 1 As shown, the transfer case control method for an unmanned vehicle according to the present invention is performed according to the following steps: Step S1: Enter driverless mode After the vehicle completes its power-on self-test, the autonomous driving system sends an autonomous driving enable signal (active status) to the transfer case controller (TMM). The TMM reads the physical position signal of the transfer case switch. If the gear indicated by the position signal matches the actual gear position of the transfer case, the TMM controls the transfer case system to enter autonomous driving mode, and simultaneously initializes all sensor data buffers, fusion parameters, and counters.

[0088] Step S2: Multi-source sensor data acquisition and preprocessing TMM synchronously acquires raw data from wheel speed sensors (4 wheels), IMUs (three-axis acceleration and three-axis angular velocity), and radar / lidar at fixed intervals (e.g., 10ms). Time synchronization is performed by interpolating high-frequency data (e.g., wheel speed 100Hz) to the time points of low-frequency data (e.g., radar 10Hz), or by uniformly downsampling to the least common multiple of the sampling frequencies of the three sensors. Spatial coordinate unification is performed by converting all sensor data to the vehicle coordinate system (with the front axle center as the origin, x-axis forward, y-axis left, and z-axis upward).

[0089] Step S3: Feature Extraction and Original Probability Calculation Following the method described in Example 2, the slip ratio of the wheel speed sensor was calculated respectively. l Wheel speed fluctuation entropy H Standard deviation of vertical vibration intensity of IMU The slope in front of the radar i and road surface smoothness Then, the original probability of each sensor for paved roads, low-adhesion roads, and off-road roads is calculated. .

[0090] Step S4: Basic probability allocation based on signal-to-noise ratio correction Obtain the real-time signal-to-noise ratio of each sensor. , , Calculate the total signal-to-noise ratio. S SNR Then follow the formula. Calculate the basic probability assignment value for each sensor for each road condition proposition.

[0091] Step S5: Fusion of DS Evidence Theory Calculate dynamic conflict factor K(t) : ,in .

[0092] Calculate fusion confidence : .

[0093] like If the value is ≥0.8, output the road condition recognition result and proceed to step S7.

[0094] Step S6: Adaptive Weighted Fusion like If the value is less than 0.8, then switch to adaptive weighted fusion.

[0095] Calculate the dynamic weights of each sensor: .

[0096] Calculate the final probability: .

[0097] Output traffic conditions .

[0098] Step S7: Generate shift decision Based on the road condition recognition results and their confidence levels, the target shift command is generated according to the mapping relationship: Paved road with confidence level ≥ 0.8 → Target gear = H Low-adhesion road with confidence level ≥ 0.7 → Target gear = HLock Off-road terrain with confidence level ≥ 0.75 → Target gear = LLock For the others, the target gear is to maintain the current gear. Step S8: Shift condition verification and execution Consult the shift condition table to check the current transmission gear, vehicle speed, and brake signal. If all conditions are met, the TMM sends the target gear signal to the solenoid valve. The solenoid valve then opens the corresponding valve, moving the engagement sleeve or shift fork to complete the gear shift. The position sensor provides feedback to confirm the successful shift.

[0099] Step S9: Dynamic Feedback Verification After the gear shift is completed, the actual adhesion coefficient is calculated by using the vehicle's longitudinal dynamics model. .like If the deviation from the expected adhesion coefficient corresponding to the road condition recognition result is greater than 0.4, the sensor is judged to be abnormal, and the historical accuracy of the corresponding sensor is reduced. And record the fault code.

[0100] Step S10: Exit Condition Monitoring Throughout the process, TMM continuously monitors the exit conditions: If an autonomous driving inactivation signal is received, the exit procedure is executed: check whether the current gear position is consistent with the gear position indicated by the switch signal. If they are inconsistent and the shift verification conditions are met, switch to the switch gear position and exit; otherwise, maintain the current gear position until the conditions are met.

[0101] If the driver is detected operating the transfer case switch, immediately exit the unmanned mode and attempt to shift gears by pressing the switch position (the shift verification conditions must be met).

[0102] If the fault flag is 1, the driver will be forced to exit the unmanned mode and the vehicle will be requested to slow down and stop.

[0103] Through the above comprehensive process, the embodiments of the present invention can realize safe, reliable and intelligent automated control of the transfer case in unmanned vehicles under complex and ever-changing working conditions, significantly improving the vehicle's passability, stability and fuel economy.

[0104] In addition, the present invention also relates to a transfer case control system for an unmanned vehicle that implements all of the above method embodiments.

[0105] The system includes the following functional modules, which can be integrated into the vehicle's existing domain controllers, such as the transfer case controller (TMM), vehicle control unit (VCU), or automated driving domain controller (ADCU): Mode Management Module: This module controls the vehicle to enter autonomous driving mode in response to a valid autonomous driving enable signal and the transfer case switch position matching the current gear. This module monitors the enable signal and switch status, coordinates the switching between autonomous and manual modes, and executes the mode entry and exit logic described in the above embodiments.

[0106] Road condition recognition module: This module acquires multi-source sensor data (including wheel speed sensor, IMU, and radar) from the vehicle and fuses the data to identify the current road conditions and obtain a road condition recognition result. Internally, this module includes the feature extraction unit, signal-to-noise ratio-based BPA correction unit, DS fusion unit, and adaptive weighted fusion unit described in the above embodiments.

[0107] Shift decision module: This module generates transfer case shift decisions based on the road condition identification results and a preset road condition-mode mapping relationship. Internally, this module stores the mapping table and confidence threshold values ​​described in the above embodiments and executes the decision logic.

[0108] The shift execution module is used to execute the transfer case shift decision and control the transfer case to switch to the target gear when the preset shift verification conditions are met. This module internally stores the shift condition table described in the above embodiments, drives the shift motor or solenoid valve to complete the gear shift, and simultaneously receives feedback from the position sensor for closed-loop control.

[0109] Exit Management Module: This module controls the vehicle to exit the autonomous driving mode in response to exit conditions. It monitors the three exit conditions described in the above embodiments in real time and executes the corresponding exit procedures, including gear shifting and maintaining a safe state when the conditions are met.

[0110] This embodiment modularizes the functions corresponding to the above-described method embodiments, forming a complete transfer case control system architecture. The modules communicate and interact with each other via the vehicle's CAN bus or internal signal interfaces, collaboratively completing the entire control process from road condition perception to gear shift execution.

[0111] The system provided in this embodiment enables safe, reliable, and intelligent control of the transfer case in autonomous vehicles under complex and changing operating conditions. The system architecture is clear, with well-decoupled modules, facilitating subsequent functional upgrades and maintenance. Furthermore, the system possesses advanced functions such as adaptive fusion, fault diagnosis, and safe exit, significantly improving the passability and driving stability of autonomous vehicles in complex scenarios such as wilderness, mines, and emergency rescue.

[0112] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A transfer case control method for an unmanned vehicle, characterized in that, Applied to autonomous vehicles with a transfer case, the method includes: In response to the autonomous driving enable signal being valid and the transfer case switch position being consistent with the current gear, the autonomous driving mode is entered. Acquire multi-source sensor data of the vehicle and calculate the raw probability of each sensor data corresponding to different road condition types; The original probabilities are fused using the DS evidence theory and based on the real-time signal-to-noise ratio of each sensor to identify the current driving conditions and obtain the road condition identification result. Based on the road condition recognition results and the preset road condition-mode mapping relationship, a transfer case shifting decision is generated; Under the condition that the preset shift verification conditions are met, the transfer case shift decision is executed, and the transfer case is controlled to switch to the target gear.

2. The transfer case control method for an unmanned vehicle according to claim 1, characterized in that, The multi-source sensor data includes wheel speed sensor data, inertial measurement unit data, and radar data; the multi-source sensor data is fused to identify the current road conditions, specifically including: Feature extraction was performed on wheel speed sensor data, inertial measurement unit data, and radar data respectively, and the original probability of each sensor data corresponding to different road condition types was calculated. By employing the DS evidence theory, the original probabilities of data from each sensor are fused to obtain the fused road condition recognition result.

3. The transfer case control method for an unmanned vehicle according to claim 2, characterized in that, Also includes: Obtain the real-time signal-to-noise ratio of each sensor; Based on the real-time signal-to-noise ratio of each sensor, the original probability of each sensor data is corrected to obtain the basic probability allocation value of each sensor for different road condition subsets. The basic probability assignment value is positively correlated with the signal-to-noise ratio of the sensor.

4. The transfer case control method for an unmanned vehicle according to claim 3, characterized in that, Methods for fusing the basic probability assignment values ​​using DS evidence theory include: The dynamic conflict factor is calculated based on the basic probability allocation values ​​of each sensor and the dynamic weights calculated according to the real-time signal-to-noise ratio of each sensor. Using the dynamic conflict factor, the basic probability assignment values ​​of multiple sensors are calculated by DS synthesis to obtain the fusion confidence of each road condition proposition; If the fusion confidence level is higher than the preset first confidence level threshold, the road condition recognition result is directly output.

5. The transfer case control method for an unmanned vehicle according to claim 4, characterized in that, If the fusion confidence level is not higher than the first confidence threshold, the method further includes: Entering the adaptive weighted fusion stage: Based on the historical recognition accuracy of each sensor and the current data quality, calculate the dynamic fusion weight of each sensor; The original probabilities of each sensor are weighted and averaged based on the dynamic fusion weights to obtain the final road condition recognition result.

6. The transfer case control method for an unmanned vehicle according to claim 1, characterized in that, The road condition-mode mapping relationship includes: When the road condition identification result is a paved road surface and the confidence level is higher than the second confidence threshold, the transfer case shifts to H mode. When the road condition identification result is a low-adhesion road surface and the confidence level is higher than the third confidence level threshold, the transfer case shift decision is to switch to HLock mode; When the road condition recognition result is an off-road surface and the confidence level is higher than the fourth confidence level threshold, the transfer case shift decision is to switch to LLock mode; When the confidence level of the road condition recognition result is lower than any of the thresholds, the transfer case shifts to maintain the current gear.

7. The transfer case control method for an unmanned vehicle according to claim 1, characterized in that, The shift verification conditions include: the matching relationship between the current transmission gear, the current vehicle speed, the braking signal, and the target gear; the matching relationship is determined according to a preset shift condition table, which specifies the allowed range of current transmission gears, the upper limit of vehicle speed, and whether the braking signal needs to be valid for different target gears.

8. The transfer case control method for an unmanned vehicle according to claim 1, characterized in that, The method further includes: after the transfer case performs a gear shift, using the vehicle longitudinal dynamics model to back-calculate the actual adhesion coefficient of the current road surface, and comparing the actual adhesion coefficient with the expected adhesion coefficient corresponding to the road condition recognition result; if the deviation exceeds a preset threshold, the sensor is determined to be abnormal, and the basic probability allocation value or fusion weight of the sensor determined to be abnormal is reduced in the subsequent road condition recognition fusion process.

9. The transfer case control method for an unmanned vehicle according to claim 1, characterized in that, The method further includes exiting the autonomous driving mode in response to any of the following conditions: Received a signal indicating that autonomous driving is disabled; Driver operation of transfer case switch detected; A system fault flag bit has been detected being set. When exiting the autonomous driving mode, if the current gear is inconsistent with the gear indicated by the switch signal, the transfer case will be controlled to switch to the gear indicated by the switch signal when the shift verification condition is met.

10. A transfer case control system for an unmanned vehicle, characterized in that, A transfer case control method for an unmanned vehicle according to any one of claims 1-9 includes: The mode management module is used to control the vehicle to enter the autonomous driving mode in response to the autonomous driving enable signal being valid and the transfer case switch position being consistent with the current gear. The road condition recognition module is used to acquire multi-source sensor data of the vehicle, calculate the original probability of each sensor data corresponding to different road condition types, and fuse the original probabilities based on the DS evidence theory and the real-time signal-to-noise ratio of each sensor to identify the current driving road condition and obtain the road condition recognition result. The shift decision module is used to generate a transfer case shift decision based on the road condition recognition result and the preset road condition-mode mapping relationship; The shift execution module is used to execute the transfer case shift decision and control the transfer case to switch to the target gear when the preset shift verification conditions are met. The exit management module is used to control the vehicle to exit the autonomous driving mode in response to exit conditions.

Images