A new energy vehicle navigation communication control method and system

Abstract

The application relates to the technical field of new energy vehicle navigation communication control, and discloses a new energy vehicle navigation communication control method and system, which comprises the following steps: predicting a poor communication area and pre-switching vehicle control right to local perception priority, in the local perception priority mode, a dynamic risk potential field is constructed by quantifying the uncertainty of multi-modal perception data, and a risk avoidance path is planned according to the potential field, and a separate safety guardian layer is arranged to execute a reflexive emergency avoidance operation when the state of the main control system is unstable, the application establishes a pre-positioned vehicle control right architecture adjustment mechanism, avoids inherent decision delay caused by conflicts between communication interruption and local perception, and solves the decision-making problem of the vehicle under the superimposed risk of external information interruption and self-perception limitation through a progressive strategy of constructing a risk potential field by quantifying perception uncertainty and setting a separate safety guardian layer.

Description

Technical Field

[0001] This invention relates to a navigation and communication control method and system for new energy vehicles, belonging to the field of new energy vehicle navigation and communication control technology. Background Technology

[0002] Currently, vehicle navigation and communication control systems are core components ensuring driving safety and efficiency. They achieve high-precision path planning and real-time traffic response through data interaction between the vehicle's communication module and a cloud service platform, a common technical solution. This approach performs well in environments with good communication signals, providing vehicles with globally optimal driving strategies, and its value has been widely validated by the market. However, as vehicle driving scenarios become increasingly complex and involve long distances, the aforementioned control method, which relies on stable external communication, must switch between two data sources with different timeliness and reliability when facing areas with foreseeable communication signal attenuation or interruption, such as long tunnels or remote mountainous areas. The commonly used strategy is to passively transfer decision-making weights to the local perception system after communication interruption. This reactive switching mechanism inherently carries a potential risk.

[0003] At a deeper level, this risk is not simply a functional deficiency, but an inherent contradiction stemming from the inertia of system design. When a vehicle is traveling at high speed toward the boundary of a communication interruption zone, the system is on the edge of an information cliff. The navigation system, based on the last information before the communication interruption, is still instructing the vehicle to maintain a driving state where everything ahead is normal. Although the local perception system is continuously working, in the priority ranking of system control, its judgment weight for disruptive risks has not been raised to the highest level. This leads to a brief but fatal decision vacuum in the critical state where communication interruption and a sudden serious accident ahead occur almost simultaneously. The system needs to process the violent conflict between navigation information failure and local perception information alarms in a very short time. This conflict processing mechanism itself increases the system's reaction time.

[0004] Specifically, existing technologies suffer from the following shortcomings: 1. Lag in control logic: the system only begins adjusting decision priorities after communication is confirmed to have been interrupted, making it unable to cope with sudden dangers occurring simultaneously near the communication interruption point; 2. Conflict in decision-making mechanisms: when lagging navigation information severely conflicts with local real-time perception, existing arbitration mechanisms increase the complexity and time overhead of the decision-making chain, easily missing the best avoidance opportunity in high-speed scenarios. To address these issues, the industry has attempted to shorten decision-making time by enhancing local sensor performance or optimizing data fusion algorithms, but these improvements do not address the root cause—they still optimize after a conflict occurs, failing to fundamentally eliminate the decision vacuum caused by the lag in control switching. As long as this passive, post-event switching mechanism exists, potential collision risks cannot be completely avoided. Therefore, the technical problem this invention aims to solve is how to establish a control mechanism that proactively avoids potential decision conflicts between navigation information and local perception information at the control architecture level before the vehicle enters a predictably communication-restricted area, thereby preventing the system from falling into decision lag in the most dangerous critical state. Summary of the Invention

[0005] This invention provides a navigation and communication control method and system for new energy vehicles. Its main purpose is to solve the problem of decision delay and safety risks caused by the conflict between navigation information lag and local perception information when the vehicle is about to enter a communication-restricted area. This is achieved by adjusting the control architecture in a forward-looking manner.

[0006] To achieve the above objectives, the present invention provides a navigation and communication control method for new energy vehicles, the method comprising: Before the vehicle enters an area with poor communication, a control mode switching trigger time is calculated based on a communication quality prediction attribute marked in the vehicle map data and the vehicle's speed and position. At the calculated trigger time, the vehicle's central control unit performs a control pre-switching operation. This operation sets an internal control mode flag, so that subsequent vehicle control commands will preferentially adopt the fused perception results from a local multimodal sensor system and suppress the adoption of path commands from the navigation system. After performing the control pre-switching operation, the quality of the data output by each sensor mode in the multimodal sensor system is evaluated in real time to generate a quantized uncertainty value for each. In a local three-dimensional grid map centered on the vehicle, a risk potential energy of each grid cell in the local three-dimensional grid map is calculated based on the quantized uncertainty value, the probability of the existence of obstacles in the grid, and the collision impact level. Based on the constructed risk potential field, a trajectory that minimizes the risk potential energy on the driving path while satisfying the vehicle's own kinematic constraints is planned and selected. Furthermore, a safety decision microcontroller, independent of the central control unit, continuously receives raw sensor data via an independent bus and calculates a collision risk index. When the calculated collision risk index exceeds the threshold, it bypasses the central control unit and directly issues emergency avoidance commands to the vehicle's brake actuators and steering actuators, and sends a bypass status notification to the central control unit for post-intervention status calibration and control revert.

[0007] Preferably, the step of evaluating the quality of the data output by each sensor mode in a multimodal sensor system in real time to generate a quantized uncertainty value for each mode is as follows: for a high-definition camera, a visual perception uncertainty value is generated by analyzing the pixel grayscale distribution and the number of effective feature points of the output image; for a lidar, a spatial perception uncertainty value is generated by evaluating the sparsity of the output point cloud data and the completeness of the target contour description; for a millimeter-wave radar, a classification confidence uncertainty value is generated by analyzing the radar cross-section characteristics of its echo signal and the degree of matching with the object model.

[0008] Preferably, each grid cell in a local 3D raster map ( Risk potential energy The calculation follows these rules: ,in, For grid cells ( Within ), it is a comprehensive uncertainty value obtained by weighting and fusing the visual perception uncertainty value, the spatial perception uncertainty value, and the classification confidence uncertainty value according to their respective weights; For grid cells ( Within a given area, the probability of an obstacle's existence is output after detection by sensors and processing by a target confirmation algorithm. The calculation based on the vehicle's current speed and braking capacity is the result if the vehicle enters the grid cell. A collision impact level; , and The weighting coefficients are determined by configuration parameters loaded during system initialization to balance the importance of the three factors.

[0009] Preferably, the safety decision microcontroller, independent of the central control unit, is a microcontroller that conforms to the Automotive Safety Integrity Level (ASIL-D) standard; the independent bus is a dedicated controller area network (CLAN) safety bus; the collision risk index calculation integrates a predicted time to collision (TTC) and a collision probability directly calculated from raw sensor data; the emergency avoidance command is specifically an emergency braking command with maximum safe deceleration and an emergency lane change assist command that limits both steering amplitude and steering angular velocity when a safe space is confirmed in the adjacent lane.

[0010] Preferably, the communication quality prediction attribute marked in the vehicle map data is one or more of the following: tunnel length and location identifiers, mountain terrain feature data, or geometric information of large underground passages.

[0011] Preferably, the method further includes: after the vehicle leaves the area with poor communication, when the network signal strength detected by an on-board communication module continuously exceeds a stable communication threshold for a period of time, the central control unit resets the internal control mode flag to restore the navigation command driven mode from the local perception priority mode.

[0012] Preferably, a bypass status notification is sent to the central control unit for post-intervention status calibration and control handover. Specifically, when the safety decision microcontroller activates the emergency avoidance command, it sends a bypass start signal to the central control unit via an independent bus. Upon receiving the bypass start signal, the central control unit suspends its own path planning and control command output. After the emergency avoidance operation is completed and the vehicle stabilizes, the safety decision microcontroller sends a bypass end signal. Upon receiving the bypass end signal, the central control unit first performs a status calibration using the vehicle's real-time pose and the latest environmental perception results, and then resumes its control functions.

[0013] Preferred, comprehensive uncertainty value The calculation also includes adaptively adjusting the weights of the visual perception uncertainty, spatial perception uncertainty, and classification confidence uncertainty used for weighted fusion based on the type of environment in which the vehicle is currently located.

[0014] Preferably, the planning and selection of a trajectory with the minimum risk potential energy on the driving path is achieved through a model predictive control algorithm. In each decision cycle, the algorithm generates multiple candidate trajectories that meet the constraints of the vehicle's maximum steering angle and maximum acceleration / deceleration, and calculates the cumulative value of the risk potential energy of all grid cells that each candidate trajectory will pass through within a prediction time window. Finally, the candidate trajectory with the minimum cumulative value of risk potential energy is selected as the execution trajectory. The trigger time for the control mode switching is calculated as when the vehicle is 300 to 800 meters away from the boundary of the poor communication area entrance. The condition for the collision risk index to exceed the threshold is that the calculated predicted collision time (TTC) value is less than 2.0 seconds.

[0015] A navigation and communication control system for new energy vehicles, the system comprising: A multimodal sensor system; A brake actuator; A steering actuator; A single independent bus; A central control unit connected to the multimodal sensor system; A safety decision microcontroller is connected to a multimodal sensor system via an independent bus, and also to the brake actuator, steering actuator, and central control unit. The central control unit is configured to: before the vehicle enters a communication-poor area, calculate a control mode switching trigger time based on a predicted communication quality attribute calibrated in the onboard map data and the vehicle's speed and position; and at the calculated trigger time, perform a control pre-switching operation to set an internal control mode flag. This flag ensures that subsequent vehicle control commands prioritize the fused perception results from the multimodal sensor system and suppress the adoption of path commands from the navigation system. After performing the control pre-switching operation, the quality of the data output by each sensor mode in the multimodal sensor system is evaluated in real time to generate... Each component has a quantified uncertainty value, and in a vehicle-centric local 3D grid map, based on the quantified uncertainty value, the probability of obstacle presence within the grid, and the collision impact level, a risk potential energy is calculated for each grid cell in the local 3D grid map. Based on the constructed risk potential field, a trajectory that minimizes the risk potential energy on the driving path while satisfying the vehicle's kinematic constraints is planned and selected. The safety decision microcontroller is configured to operate independently of the central control unit, continuously receive raw sensor data from the multimodal sensor system, and calculate a collision risk index. When the calculated collision risk index exceeds a threshold, it bypasses the central control unit, directly issues emergency avoidance commands to the brake actuators and steering actuators, and sends a bypass status notification to the central control unit for post-intervention status calibration and control revert.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By pre-judging changes in the communication environment to initiate dynamic switching of control modes, a forward-looking vehicle control architecture adjustment mechanism is established. Before entering a signal-restricted area, the system does not simply switch to a backup information source, but fundamentally transfers absolute control of navigation commands to the local perception system in advance. This non-conflict-based arbitration approach allows the vehicle's response path to naturally bypass the information source conflict judgment process when facing sudden obstacles. Structurally, it avoids the inherent time delay caused by data comparison and decision arbitration, making the triggering of emergency avoidance actions depend solely on a single link of local physical perception, directly improving the vehicle's reaction certainty in specific risky road sections. 2. When a vehicle enters an environment with reduced visibility, this invention performs real-time uncertainty assessment on the data quality of multiple sensors, such as high-definition cameras, lidar, and millimeter-wave radar, and integrates the quantified uncertainty value into a three-dimensional grid map to construct a dynamic risk potential field. This process transforms the fuzzy obstacles that are difficult to identify accurately due to the deterioration of sensor data quality in traditional technologies into areas with clear high-risk potential energy on the map. The vehicle's risk avoidance decision-making basis also changes from identifying specific obstacles to avoiding high-risk potential energy areas. This allows the system to generate effective tactical avoidance paths even when it cannot accurately perceive obstacle details, solving the problem of a sharp decline in decision-making ability when some functions of the local perception system are impaired. 3. The independent threat perception and reflective safety decision triggering mechanism established in this invention constitutes a safety protection link that operates in parallel with the main central control unit and has a higher execution priority. In extreme cases where the main control system becomes unstable in control mode switching due to unforeseen communication interference, or its own decision logic becomes sluggish, this independent link directly assesses the urgency of the collision through a dedicated processor and sensors, and bypasses the main control unit to issue basic avoidance commands to the vehicle's braking and steering systems. This architecture provides the vehicle control system with a low-level safety guarantee that does not depend on the current operating state of the main system, and addresses the protection gaps that may occur when the main control system is in an abnormal state. This invention combines two mechanisms—predicting changes in the communication environment and dynamically switching control modes, and constructing a real-time risk potential field—to form a progressive risk response strategy. First, the vehicle mitigates potential information conflict risks before entering a specific area through prediction and control switching. Then, if local perception capabilities are impaired due to environmental degradation, the risk potential field mechanism serves as a second layer of protection, solving the decision-making problem under conditions of ambiguous perception information. The synergistic effect of these two mechanisms enables the vehicle to possess a hierarchical and systematic response capability when facing two different and potentially overlapping risks: external information interruption and limited self-perception, rather than relying on a single emergency plan. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the forward-looking two-layer security control logic of the present invention. Figure 2 This is a diagram showing the relationship between navigation and local perception mode switching in this invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides a navigation and communication control method and system for new energy vehicles, comprising three stages: predicting the communication environment and performing pre-control switching to construct a dynamic risk potential field and plan a risk avoidance path, and setting up an independent reflective safety protection layer. During vehicle operation, when the communication link is stable, the vehicle control system plans the path based on data interacted with the cloud service platform. However, when the vehicle enters a predictably poor communication area such as a long tunnel or mountainous region, and simultaneously encounters a sudden road accident ahead, the traditional method of passively switching control after communication interruption will result in decision-making delays due to conflicts between handling navigation information failure and local perception alarms. To address this issue, the first stage of this invention is configured to perform a pre-control switching operation. Specifically, during path planning, the vehicle's central control unit scans onboard map data, which includes a communication quality prediction attribute, such as tunnel length and location markers, mountainous terrain feature data, or geometric information of large underground passages. The central control unit then uses this communication quality prediction attribute, combined with the vehicle's real-time driving data... Based on driving speed and location information, a trigger time for switching control modes is calculated. One procedure for determining this trigger time is to set it when the vehicle is 300 to 800 meters away from the boundary of the entrance to a communication-poor area. For example, when a vehicle is driving towards the entrance at 100 km / h, if the trigger distance is set to 500 meters, the system calculates that it will enter the signal attenuation zone in approximately 18 seconds and uses this as the trigger time. At the calculated trigger time, the vehicle's central control unit performs a pre-control switching operation. By setting an internal control mode flag, the logic for generating vehicle control commands is switched from navigation command-driven mode to local perception priority mode. In this mode, the fused perception results from a multimodal sensor system located locally in the vehicle are set as the highest decision priority, while path commands from the navigation system are suppressed. By adjusting the control architecture before entering a signal-restricted area, the vehicle's response path when facing subsequent sudden obstacles does not include the comparison and arbitration of conflicting information sources, thus avoiding corresponding decision delays.

[0020] After performing the control handover operation, if the vehicle encounters reduced visibility in a poorly oriented area, leading to decreased data quality from the local sensor system and difficulty in identifying obstacles with high confidence, the second stage of this invention involves constructing a dynamic risk potential field for obstacle avoidance path planning. This stage first evaluates the output data quality of each sensor mode in the multimodal sensor system in real time to generate its respective quantized uncertainty value. This evaluation process is as follows: for high-definition cameras, a visual perception uncertainty value is generated by analyzing the pixel grayscale distribution and the number of effective feature points in the output image; for lidar, a spatial perception uncertainty value is generated by evaluating the sparsity of the output point cloud data and the completeness of the target contour description; for millimeter-wave radar, a classification confidence uncertainty value is generated by analyzing the radar cross-section (RCS) characteristics of its echo signal and the degree of matching with a preset object model. Subsequently, in a local 3D grid map centered on the vehicle, based on the quantized uncertainty values ​​generated in the aforementioned steps, and combined with the probability of obstacle presence and collision impact level within the grid, the risk potential energy of each grid cell in the local 3D grid map is calculated. Each grid cell... Risk potential The calculation follows these rules: In this formula, Within a grid cell, the comprehensive uncertainty value is obtained by weighting and fusing the visual perception uncertainty value, spatial perception uncertainty value, and classification confidence uncertainty value according to their respective weights. Furthermore, the weights can be adaptively adjusted according to the type of environment in which the vehicle is currently located. The probability of an obstacle existing within a grid cell, after being detected and processed by a sensor. This is a collision impact level calculated based on the vehicle's current speed and braking capacity if the vehicle enters a grid cell. , and The weighting coefficients, determined by the configuration parameters loaded during system initialization, are used to balance the importance of the three factors. Based on the constructed risk potential field, the vehicle's path planning module, such as a model predictive control algorithm, plans and selects a trajectory that minimizes the cumulative risk potential energy on the driving path while satisfying the vehicle's kinematic constraints, as the execution trajectory.

[0021] In extreme cases, if the vehicle's control mode switching process becomes unstable due to unforeseen transient electromagnetic interference, and a sudden hazard occurs ahead, this invention provides a third stage: an independent safety protection layer. Its core is a safety decision microcontroller independent of the central control unit. This microcontroller conforms to the Automotive Safety Integrity Level (ASIL-D) standard and continuously receives raw sensor data via an independent controller area network (CLAN) security bus, calculating a collision risk index. This index integrates the estimated time to collision (TTC) and the probability of collision. When the calculated collision risk index exceeds a threshold, for example, when the TTC is less than 2.0 seconds, the safety decision microcontroller bypasses the central control unit and directly issues emergency avoidance commands to the vehicle's brake and steering actuators. These commands include an emergency braking command with maximum safe deceleration and a command to confirm the adjacent lane... When a safe space exists, the emergency lane change assist command limits both steering range and steering angular velocity. Simultaneously with activating the emergency avoidance command, the safety decision microcontroller sends a bypass status notification to the central control unit for post-intervention status calibration and control revert. Specifically, it sends a bypass start signal, which the central control unit receives and suspends its own command output. After the emergency avoidance operation is completed and the vehicle stabilizes, the safety decision microcontroller sends a bypass end signal. Upon receiving this signal, the central control unit first performs status calibration using the vehicle's real-time pose and environmental perception results, and then resumes its control functions. When the vehicle leaves an area with poor communication, the onboard communication module continuously monitors the network signal strength. Once the signal strength exceeds a stable communication threshold (e.g., -90dBm) within a preset duration (e.g., 10 seconds), the central control unit resets its internal control mode flag to revert from local perception priority mode to navigation command driven mode.

[0022] Example 1: When a new energy vehicle is traveling at 90 km / h along a mountain highway and is about to enter a 4-kilometer tunnel marked as a weak communication signal area in the vehicle's onboard map data, the vehicle's central control unit calculates, based on the predicted communication quality and the current vehicle speed, that the control mode switching trigger point is 600 meters from the tunnel entrance. When the vehicle reaches this location, the system performs a pre-control switching operation, sets the internal control mode flag, and switches the vehicle control mode from navigation command-driven mode to local perception priority mode. After the vehicle enters the tunnel, the communication connection with the cloud service platform is interrupted, and the vehicle's control relies entirely on the fusion perception results of its multimodal sensor system. When the vehicle reaches the middle of the tunnel, it encounters high-concentration exhaust fumes from a heavy truck ahead, creating a localized low-visibility area. In this environment, the onboard high-definition... As the camera's output image clarity decreases and the LiDAR point cloud data becomes sparse, the system immediately performs a real-time assessment of the uncertainty in the multimodal perception data, calculating that the visual perception uncertainty and spatial perception uncertainty values ​​for the corresponding area have increased. Simultaneously, the millimeter-wave radar detects a stationary target with unclear radar cross-section characteristics ahead, whose classification confidence uncertainty value is also at a high level. Under these conditions, the pre-executed control pre-switching operation eliminates the need for the system to handle conflicts between expired navigation commands and local perception results, providing an interference-free decision-making basis for subsequent risk potential field construction. Instead of attempting to identify the specific category of the stationary target in the low-quality data, the system shifts its decision-making basis to the quantification of perception uncertainty. Within a vehicle-centric local 3D grid map, it determines the target's classification based on the high level of comprehensive uncertainty value in the area. The probability of the presence of non-zero obstacles detected by millimeter-wave radar and follow the risk potential calculation rules. This provides high-risk potential values ​​for relevant grid cells in the low-visibility area ahead.

[0023] Based on the constructed dynamic risk potential field, the vehicle's model predictive control algorithm, when evaluating the future trajectory, determines that the cumulative risk value of a straight-line trajectory passing through a high-risk potential energy area exceeds the safety threshold. It then selects a candidate trajectory with the minimum cumulative risk value that satisfies the vehicle's kinematic constraints. This trajectory corresponds to a controlled lane change to an adjacent lane accompanied by deceleration. The vehicle's brake and steering actuators respond according to this trajectory command, causing the vehicle to merge into the adjacent lane, thus bypassing a detached tire fragment obscured by exhaust fumes. After passing through the area, the vehicle's local perception system returns to normal operation and continues driving in this mode. When the vehicle exits the tunnel, the onboard communication module detects that the network signal strength exceeds the stable communication threshold and remains stable. The central control unit resets the internal control mode flag, and the vehicle returns to the navigation command drive mode, continuing to drive based on the resynchronized cloud information.

[0024] Example 2: This example uses a hardware-in-the-loop simulation test to quantitatively evaluate the response time of the new energy vehicle navigation and communication control method of the present invention under specific critical conditions. The test aims to compare the system using the present invention with the system using the traditional passive switching scheme, and the total response time required from the appearance of the obstacle to the system issuing an emergency avoidance command when communication interruption and sudden obstacle in front occur simultaneously.

[0025] The experiment employed a hardware-in-the-loop simulation platform, which included a physical central control unit running the control algorithm of this invention, and a real-time simulator capable of simulating vehicle dynamic sensor data flow and network communication environment. The simulator generated test scenarios based on a computational fluid dynamics model, with the functional specifications and performance boundaries of the simulated sensor data output set as follows: a high-definition camera frame rate of 30 frames / second, a lidar scanning frequency of 10 Hz, and a millimeter-wave radar data refresh cycle of 50 milliseconds. The test scenario was set as follows: a test vehicle travels at 100 km / h on a simulated highway towards a tunnel entrance, which is defined as a communication interruption point; upon the vehicle reaching… The moment of reaching the tunnel entrance is defined as t=0. A static obstacle appears 150 meters in front of the vehicle. For comparison, a control group and the sample group of the present invention are set up: the control logic of the control group is set to switch to the local perception priority mode only after a 500-millisecond confirmation delay after detecting a communication interruption; the control logic of the sample group of the present invention is configured according to the complete scheme of the present invention, including performing a control pre-switching operation 500 meters in front of the tunnel entrance, and making decisions based on the constructed risk potential field in the local perception priority mode. In the experiment, the starting point for measuring the total system response time is t=0, and the ending point is the moment when the central control unit issues an emergency avoidance command to the actuator.

[0026] During the experiment, the control group's system simultaneously received a communication interruption signal and obstacle data detected by the sensors at t=0. Its control logic first entered a 500-millisecond communication interruption confirmation period. Then, when dealing with the conflict between the delayed navigation information and the local perception information, an additional decision arbitration delay was generated. Finally, an emergency avoidance command was issued at t=1150 milliseconds. The system of the present invention, having completed the switch to the local perception priority mode before entering the tunnel, did not need to perform mode switching or information conflict arbitration when the obstacle appeared at t=0. Instead, it directly used the sensor data for risk potential field construction and path planning, and issued an emergency avoidance command at t=250 milliseconds. The specific experimental data are shown in Table 1.

[0027] Table 1: Comparison of response times under different control schemes; The data in Table 1 show that the total response time of the sample group of the present invention is less than that of the control group. This difference is due to the fact that the sample group of the present invention avoids the delay in communication interruption confirmation and the delay in decision arbitration caused by data source conflict by switching control in advance. The test results confirm that the method of the present invention can shorten the risk avoidance decision time of the vehicle under the critical condition of communication interruption and sudden danger ahead.

[0028] Example 3: This example combines Figures 1 to 2 Description of navigation and communication control methods and systems for new energy vehicles, such as Figure 1As shown, the process begins by determining whether the vehicle is driving normally in an area with good communication. When the communication quality prediction attribute of the onboard map indicates that the vehicle is about to enter an area with poor communication, the system performs a control pre-switching, switching from navigation command mode to local perception priority mode and suppressing the adoption of navigation commands. In this mode, the system quantifies the uncertainty of multimodal perception based on the raw perception data continuously output by the multimodal sensor system, and constructs a dynamic risk potential field in a local three-dimensional grid map by combining obstacle probability and collision impact level. Subsequently, the central control unit generates and issues control commands based on this potential field to plan an avoidance path. On the other hand, a system conforming to the ASIL-D standard for vehicle safety integrity is also implemented. The precise safety decision microcontroller operates independently of the main control unit, continuously calculating collision risk indicators, including the estimated time to collision (TTC) and the probability of collision, and determining whether these indicators exceed thresholds, such as whether the TTC is less than 2.0 seconds. If the TTC does not exceed the threshold, the central control unit continues to take control of the vehicle. If the threshold is exceeded, the safety decision microcontroller will bypass the main control unit to perform a reflexive emergency avoidance operation, directly issuing commands to the braking and steering actuators, and sending a bypass status notification for subsequent status calibration and control handover. In addition, the system continuously monitors whether it has driven out of the communication failure area. When the network signal is detected to be continuously and stably restored, the internal control mode flag is reset, allowing the vehicle to return to the navigation command drive mode.

[0029] like Figure 2 As shown in the figure, the horizontal axis represents the distance of the vehicle relative to the tunnel entrance, and the vertical axis represents the control mode status. The solid square marks represent the navigation mode, and the dashed triangle marks represent the local perception mode. As can be seen from the figure, in the range of 800 meters to 400 meters from the tunnel entrance, the vehicle is in a state where the navigation mode is active and the local perception mode is off. When the vehicle travels to 400 meters from the tunnel entrance, the system performs a forward switch and activates the local perception mode. When the vehicle arrives at the tunnel entrance, that is, at a relative distance of 0 meters, the navigation mode is turned off due to communication interruption, while the local perception mode remains active. This ensures that control is smoothly transferred to the local perception system before entering an area with poor signal.

[0030] Example 4: This example illustrates a standardized offline calibration procedure for determining key parameters in the risk potential energy calculation formula. This addresses the issue of ensuring the reproducibility of decision-making logic when the system is deployed to vehicle platforms with different physical characteristics. When the control system of this invention needs to be adapted to a new energy vehicle with specific braking performance and sensor layout, the weighting coefficients in its risk potential field calculation logic need to be adjusted. , , and collision impact level Calibration is performed; the procedure first defines the collision impact level. Calculations are performed, and the settings are designed to balance the severity of the collision consequences with the vehicle's own emergency braking capabilities; for any grid unit ,That Defined as an indicator related to the vehicle's current speed and shortest braking distance, it is calculated by calibrating the vehicle's maximum safe deceleration under a specific road surface adhesion coefficient through real-vehicle testing. It is 9.8 During system operation, the collision impact level Based on the vehicle's current speed and the distance between the grid and the vehicle The calculation is performed in real time and is based on the formula for the shortest braking distance. and will Set as This calculation result increases when the vehicle approaches an obstacle at high speed, thus providing an input related to the vehicle's physical limits for calculating the risk potential energy.

[0031] Based on this, the weighting coefficients , , Calibration is performed, a process that integrates the vehicle dynamics model with the aforementioned... The computational logic is executed on a ring-on-a-loop simulation platform. The calibration process introduces a test case set containing multiple scenarios, covering both clear obstacles with high uncertainty in sensor data and ambiguous obstacles, as well as scenarios with high uncertainty due to environmental factors but no real obstacles. The calibration objective is to find a set of optimal weight coefficients that minimizes the system's overall decision cost function across all test scenarios. Here, the cost function is defined as the penalty for avoidance response delay and erroneous avoidance actions. The specific calibration steps are: setting... The initial value is 1.0, and its sum is constrained to be a constant; subsequently, the gradient descent optimization algorithm is used to repeatedly run simulations on the test case set. After each iteration, the values ​​of the three weight coefficients are adjusted according to the output of the cost function until the cost function converges to the minimum value; in a specific implementation of this procedure, the system determines when , , At that time, the comprehensive cost function reaches its minimum point; by executing the above procedure, a set of weight coefficient combinations for this specific vehicle model is obtained. , , This set of parameters is embedded in the configuration file of the vehicle's central control unit, thus providing a reproducible method for determining parameters for the deployment of risk potential field decision-making logic on different vehicle platforms.

[0032] Example 5: To enable the central control unit to directly use the communication quality prediction attributes calibrated in the vehicle map data to calculate the trigger time for control mode switching, the map database must first be processed offline to generate a layer of poor communication areas with clearly defined boundaries. This processing uses a set of preset logical rules to convert the geographic attribute information in the map into area identifiers that the control system can recognize. The rules are as follows: tunnel sections with a length exceeding 500 meters, mountain road sections with a continuous slope exceeding 30 degrees on both sides, and urban canyon road sections with an average building height exceeding 50 meters and a road width of less than 20 meters are all marked as poor communication areas. After this process, the map data contains road section identifiers with definite start and end points, providing unambiguous geographic location input for subsequent control pre-switching operations.

[0033] To set collision risk thresholds for a safety decision-making microcontroller independent of the central control unit, a calibration procedure for emergency avoidance triggering conditions must be implemented to balance the timeliness of emergency response with the ability to suppress false triggering; this procedure aims to set the triggering threshold of the predicted time of collision (TTC). The settings are matched to the vehicle's own physical braking limits and the system's internal delays; its threshold consists of the following components: ,in The internal processing latency required for the microcontroller to receive raw sensor data and issue commands for safety decision-making was calibrated to 50 milliseconds through system testing. The time required for the vehicle braking system to build up full braking pressure after receiving a command was calibrated to 200 milliseconds through bench testing. A time safety margin, set according to vehicle safety requirements, is defined as 1.5 seconds. Substituting the above parameters into the calculation, the following can be obtained for this vehicle model. The time is 1.75 seconds. This value is stored in the non-volatile memory of the safety decision microcontroller as the basis for its judgment to bypass the central control unit and directly issue emergency avoidance commands.

[0034] Example 6: To quantify the image features output by a high-definition camera into visual perception uncertainty values ​​that can be used in a risk potential field, an offline calibration and online calculation procedure needs to be performed. This procedure first involves acquiring thousands of road images under clear and well-lit conditions, and then statistically determining the number of baseline effective feature points. With a baseline image entropy value During actual vehicle operation, the system calculates the number of valid feature points in the current frame in real time. With image entropy The uncertainty value of visual perception is calculated using a normalized weighting function. The function is The weighting coefficient and The values ​​were set to 0.7 and 0.3. To set objective trigger conditions for the system to recover from the local perception priority mode to the navigation command driven mode, the performance of the vehicle communication module needs to be calibrated to determine the stable communication threshold. This process involves controlling the network signal strength in a laboratory environment to find the lowest signal strength value at which the vehicle communication module can maintain a preset data throughput and a packet loss rate of less than 0.1%, which was determined to be -90dBm. At the same time, to avoid frequent switching of control modes due to instantaneous signal fluctuations, a duration of 10 seconds was set as the time threshold for judging whether communication has been stably restored by analyzing the fluctuation period of signal strength in the road test data. Only when the signal strength is continuously above -90dBm within 10 seconds will the central control unit reset the internal control mode flag.

[0035] To address potential failures of critical sensors in local perception-first mode, the system integrates an online fault-tolerance mechanism. This mechanism continuously monitors the data stream integrity and status flags of core sensors, including millimeter-wave radar and lidar. If the data stream of any core sensor is interrupted or continuously outputs an error status for more than a 200-millisecond time window, the system determines that the sensor has failed. Once a failure is determined, the system immediately executes a preset minimum risk strategy, which includes: gradually decelerating the vehicle at a controlled rate until it stops within the current lane, activating the vehicle's hazard warning lights, and issuing a takeover request to the driver.

[0036] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A navigation and communication control method for new energy vehicles, characterized in that, The method includes: Before the vehicle enters an area with poor communication, a control mode switching trigger time is calculated based on a communication quality prediction attribute marked in the vehicle map data and the vehicle's speed and position. At the calculated trigger time, the vehicle's central control unit performs a control pre-switching operation. This operation sets an internal control mode flag, so that subsequent vehicle control commands will preferentially adopt the fused perception results from a local multimodal sensor system and suppress the adoption of path commands from the navigation system. After performing the control pre-switching operation, the quality of the data output by each sensor mode in the multimodal sensor system is evaluated in real time to generate a quantized uncertainty value for each mode. Specifically: for high-definition cameras, a visual perception uncertainty value is generated by analyzing the pixel grayscale distribution and the number of effective feature points in the output image; for LiDAR, a spatial perception uncertainty value is generated by evaluating the sparsity of the output point cloud data and the completeness of the target contour description; for millimeter-wave radar, a classification confidence uncertainty value is generated by analyzing the radar cross-section characteristics of its echo signal and the degree of matching with the object model. Furthermore, in a local 3D grid map centered on the vehicle, a risk potential energy is calculated for each grid cell in the local 3D grid map based on the quantized uncertainty value, the probability of obstacle presence within the grid, and the collision impact level; wherein, each grid cell in the local 3D grid map ( Risk potential energy The calculation follows these rules: ,in, For grid cells ( Within ), it is a comprehensive uncertainty value obtained by weighting and fusing the visual perception uncertainty value, the spatial perception uncertainty value, and the classification confidence uncertainty value according to their respective weights; For grid cells ( Within a given area, the probability of an obstacle's existence is output after detection by sensors and processing by a target confirmation algorithm. The calculation based on the vehicle's current speed and braking capacity is the result if the vehicle enters the grid cell. A collision impact level; , and The weighting coefficients are determined by configuration parameters loaded during system initialization to balance the importance of the three factors; Based on the constructed risk potential field, a trajectory that minimizes the risk potential energy on the driving path while satisfying the vehicle's own kinematic constraints is planned and selected. Furthermore, a safety decision microcontroller, independent of the central control unit, continuously receives raw sensor data via an independent bus and calculates a collision risk index. When the calculated collision risk index exceeds the threshold, it bypasses the central control unit and directly issues emergency avoidance commands to the vehicle's brake actuators and steering actuators, and sends a bypass status notification to the central control unit for post-intervention status calibration and control revert.

2. The navigation and communication control method for new energy vehicles according to claim 1, characterized in that, The safety decision microcontroller, independent of the central control unit, is a microcontroller that conforms to the Automotive Safety Integrity Level (ASIL-D) standard; the independent bus is a dedicated controller area network (CLAN) safety bus; the collision risk index calculation integrates a predicted time to collision (TTC) and a collision probability directly calculated from raw sensor data; the emergency avoidance commands are specifically an emergency braking command with maximum safe deceleration and an emergency lane change assist command that limits both steering range and steering angular velocity when a safe space is confirmed in the adjacent lane.

3. The navigation and communication control method for new energy vehicles according to claim 1, characterized in that, The method also includes: after the vehicle leaves an area with poor communication, when the network signal strength detected by an on-board communication module exceeds a stable communication threshold for a period of time, the central control unit resets the internal control mode flag to restore the navigation command driven mode from the local perception priority mode.

4. The navigation and communication control method for new energy vehicles according to claim 1, characterized in that, A bypass status notification is sent to the central control unit for post-intervention status calibration and control handover. Specifically, when the safety decision microcontroller activates the emergency avoidance command, it sends a bypass start signal to the central control unit via an independent bus. Upon receiving the bypass start signal, the central control unit suspends its own path planning and control command output. After the emergency avoidance operation is completed and the vehicle stabilizes, the safety decision microcontroller sends a bypass end signal. Upon receiving the bypass end signal, the central control unit first performs a status calibration using the vehicle's real-time pose and the latest environmental perception results, and then resumes its control functions.

5. The navigation and communication control method for new energy vehicles according to claim 1, characterized in that, Overall uncertainty value The calculation also includes adaptively adjusting the weights of the visual perception uncertainty, spatial perception uncertainty, and classification confidence uncertainty used for weighted fusion based on the type of environment in which the vehicle is currently located.

6. The navigation and communication control method for new energy vehicles according to claim 1, characterized in that, Planning and selecting a trajectory with the lowest risk potential energy along the driving path is achieved through a model predictive control algorithm. In each decision cycle, the algorithm generates multiple candidate trajectories that meet the constraints of the vehicle's maximum steering angle and maximum acceleration / deceleration. It also calculates the cumulative risk potential energy of all grid cells that each candidate trajectory will pass through within a prediction time window. Finally, the candidate trajectory with the lowest cumulative risk potential energy is selected as the execution trajectory. The trigger time for switching control modes is calculated as when the vehicle is 300 to 800 meters away from the boundary of the poor communication zone entrance. Furthermore, the condition for the collision risk index to exceed the threshold is that the calculated predicted collision time (TTC) value is less than 2.0 seconds.

7. A navigation and communication control system for new energy vehicles, characterized in that, The system includes: A multimodal sensor system; A brake actuator; A steering actuator; A single independent bus; A central control unit connected to the multimodal sensor system; A safety decision-making microcontroller is connected to a multimodal sensor system via an independent bus, and also to the brake actuator, steering actuator, and central control unit. The central control unit is configured to: before the vehicle enters a communication-poor area, calculate a control mode switching trigger time based on a predicted communication quality attribute calibrated in the onboard map data and the vehicle's speed and position; and at the calculated trigger time, perform a control pre-switching operation to set an internal control mode flag. This flag ensures that subsequent vehicle control commands prioritize the fused perception results from the multimodal sensor system and suppress the adoption of path commands from the navigation system. After performing the control pre-switching operation, it continuously evaluates the performance of each sensor mode in the multimodal sensor system. The quality of the output data is evaluated to generate a quantified uncertainty value for each of the following: For high-definition cameras, a visual perception uncertainty value is generated by analyzing the pixel grayscale distribution and the number of effective feature points in the output image; for LiDAR, a spatial perception uncertainty value is generated by evaluating the sparsity of the output point cloud data and the completeness of the target contour description; for millimeter-wave radar, a classification confidence uncertainty value is generated by analyzing the radar cross-section characteristics of its echo signal and the degree of matching with the object model. Furthermore, in a vehicle-centric local 3D grid map, based on the quantified uncertainty value and the probability of obstacle presence and collision impact level within the grid, a risk potential energy is calculated for each grid cell in the local 3D grid map; where each grid cell in the local 3D grid map ( Risk potential energy The calculation follows these rules: ,in, For grid cells ( Within ), it is a comprehensive uncertainty value obtained by weighting and fusing the visual perception uncertainty value, the spatial perception uncertainty value, and the classification confidence uncertainty value according to their respective weights; For grid cells ( Within a given area, the probability of an obstacle's existence is output after detection by sensors and processing by a target confirmation algorithm. The calculation based on the vehicle's current speed and braking capacity is the result if the vehicle enters the grid cell. A collision impact level; , and The weighting coefficients are determined by the configuration parameters loaded during system initialization to balance the importance of the three factors; based on the constructed risk potential field, a trajectory that minimizes the risk potential energy on the driving path while satisfying the vehicle's own kinematic constraints is planned and selected; the safety decision microcontroller is configured to: operate independently of the central control unit, continuously receive raw sensor data from the multimodal sensor system and calculate a collision risk index; and when the calculated collision risk index exceeds the threshold, bypass the central control unit, directly issue emergency avoidance commands to the brake actuators and steering actuators, and send a bypass status notification to the central control unit for post-intervention status calibration and control handover.

Images