Automatic driving navigation control method based on fixed calibration object guidance

Abstract

The invention relates to the field of automatic driving, and discloses an automatic driving navigation control method based on fixed calibration object guidance, which comprises the following steps of: detecting and decoding one or more fixed calibration objects preset in an environment through a vehicle-mounted sensor to obtain function types, positions and basic potential field parameters of the fixed calibration objects; acquiring a real-time state vector of the vehicle; based on the vehicle state vector, performing adaptive correction on the basic potential field parameter, and generating a composite adaptive virtual potential field according to a function type combination decoded from a calibration object; and superposing the virtual potential field to a basic navigation cost map of the vehicle to guide a path planning module to generate a driving path capable of avoiding temporary obstacles or following temporary guidance. According to the method, a map does not need to be built in advance, flexible, safe and real-time local path guidance of the automatic driving vehicle can be realized through a low-cost calibration object, and the adaptability and reliability of the system in temporary traffic scenes such as construction and accidents are enhanced.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to an autonomous driving navigation control method based on fixed calibration objects. Background Technology

[0002] Currently, the path planning and navigation of autonomous vehicles largely rely on pre-collected and produced high-precision maps. High-precision maps provide rich static environmental elements such as road geometry, topological relationships, and traffic signs with centimeter-level accuracy, providing crucial basic data support for vehicle localization and global path planning.

[0003] However, this technological approach, which heavily relies on high-precision maps, increasingly reveals its inherent limitations when facing dynamic and ever-changing real-world traffic environments. Road conditions in the real world are not static; unexpected events such as temporary road construction, lane closures due to traffic accidents, and temporary detours guided by traffic control frequently occur. For these highly dynamic scenarios, the production and updating of high-precision maps are costly, and their update cycle lags far behind the rate of change in the real world, leading to inconsistencies between map information and actual road conditions.

[0004] An autonomous vehicle that relies entirely on map data will be unable to correctly understand the current scenario when encountering temporary changes not marked on the map, potentially making incorrect decisions, such as attempting to enter a closed construction area or failing to recognize newly opened temporary guidance lanes. This not only severely limits the vehicle's applicability and operational efficiency but also poses significant safety hazards. Therefore, developing a technical solution that can supplement high-precision maps in real time and at low cost, enabling autonomous vehicles to flexibly and safely respond to localized temporary road condition changes, has become an urgent technical challenge to be solved in this field. Summary of the Invention

[0005] The technical problem this invention aims to solve is that existing autonomous driving technologies lack effective guidance mechanisms when faced with temporary road closures or obstacles not covered by high-precision maps, which may lead to vehicle interruption or safety risks.

[0006] To solve the above-mentioned technical problems, the present invention is implemented through the following technical solution: The first aspect of this invention provides an autonomous driving navigation control method based on fixed calibration objects, which is applied to autonomous vehicles and includes the following steps: Step S100: Detect at least one fixed marker in the driving environment of the autonomous vehicle.

[0007] Step S200: Decode the at least one fixed calibrator to obtain the potential field parameters used to define the virtual potential field.

[0008] Step S300: Based on the potential field parameters, generate the virtual potential field on the vehicle's navigation cost map.

[0009] Step S400: Based on the navigation cost map overlaid with the virtual potential field, replan the local driving path for the autonomous vehicle.

[0010] In an optional implementation, step S300, which involves generating the virtual potential field based on the potential field parameters, further includes: Real-time acquisition of a set of vehicle statuses of the autonomous vehicle; The virtual potential field is generated based on the vehicle's own state and the potential field parameters.

[0011] In a further embodiment, the process of generating the virtual potential field based on the vehicle's own state and the potential field parameters specifically involves: Extract the fundamental strength and fundamental influence range that define the potential field shape from the potential field parameters; The basic strength and / or basic influence range are corrected using the vehicle's own state to obtain the final potential field strength and / or final potential field influence range, and the virtual potential field is generated accordingly.

[0012] Specifically, this correction process can be characterized by the following formula: A final =A base ·f A (S); σ final =σ base ·f σ (S); Among them, A base and σ base These represent the basic strength and basic influence range decoded from the fixed calibration object, respectively; S is the state vector composed of the vehicle's own state; f A (S) and f σ (S) is a preset nonlinear correction function that takes the state vector S as input; A final and σ final The final strength and final influence range used to generate the virtual potential field after modification.

[0013] In a further embodiment, the vehicle's own state includes the vehicle's current speed and / or vehicle category. The process of correcting using the vehicle's own state specifically involves: The correction function f σ The output value of (S) increases as the current speed of the vehicle increases; Furthermore, the geometry of the virtual potential field is selected or adjusted according to the vehicle category. For example, when the vehicle category is a large vehicle, the virtual potential field is determined to be elliptical, while when the vehicle category is a small vehicle, it is determined to be circular.

[0014] In another optional implementation, when multiple fixed calibration objects are detected, the step of generating the virtual potential field in step S300 specifically involves: Based on multiple sets of potential field parameters decoded from the multiple fixed calibration objects, and according to a preset composite rule, a composite virtual potential field is generated as the virtual potential field.

[0015] In one specific embodiment, the preset composite rule includes a guiding channel rule. This rule is executed when the system detects that the plurality of fixed calibration objects include both repulsive and attractive calibration objects.

[0016] When executing this rule, the system will arithmetically superimpose the repulsive potential field generated by the repulsive type calibration object and the gravitational potential field generated by the gravitational type calibration object. Since the strength value of the gravitational potential field is negative, the superposition will form a composite virtual potential field with a low-cost valley region and guiding function.

[0017] In another specific embodiment, the preset composite rule includes a conditional activation rule. This rule is executed when the system detects that a condition-triggered calibrator is included among the plurality of fixed calibrators.

[0018] When executing this rule, the system continuously determines whether the preset activation conditions are met based on the real-time sensor data of the autonomous vehicle. The system only generates the virtual potential field corresponding to the condition-triggered calibration object when the determination result is that the preset activation conditions are met; otherwise, the virtual potential field is not generated.

[0019] In an alternative implementation, the method further includes a step after path replanning is completed: removing the previously generated virtual potential field from the navigation cost map after detecting the deregistration object or the sensors of the autonomous vehicle can no longer detect the fixed calibration object.

[0020] In an optional implementation, step S400, which involves replanning the local driving path based on the navigation cost map overlaid with the virtual potential field, specifically comprises: The virtual potential field ΔC(p) generated in step S300 is compared with the vehicle's inherent basic navigation cost map C. base (p) Perform arithmetic superposition to form the total cost map C. total (p). Their superposition relationship can be expressed by the following formula: C total (p)=Cbase (p)+ΔC(p); Where p is any point on the map.

[0021] Then, a path planning algorithm is used on the total cost map C. total (p) performs a path search to obtain the path with the lowest cost as the local driving path.

[0022] A second aspect of the present invention provides an autonomous driving navigation control system based on fixed calibration objects, the system comprising: A perception decoding module is configured to detect at least one fixed marker in the driving environment of an autonomous vehicle and decode the fixed marker to obtain potential field parameters. A potential field generation module, which is connected to the perception decoding module, is configured to receive the potential field parameters and generate a virtual potential field on the vehicle's navigation cost map based on the potential field parameters. The path planning module, which is connected to the potential field generation module, is configured to replan a local driving path for the autonomous vehicle based on a navigation cost map superimposed with the virtual potential field.

[0023] In summary, the present invention has at least one of the following beneficial technical effects: 1. This invention guides vehicles by generating a virtual potential field on a navigation cost map, rather than providing a fixed detour path, thereby improving the flexibility and robustness of navigation. When a new temporary obstacle appears within the detour area, the path planner, based on a total cost map that integrates the virtual potential field and real-time environmental perception, can autonomously avoid the new obstacle and continue the detour, thus avoiding the driving interruption problem caused by relying on a preset rigid path.

[0024] 2. This invention improves the safety of path planning and its applicability to different vehicle types by adaptively correcting the shape of the virtual potential field by incorporating the vehicle's own state. For example, the system can adjust the range of influence of the potential field according to the vehicle's current speed, or adjust the geometry of the potential field according to the vehicle category, ensuring that the path planned for large, heavy-duty vehicles has a sufficiently large turning radius and safety margin, while also allowing smaller vehicles to take more efficient and compact paths, thus achieving unified and safe guidance for heterogeneous vehicle fleets.

[0025] 3. This invention achieves complex guidance and refined management of vehicle behavior by designing composite rules based on multiple fixed calibrators. For example, by combining repulsive and attractive calibrators, a directional guidance channel can be formed, actively guiding the vehicle to travel along a specific safety corridor, rather than simply passively avoiding obstacles. By setting condition-triggered calibrators, intelligent management of dynamic scenarios such as intersection conflicts can be achieved, significantly improving the autonomous decision-making ability and operational efficiency of the autonomous driving system in complex and temporary traffic scenarios. Attached Figure Description

[0026] Figure 1 This is a structural block diagram of an embodiment of an autonomous driving navigation control system of the present invention; Figure 2 This is a flowchart of an embodiment of an autonomous driving navigation control method of the present invention; Figure 3 This is a schematic diagram illustrating the working relationship between the sensing decoding module and the fixed calibration object according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the internal logic of a composite adaptive virtual potential field generation module according to an embodiment of the present invention; Figure 5 This is a schematic diagram illustrating an embodiment of the present invention applied to a temporary maintenance scenario on a single-sided road in a mining area.

[0027] Among them, 10 is the perception decoding module; 20 is the vehicle state acquisition module; 30 is the composite adaptive virtual potential field generation module; 31 is the adaptive correction unit; 32 is the composite generation unit; 40 is the path planning and control module; 50 is the fixed calibration object; 50a is the first repulsive type fixed calibration object; 50b is the second repulsive type fixed calibration object; 50c is the gravitational type fixed calibration object; 60 is the two-way road; 61 is the temporary maintenance lane; 62 is the temporarily opened dirt road; and 70 is the large autonomous driving mining truck. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0029] See attached document Figure 1 , Figure 1 This is a structural block diagram of an autonomous driving navigation control system according to an embodiment of the present invention. The system can be integrated into an autonomous vehicle and includes: a perception decoding module 10, a vehicle state acquisition module 20, a composite adaptive virtual potential field generation module 30, and a path planning and control module 40.

[0030] The perception decoding module 10 and the vehicle state acquisition module 20 are both electrically or communicatively connected to the composite adaptive virtual potential field generation module 30. The composite adaptive virtual potential field generation module 30 is electrically or communicatively connected to the path planning and control module 40.

[0031] See attached document Figure 2 , Figure 2 This is a flowchart of an autonomous driving navigation control method according to an embodiment of the present invention. The navigation control method provided by the present invention may include the following steps: In the initial stage, autonomous vehicles rely on their internally stored basic navigation cost map C base (p) The system travels along a global path, with the basic navigation cost map representing a fixed road network and static obstacle information. During travel, the system performs the following steps: In step S100, the perception decoding module 10 continuously detects the driving environment through the vehicle-mounted sensors.

[0032] When at least one preset fixed calibration object is detected, the module identifies and locates it, and decodes the information stored therein to obtain a set of potential field parameters I for defining the virtual potential field.

[0033] The potential field parameter I includes at least: the unique identifier i of the calibration object. d Function type (type), position (p) of the calibrator in the world coordinate system. i and a set of fundamental potential field parameters θ base , where θ base This also includes the basic strength A. base and the basic influence range σ base .

[0034] The obtained potential field parameter I is transmitted to the composite adaptive virtual potential field generation module 30.

[0035] Step S200: At the same time, the vehicle status acquisition module 20 collects a set of vehicle statuses representing the current operating conditions of the vehicle in real time via the vehicle data bus and constructs them into a state vector S.

[0036] The state vector S includes: the vehicle's current speed v, current acceleration a, a vehicle category identifier M representing the vehicle's physical properties, and a task priority identifier T representing the current driving task attributes. The obtained state vector S is transmitted to the composite adaptive virtual potential field generation module 30.

[0037] In step S300, after receiving the potential field parameter I and the state vector S, the composite adaptive virtual potential field generation module 30 executes its internal generation logic.

[0038] This logic is first based on the state vector S and the fundamental intensity A in the potential field parameter I. base and the basic influence range σ base Adaptive correction is performed to obtain a set of final potential field parameters that match the current state of the vehicle.

[0039] Subsequently, if there are multiple fixed calibration objects, the multiple modified independent potential fields are combined according to the preset composite rules.

[0040] Finally, the module generates a complete composite adaptive virtual potential field ΔC(p) with a specific cost value at spatial point p. This generated virtual potential field ΔC(p) is then transmitted to the path planning and control module 40.

[0041] In step S400, after receiving the composite adaptive virtual potential field ΔC(p), the path planning and control module 40 compares it with the basic navigation cost map C stored inside the vehicle. base (p) Perform arithmetic superposition to generate a temporary total cost map C. total (p). This superposition relationship can be characterized by the following formula: C total (p)=C base (p)+ΔC(p); Where p is any coordinate point on the cost map.

[0042] In step S500, the path planning and control module 40 then, within its internal path planner, calculates the path based on the updated total cost map C. total (p) Perform path search calculations to find a local driving path with the lowest cost that starts from the vehicle's current position and avoids the virtual potential field area.

[0043] In step S600, the path planning and control module 40 sends the newly generated local driving path to the vehicle's underlying actuators to control the vehicle's steering, driving, and braking so that it can travel along the path.

[0044] When the perception decoding module 10 detects the removed calibration object or can no longer detect the aforementioned fixed calibration object, the composite adaptive virtual potential field generation module 30 stops generating the corresponding virtual potential field ΔC(p), thus preventing the total cost map C from being generated. total (p) Restore to the basic navigation cost map C base (p) The vehicle resumes normal navigation.

[0045] See attached document Figure 3 , Figure 3This is a schematic diagram illustrating the working relationship between the perception decoding module 10 and the fixed calibration object 50 according to an embodiment of the present invention. In a specific embodiment of the present invention, the perception decoding module 10 is used to detect, locate, and extract information from the fixed calibration object 50 deployed in the environment.

[0046] Fixed calibration object 50 is a physical entity that carries structured information and is pre-deployed in the driving environment.

[0047] In one specific embodiment, the fixed calibration object 50 may be a plate-like structure with a specific two-dimensional coding pattern. This two-dimensional coding pattern may adopt forms such as QR codes or April Tags, which have high information density and decoding robustness. The surface material of the plate-like structure is selected from materials with weather resistance and low reflectivity to ensure the detection stability of the vehicle-mounted sensor under different lighting and weather conditions.

[0048] The perception decoding module 10 includes an image acquisition unit and a lidar unit. During operation, the image acquisition unit (such as an onboard camera) acquires scene images containing a fixed calibration object 50. Simultaneously, the lidar unit acquires point cloud data of the same scene.

[0049] The perception decoding module 10 internally executes the detection process. First, it processes the image data acquired by the image acquisition unit, using an image recognition algorithm to locate the pixel region of the two-dimensional coded pattern in the image. After successful location, it calls the decoding program to decode the pixel region, thereby extracting a string of raw data.

[0050] Subsequently, the perception decoding module 10 parses the original data string and converts it into a structured potential field parameter I. The data structure of the potential field parameter I is consistent with that described in step S100 above, and includes the calibrator unique identifier i. d Function type and basic potential field parameter θ base Fields such as...

[0051] To determine the precise position p of the fixed calibration object 50 in the world coordinate system i The perception decoding module 10 also performs a data fusion process. This process associates the point cloud data acquired by the lidar unit with the recognition results of the image acquisition unit.

[0052] Specifically, based on the known geometric dimensions of the fixed calibration object 50 and its position in the image, combined with sensor calibration parameters, a point cloud cluster corresponding to the fixed calibration object 50 is matched in the point cloud data. By calculating the geometric center or the center of the fitted plane of the point cloud cluster, the three-dimensional coordinates of the fixed calibration object 50 in a high-precision world coordinate system are obtained, and its two-dimensional projection is used as its position p. i .

[0053] Ultimately, the perception decoding module 10 will contain the precise location p i The complete structured potential field parameters I are output to the composite adaptive virtual potential field generation module 30.

[0054] The vehicle status acquisition module 20 works in parallel with the vehicle's internal data network. Its function is to acquire and integrate a set of status information that can accurately characterize the current dynamics and attributes of the autonomous vehicle in real time.

[0055] In a specific embodiment of the present invention, the vehicle state acquisition module 20 reads data packets from the controller area network (CAN) bus or vehicle control unit (VCU) of the autonomous vehicle at a preset sampling frequency (e.g., 10 times per second or higher). The module parses the read data packets, extracts key physical quantities, and constructs them into a structured vehicle state vector S.

[0056] The vehicle's own state vector S is the core data input for the subsequent potential field adaptive correction step, and its specific structure is as follows: S = {v, a, M, T}; Here, v and a are scalar values ​​representing the vehicle's current speed and current acceleration, respectively. These two values ​​are obtained directly from the messages of the vehicle's powertrain or chassis control system, reflecting the vehicle's most direct kinematic characteristics.

[0057] M is the vehicle category identifier. This identifier is a pre-configured, discrete value stored in the system's non-volatile memory, used to distinguish vehicles with different physical attributes. For example, M might be set to a value of 1 for a large mining truck, and a value of 2 for a light pickup truck or maintenance vehicle. This identifier is indirectly related to the vehicle's inherent physical constraints such as minimum turning radius, mass, and geometric dimensions.

[0058] T is the task priority identifier. This identifier is issued by the vehicle's task management or dispatching system and reflects the nature or urgency of the current transportation task. For example, for a routine material transportation task, T can be set to a value of 1; for an urgent equipment repair or rescue task, T can be set to a value of 2.

[0059] In each calculation cycle, the vehicle state acquisition module 20 outputs the vehicle's own state vector S, which integrates the latest values, to the composite adaptive virtual potential field generation module 30 as the data basis for its subsequent calculations.

[0060] See attached document Figure 4 , Figure 4This is a schematic diagram of the internal logic of a composite adaptive virtual potential field generation module 30 according to an embodiment of the present invention. The operation process of this composite adaptive virtual potential field generation module 30 is essentially an operation on the basic navigation cost map C. base (p) Perform temporary, dynamic data modifications. Basic navigation cost map C base (p) is a one-dimensional or three-dimensional data raster, where each raster point p stores a value representing the passage cost.

[0061] The virtual potential field ΔC(p) generated by this invention is another data layer with the same dimension and resolution as the basic navigation cost map. The specific process of generating the virtual potential field ΔC(p) involves calculating the cost increment value that should be superimposed at each grid point p within the affected area. A repulsive virtual potential field corresponds to a cost increment region consisting of positive values, while a gravitational virtual potential field corresponds to a cost increment region consisting of negative values.

[0062] The composite adaptive virtual potential field generation module 30 receives the potential field parameter I output by the sensing and decoding module 10. i After obtaining the vehicle's own state vector S output by the vehicle state acquisition module 20, the potential field parameters are first processed by the internal adaptive correction unit 31.

[0063] The function of the adaptive correction unit 31 is to convert the universal basic potential field parameters obtained from the fixed calibration object into a set of personalized final potential field parameters that match the specific state of the current vehicle, and to calculate the data of the independent virtual potential field based on these parameters.

[0064] Specifically, for each detected fixed calibrator m i The adaptive correction unit 31 extracts its basic potential field parameter θ. base,i That is, the basic strength A base,i and the basic influence range σ base,i Subsequently, a set of preset adaptive correction functions f are applied. A and f σ By combining the vehicle's own state vector S with these two basic parameters through mathematical operations, the final potential field strength A is obtained. final,i and the final potential field influence range σ final,i The correction process is defined by the following formula: A final,i =A base,i ·f A (S); σ final,i =σ base,i ·f σ (S); Where i is the index of the calibrator; Abase,i and σ base,i f represents the original parameters decoded from the ii-th calibration object; S represents the vehicle's own state vector at the current time; f A and f σ It is a function with S as its independent variable that is pre-defined and stored in the system, and its output is a dimensionless correction coefficient.

[0065] In one specific implementation, the correction function f σ It can be designed as a linear function related to the vehicle's current speed v, for example: f σ (S)=k v ·v+c, where k v c are preset constant coefficients. This design ensures that the faster the vehicle speed, the larger the calculated final influence range σ. final,i The larger the value, the wider the virtual potential field will cover on the cost map, thus providing a greater safety margin.

[0066] In another specific implementation, the adaptive correction unit 31 also determines the geometric model of the virtual potential field based on the vehicle category identifier M. For example, when the value of M corresponds to a large truck, the system selects an elliptical Gaussian model to generate the potential field, with the major axis of the ellipse aligned with the vehicle's current direction of travel to better characterize the lateral space required for turning. When the value of M corresponds to a light vehicle, a standard circular Gaussian model is selected.

[0067] After adaptive correction, the adaptive correction unit 31 uses the final potential field parameters to perform adaptive correction on each independent calibration object m. i Calculate the cost value P of the independent virtual potential field at the spatial point p = (x, y). i (p). This calculation process can be characterized by a two-dimensional Gaussian function: Among them, ||pp i || represents the position p of map point p relative to the i-th marker p. i The Euclidean distance between them. This formula is used to calculate the cost increment for each grid point within the affected area, these calculated independent virtual potential fields P i (p) will be used as the input to the next processing unit.

[0068] See attached document Figure 4 The composite adaptive virtual potential field generation module 30 calculates an independent virtual potential field cost value P for each calibration object through the adaptive correction unit 31. i (p) After that, the composite generation unit 32 contained within it processes these independent virtual potential fields.

[0069] When the sensing and decoding module 10 detects multiple fixed calibration objects simultaneously, the composite generation unit 32, according to a set of preset composite rules, converts multiple independent virtual potential fields P i (p) combine to form a single, functionally guided composite adaptive virtual potential field ΔC(p).

[0070] The preset composite rules are stored in the system, and the mathematical operation to be performed is determined based on the combination of function types decoded from each fixed calibrator.

[0071] In one specific implementation, the preset composite rule includes a repulsive superposition rule. This rule is executed when the system detects multiple fixed calibrators with a repulsive function.

[0072] At this point, the composite generation unit 32 will generate the independent virtual potential fields P corresponding to each of these repulsive calibrators. i (p) is summed point-by-point to generate the final composite virtual potential field ΔC(p). The computation process can be represented by the following equation: ΔC(p)=∑ i∈RepulsiveSet P i (p); Here, RepulsiveSet is the set of indices of all currently detected repulsive markers. This rule can be used to enclose an irregularly shaped or larger, uniform, high-cost restricted area by deploying several independent markers.

[0073] In another specific implementation, the preset composite rule includes a guiding channel rule. This rule is executed when the system detects at least one calibrator of repulsive type and one calibrator of attractive type.

[0074] The basic strength A of the gravitational calibration object base It is set to a negative value. When this rule is executed, the composite generation unit 32 performs an arithmetic superposition of the repulsive potential field generated by the repulsive type calibration object and the gravitational potential field generated by the gravitational type calibration object. Since the cost value of the gravitational potential field is negative, its effect is to generate a strip-shaped region with a lower cost value in the high-cost region composed of the repulsive potential field, thereby forming a low-cost channel with a clear guiding direction.

[0075] In another specific implementation, the preset composite rule includes a conditional activation rule. This rule activates when the system detects a fixed calibrator m with a function type of conditional triggering. k It is executed at that time.

[0076] The independent virtual potential field P corresponding to the calibration object k(p) It does not take effect immediately; rather, its generation depends on external conditions. The composite generation unit 32 continuously acquires real-time environmental perception data S from other vehicle sensors (such as forward-facing millimeter-wave radar, cameras, etc.). sensor And based on this data, evaluate a pre-defined Boolean condition function Cond(S) sensor The truth value of ) is determined. The final generated virtual potential field is determined by the following formula: ΔC(p)=P k (p)·H(Cond(S sensor )); Here, H(·) is the Heaviside step function, which has a value of 1 when its independent variable is true and a value of 0 when it is false. This rule allows the generation of the virtual potential field to be linked with the dynamically changing environment in real time. For example, a temporary repulsive field is generated at the exit of the secondary road only when there is oncoming traffic on the main road, in order to prevent vehicles from merging in.

[0077] Finally, the composite generation unit 32 outputs the data of the composite adaptive virtual potential field ΔC(p) calculated according to the rules to the path planning and control module 40.

[0078] After receiving the complete composite adaptive virtual potential field ΔC(p) output by the composite adaptive virtual potential field generation module 30, the path planning and control module 40 executes the subsequent path replanning and vehicle control process.

[0079] The path planning and control module 40 internally maintains and uses the basic navigation cost map C. base (p). The basic navigation cost map is pre-built and represents permanent or semi-permanent environmental information such as road boundaries and static obstacles.

[0080] Upon receiving the virtual potential field ΔC(p), the path planning and control module 40 fuses it with the basic navigation cost map to generate a temporary total cost map C reflecting the current guidance information. total (p). This fusion process is a point-by-point arithmetic summation operation: C total (p)=C base (p)+ΔC(p); Through this calculation, a virtual high-cost region (repulsive field) or low-cost region (gravitational field) defined by a fixed calibration object is superimposed onto the vehicle's navigation space.

[0081] Subsequently, the path search unit within the path planning and control module 40 employs existing industry-standard path planning algorithms, such as the Hybrid A* algorithm or the D*Lite algorithm. This is used in conjunction with the updated total cost map C. totalUsing p as the sole basis for its calculation, starting from the vehicle's current position and attitude, it searches for a continuous path that minimizes the total travel cost to reach the local target point. Since the virtual potential field ΔC(p) has already changed the cost value of the local region, the algorithm's cost function evaluation results will naturally guide it to avoid high-cost regions and tend towards low-cost regions during the search process.

[0082] The calculated lowest-cost path is determined as the new local driving path. This path is transmitted to the control command generation unit of the path planning and control module 40. This unit is responsible for converting the path into specific, time-sequential vehicle control commands (e.g., target speed and steering angle at specific times) and issuing these commands to the vehicle's underlying actuators (such as the steering system, drive system, and braking system) to control the vehicle to travel along the newly planned path.

[0083] During the execution of this method, when the perception decoding module 10 can no longer detect the fixed marker that previously triggered the generation of the virtual potential field (e.g., the vehicle has passed through the area), or detects a specific marker with a cancellation function, this information is transmitted to the composite adaptive virtual potential field generation module 30. The composite adaptive virtual potential field generation module 30 then stops generating the corresponding virtual potential field ΔC(p), or outputs a virtual potential field where the cost value of all points is zero. Therefore, the total cost map C... total (p) Restore to the base navigation cost map C base (p) In a consistent state, the vehicle's path planning behavior then reverts to the normal mode.

[0084] To further illustrate the present invention, a specific embodiment will be described below.

[0085] See attached document Figure 5 , Figure 5 This is a schematic diagram of a temporary maintenance scenario for a single-sided road in a mining area, according to an embodiment of the present invention. In this scenario, the temporary maintenance lane 61 of a two-way road 60 is closed due to temporary maintenance, and vehicles need to detour to a temporarily opened dirt road 62.

[0086] In this embodiment, the scene is arranged as follows: Two repulsive fixed markers, a first repulsive fixed marker 50a and a second repulsive fixed marker 50b, are placed at the entrance of the closed lane 61, with their lateral spacing designed to form a continuous virtual barrier. A gravitational fixed marker 50c, with a functional type of attraction, is placed at the entrance of the temporary bypass dirt road 62.

[0087] A large, autonomous mining truck 70 (its vehicle category identifier M is preset to "large vehicle") travels along road 60 toward the maintenance area. Its perception decoding module 10 sequentially detects fixed markers 50a, 50b, and 50c, and successfully decodes their respective function types, world coordinate system positions, and fundamental potential field parameters (A). base and σ base Meanwhile, the vehicle status acquisition module 20 continuously acquires the current speed v of the mining vehicle.

[0088] The data is transmitted to the composite adaptive virtual potential field generation module 30. The adaptive correction unit 31 of this module first processes the parameters of the three calibration objects. Since the vehicle category M is "large vehicle," the system selects an elliptic Gaussian model to generate the potential field. Simultaneously, based on the current speed v, the system uses a correction function f... σ (S) Calculate a larger final impact range σ final This is to ensure that the large vehicle is given sufficiently smooth steering guidance.

[0089] Subsequently, the composite generation unit 32 of this module identifies the simultaneous presence of repulsive and attractive calibrators in the current scene, thus triggering the execution of the guidance channel rules. This unit calculates two independent repulsive potential fields with positive cost values ​​generated by calibrators 50a and 50b, and one independent attractive potential field with a negative cost value generated by calibrator 50c. Then, these three independent potential fields are superimposed through point-by-point arithmetic summation to generate a single composite adaptive virtual potential field ΔC(p).

[0090] The generated ΔC(p) creates a continuous high-cost zone at the entrance of the closed lane 61, while creating a low-cost attraction zone at the entrance of the temporary bypass dirt road 62.

[0091] The path planning and control module 40 adds ΔC(p) to the basic navigation cost map to obtain an updated total cost map. Its internal path search algorithm operates on this total cost map and outputs a local path with the lowest cost. This path smoothly avoids the high-cost area formed by markers 50a and 50b and is guided to the low-cost area formed by marker 50c, ultimately leading to the temporary detour dirt road 62. The vehicle control system executes this path, enabling the large mining truck to complete the detour safely and smoothly.

[0092] This embodiment demonstrates how the present invention provides a flexible and safe navigation and guidance scheme for autonomous vehicles in temporary and complex road scenarios by combining and deploying different types of fixed markers and adaptively adjusting to the vehicle's own state. This scheme does not require pre-coverage of high-precision maps.

[0093] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automated driving navigation control method based on fixed calibration objects, characterized in that, Includes the following steps: Detect at least one fixed marker in the driving environment of an autonomous vehicle; Decode the at least one fixed calibrator to obtain the potential field parameters used to define the virtual potential field; Based on the potential field parameters, the virtual potential field is generated on the vehicle's navigation cost map; Based on the navigation cost map overlaid with the virtual potential field, a local driving path is replanned for the autonomous vehicle.

2. The autonomous driving navigation control method based on fixed calibration objects according to claim 1, characterized in that, The step of generating the virtual potential field on the vehicle's navigation cost map based on the potential field parameters further includes: The vehicle's own status is acquired in real time; The virtual potential field is generated based on the vehicle's own state and the potential field parameters.

3. The autonomous driving navigation control method based on fixed calibration objects according to claim 2, characterized in that, The step of generating the virtual potential field based on the vehicle's own state and the potential field parameters includes: Using the vehicle's own state, the basic strength and / or basic influence range used to define the potential field shape in the potential field parameters are modified to obtain the final potential field strength and / or the final potential field influence range.

4. The autonomous driving navigation control method based on fixed calibration objects according to claim 3, characterized in that, The vehicle's own status includes the vehicle's current speed and / or vehicle category; The step of modifying the basic strength and / or basic influence range used to define the potential field shape in the potential field parameters by utilizing the vehicle's own state includes: The final potential field influence range is adjusted according to the vehicle's current speed, and / or the geometry of the virtual potential field is adjusted according to the vehicle category.

5. The autonomous driving navigation control method based on fixed calibration objects according to claim 1, characterized in that, When multiple fixed calibration objects are detected, the step of generating the virtual potential field includes: Based on multiple sets of potential field parameters decoded from the multiple fixed calibration objects, and according to a preset composite rule, a composite virtual potential field is generated as the virtual potential field.

6. The autonomous driving navigation control method based on fixed calibration objects according to claim 5, characterized in that, The preset composite rules include guidance channel rules, specifically: When the plurality of fixed calibration objects include both repulsive and attractive calibration objects, the repulsive potential field generated by the repulsive calibration object and the attractive potential field generated by the attractive calibration object are superimposed to form a composite virtual potential field with a guiding channel.

7. The autonomous driving navigation control method based on fixed calibration objects according to claim 5, characterized in that, The preset composite rules include conditional activation rules, specifically: When the plurality of fixed calibration objects include a condition-triggered calibration object, the system determines whether a preset activation condition is met based on the real-time sensor data of the autonomous vehicle, and generates a virtual potential field corresponding to the condition-triggered calibration object only when the preset activation condition is met.

8. The autonomous driving navigation control method based on fixed calibration objects according to claim 1, characterized in that, The method further includes: After detecting the removal of the marker or the autonomous vehicle leaving the perception range of the fixed marker, the virtual potential field is removed from the navigation cost map.

9. The autonomous driving navigation control method based on fixed calibration objects according to claim 1, characterized in that, The step of replanning a local driving path for the autonomous vehicle based on the navigation cost map superimposed with the virtual potential field includes: The virtual potential field is superimposed on the basic navigation cost map to form the total cost map; The path planning algorithm is used to search for the path with the lowest cost on the total cost map, which is then used as the local driving path.

10. An automated driving navigation control system based on a fixed benchmark, used to perform the method as described in any one of claims 1-9, characterized in that, include: The perception decoding module is used to detect at least one fixed marker in the driving environment of the autonomous vehicle and decode it to obtain potential field parameters. A potential field generation module is used to generate a virtual potential field on the vehicle's navigation cost map based on the potential field parameters. The path planning module is used to replan local driving paths for the autonomous vehicle based on a navigation cost map overlaid with the virtual potential field.

Images