AVP parking lot parking space reservation and automatic parking collaborative control method and system

Abstract

The application belongs to the technical field of intelligent traffic control, and specifically discloses an AVP parking lot parking space reservation and automatic parking cooperative control method and system, wherein the method comprises the following steps: responding to a parking request, obtaining traffic density and queuing time, dimensionally reducing and mapping the traffic density and the queuing time into a space cost, and planning a reference path according to the space cost; when the vehicle is stationary, obtaining a physical baseline formed by two positioning points of the vehicle body, compensating the physical baseline by using a pitch angle for horizontal projection, and combining the compensated effective baseline and the plane coordinates of the positioning points to solve an initial heading; the application aims to solve the problem that there is a barrier between macroscopic space-time resource scheduling and microscopic vehicle chassis kinematics constraint in a high-density underground parking lot environment, which leads to the problems that the prior art is difficult to realize dimension unified optimization, zero-speed accurate pose determination, space-time phase isolation avoidance, adhesion self-adaptive coupling and network interruption safety passive fallback.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, specifically to a method and system for coordinated control of AVP parking space reservation and automatic parking. Background Technology

[0002] In high-concurrency closed parking lots, existing parking control systems have many limitations.

[0003] First, in the macro-path planning stage, existing optimization algorithms typically use multi-objective weighting of geometric distance, expected queuing time and congestion density. This approach has the problem of inconsistent physical dimensions, which not only easily leads to deviations in calculation results, but may also guide a large number of vehicles to the same main road, causing congestion deadlock.

[0004] Secondly, when a vehicle loses satellite signal after entering an underground parking garage, existing indoor positioning technology struggles to accurately calculate the initial heading angle at the moment the vehicle starts at zero speed. Furthermore, when facing an entrance gate with a slope, the three-dimensional elevation difference can cause nonlinear projection errors in the mapping of the vehicle's physical baseline onto a two-dimensional map, thus affecting the accuracy of subsequent trajectory coordinates.

[0005] Secondly, during multi-vehicle merging, traditional control schemes rely on passive braking and avoidance within line of sight, lacking a collaborative mechanism to transform spatial conflicts into phase isolation on the time axis, easily causing intersection stalls. Furthermore, underground parking lots often use epoxy resin flooring, which reduces the road surface adhesion coefficient in wet conditions. Existing local obstacle avoidance algorithms typically use fixed potential field repulsion weights, failing to dynamically couple with the vehicle's underlying tire friction limits, leading to a risk of sideslip when vehicles execute obstacle avoidance commands on slippery surfaces. Simultaneously, due to the obstruction of underground building structures, cross-node communication links are prone to disconnection, potentially causing vehicles to lose control guidance, lacking a degradation and backoff safety mechanism based on underlying physical sensors.

[0006] Therefore, how to combine macroscopic spatiotemporal resource scheduling with microscopic vehicle physical kinematic constraints to provide a fully automatic valet parking collaborative control method that can achieve dimensional uniform optimization, zero-velocity attitude determination, spatiotemporal phase isolation and avoidance, adhesion adaptive coupling, and disconnection safety passive retreat is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0007] This invention provides a collaborative control method and system for AVP parking space reservation and automatic parking. The purpose is to solve the problem that there is a barrier between macroscopic spatiotemporal resource scheduling and microscopic vehicle chassis physical kinematic constraints in high-density underground parking environments, which makes it difficult for existing technologies to achieve dimensionless optimization, zero-speed precise attitude determination, spatiotemporal phase isolation and avoidance, adhesion adaptive coupling, and safe passive back-off when the network is disconnected.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: AVP parking lot space reservation and automatic parking collaborative control method and system, including: responding to parking requests, acquiring traffic flow density and queuing time, mapping the traffic flow density and queuing time to spatial cost in a dimension reduction manner, and planning a reference path based on the spatial cost; when the vehicle is stationary, acquiring a physical baseline formed by two positioning points of the vehicle body, performing horizontal projection compensation on the physical baseline using pitch angle, and solving for the initial heading by combining the compensated effective baseline and the plane coordinates of the positioning points; constructing a spatiotemporal corridor based on the initial heading and the reference path, and when path spatial overlap is detected, calculating based on the departure time of the preceding vehicle. Upon reaching the arrival time limit, a phase constraint with time redundancy is applied to subsequent vehicles, and a cruise curve is generated based on the phase constraint. The adhesion coefficient of the current road surface is estimated, and the repulsion weight of the potential field method is dynamically adjusted according to the adhesion coefficient. The repulsion field corresponding to the repulsion weight is embedded into the prediction algorithm, and path tracking and local obstacle avoidance are performed along the cruise curve. When the communication interruption exceeds the threshold, the path tracking command is frozen. Using local grid data and dead reckoning, the vehicle is guided to slow down and drive to the nearest idle node and park. After communication is restored and the vehicle reaches the target parking space, it reverses into the parking space and sends a locking signal.

[0009] In one aspect of the invention, responding to a parking request, obtaining traffic flow density and queuing time, mapping the traffic flow density and queuing time to a spatial cost in a reduced dimension, and planning a reference path based on the spatial cost include: Extract the geometric distance from the current node to the target storage location; Extract the traffic flow density around the current node and the maximum design density of the road segment, and calculate the density ratio; Multiply the density ratio by a preset distance constant to obtain the density cost; Multiply the queuing time by the set average vehicle speed to obtain the time cost; The geometric distance, density cost, and time cost are superimposed by a scalar to obtain the spatial cost, and the path with the minimum spatial cost is selected as the reference path.

[0010] In one aspect of the invention, when the vehicle is stationary, a physical baseline formed by two positioning points of the vehicle body is obtained, and the physical baseline is horizontally projected and compensated using a pitch angle. The initial heading is then calculated by combining the compensated effective baseline with the planar coordinates of the positioning points, including: Confirm that the vehicle is stationary and establish an external ranging communication link; Two positioning points located at the front center axle and rear center axle of the vehicle are obtained respectively, and the physical straight-line distance between the two positioning points is determined as the physical baseline. Obtain the pitch angle output by the vehicle status monitoring module, and extract the cosine value of the pitch angle; Multiply the physical baseline by the cosine value to obtain the effective baseline after horizontal projection compensation; Obtain the difference in planar coordinates between the two positioning points in the global coordinate system, and solve for the initial heading using the arctangent function in conjunction with the effective baseline.

[0011] In one aspect of the invention, the step of constructing a spatiotemporal corridor based on the initial heading and the reference path, and when path spatial overlap is detected, calculating an arrival time limit based on the departure time of the preceding vehicle, applying a phase constraint with temporal redundancy to the subsequent vehicle, and generating a cruise curve based on the phase constraint, includes: Based on the initial heading, the vehicle's physical boundary is mapped along the reference path to generate the three-dimensional spatiotemporal corridor containing a time axis; Extract the overlapping regions where the spatiotemporal tunnels of multiple vehicles physically intersect at the topological nodes; Predict the expected end time when the preceding vehicle has completely driven out of the overlapping area; The expected end time plus the safety redundancy interval is determined as the starting time limit for the subsequent vehicles to be allowed to enter the overlapping area. Using the starting time limit as the time-forced anchor point for phase constraints, the longitudinal deceleration profile of the subsequent vehicle before entering the overlapping area is reverse-engineered to generate the cruise curve.

[0012] In one aspect of the invention, estimating the adhesion coefficient of the current road surface, dynamically adjusting the repulsive force weight of the potential field method based on the adhesion coefficient, and embedding the repulsive force field corresponding to the repulsive force weight into the prediction algorithm, and performing path tracking and local obstacle avoidance along the cruise curve, includes: Collect the vehicle's drive wheel speed and actual longitudinal vehicle speed; The real-time slip ratio is calculated based on the drive wheel speed and the actual longitudinal vehicle speed, and the adhesion coefficient of the road surface is estimated. The repulsion weight is determined to be inversely proportional to the adhesion coefficient and directly proportional to the actual longitudinal vehicle speed, and the dynamic repulsion weight is calculated. The repulsive weights are applied to the potential field method to generate penalty constraint terms for dynamic obstacles; Substituting the penalty constraint into the quadratic programming solver of the model predictive control algorithm, the algorithm outputs flexible steering and braking commands that track the cruise curve and avoid obstacles.

[0013] In one aspect of the invention, when the determination of communication interruption exceeds a threshold, freezing the path tracking command, and using local grid data and dead reckoning, guiding the vehicle to slow down and drive to the nearest available node and park, includes: Monitor the status of communication heartbeat packets and record the time span of consecutive lost heartbeat packets; When the time span exceeds the set threshold, the original global path tracking control is forcibly cut off; Extract the local raster data of the topology environment last effectively cached before the communication interruption; The active power output command is cut off, and dead reckoning is performed solely based on the accumulation of relative position parameters; Based on the dead reckoning results and the local grid data, the nearest available node is matched, and the vehicle is controlled to travel to the available node at a speed lower than the safety limit and the parking brake is triggered.

[0014] In another aspect, the present invention also relates to an AVP parking space reservation and automatic parking cooperative control system, used to implement the aforementioned AVP parking space reservation and automatic parking cooperative control method, comprising: The path planning unit is used to respond to parking requests, obtain traffic flow density and queuing time, reduce the traffic flow density and queuing time to a spatial cost, and plan a reference path based on the spatial cost. The heading initialization unit is used to obtain the physical baseline formed by the two positioning points of the vehicle body when the vehicle is stationary, perform horizontal projection compensation on the physical baseline using the pitch angle, and solve the initial heading by combining the compensated effective baseline and the plane coordinates of the positioning points. The utility tunnel scheduling unit is used to construct a spatiotemporal utility tunnel based on the initial heading and the reference path. When the path space overlap is detected, the unit calculates the arrival time limit based on the departure time of the preceding vehicle, applies a phase constraint with time redundancy to the following vehicle, and generates a cruise curve based on the phase constraint. The trajectory tracking unit is used to estimate the adhesion coefficient of the current road surface, dynamically adjust the repulsive force weight of the potential field method according to the adhesion coefficient, and embed the repulsive force field corresponding to the repulsive force weight into the prediction algorithm to perform path tracking and local obstacle avoidance along the cruise curve. The downgrade rollback unit is used to freeze path tracking commands when the communication interruption exceeds the threshold. It uses local grid data and dead reckoning to guide the vehicle to slow down and drive to the nearest idle node and park. The in-warehouse locking unit is used to reverse into the warehouse and send a locking signal after communication is restored and the vehicle arrives at the target warehouse location.

[0015] In one aspect of the invention, the system further includes a state synchronization mechanism that uses a precise time protocol to achieve global clock alignment and broadcasts a synchronous ranging signal based on the two-way time-of-flight ranging principle.

[0016] In one aspect of the invention, the physical baseline is composed of positioning points located at the front center axle and the rear center axle of the vehicle, respectively, and the line connecting the two positioning points is parallel to the longitudinal plane of symmetry of the vehicle.

[0017] In one aspect of the invention, a parking space database is included, wherein the parking space attributes updated in real time include at least: parking space geometry, ground slope, and charging pile power parameters.

[0018] Compared with the prior art, the present invention has the following beneficial effects: In the macro-path planning stage, this invention resolves the physical dimension conflict problem in multi-objective weighted optimization by mapping traffic density and queuing time to a unified spatial length cost, thus avoiding congestion caused by concentrated traffic flow and improving the overall topology network's circulation efficiency. In the vehicle startup initialization stage, a horizontal projection compensation mechanism using the physical baseline and pitch angle of dual positioning points reduces the nonlinear mapping error caused by the three-dimensional slope elevation difference, enabling vehicles to calculate their initial heading at zero speed, compensating for the shortcomings of single-point positioning in initial heading acquisition. In the multi-vehicle intersection and merging stage, by constructing a spatiotemporal corridor and applying time phase translation constraints with safety redundancy, spatial avoidance is transformed into phase sequence isolation on the time axis, achieving intersection... Smooth alternating traffic flow; during the dynamic trajectory tracking phase, the real-time observed road surface adhesion coefficient is dynamically coupled with the repulsive force weight of the potential field method, adaptively adjusting the repulsive potential energy on low-adhesion roads, prompting the prediction algorithm to output obstacle avoidance commands in advance, reducing the risk of sideslip; in the face of cross-node communication disconnection, the system can pause the original path tracking command and switch to a relative positioning mode based on local grid and dead reckoning, guiding the vehicle to slow down and stop at the nearest idle node, providing a safety guarantee for retreat in the context of communication interruption; in the final parking stage, relying on multimodal perception and nonholonomic kinematic constraints to reduce cumulative drift error, and using locking signals to update the parking space status, realizing a closed-loop business process from resource scheduling to physical movement in the parking scenario. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts for the AVP parking space reservation and automatic parking collaborative control method of the present invention.

[0021] Figure 2This is the second flowchart of the AVP parking space reservation and automatic parking collaborative control method of the present invention.

[0022] Figure 3 This is a framework diagram of the AVP parking space reservation and automatic parking collaborative control system of the present invention. Detailed Implementation

[0023] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0024] Example 1 Please see Figures 1-3 As shown in the figure, this embodiment discloses a method and system for coordinated control of AVP parking space reservation and automatic parking, wherein the method specifically includes: Step 1: Respond to parking requests, obtain traffic density and queuing time, map the traffic density and queuing time into spatial cost in a dimension-reduced manner, and plan a reference path based on the spatial cost.

[0025] During this step, the system needs to process core parameters such as parking requests, traffic density, queuing time, spatial cost, and reference paths. A parking request refers to the trigger command data packet received by the system, containing a unique vehicle identifier and the topological logical endpoint coordinates of the target parking space. Traffic density refers to the scalar ratio of the currently physically occupied area to the total available passage area within a discrete grid or specific road segment arc length on the topology map. Queuing time, derived from queuing theory, is the expected dwell time of a vehicle due to congestion when passing through a specific physical bottleneck (such as an intersection). Spatial cost addresses the non-convex optimization problem caused by the inability to directly add different physical dimensions such as distance, time, and density in traditional optimization algorithms. By introducing a specific physical transformation mapping matrix, non-spatial obstacles are transformed into a single length dimension (meters), forming a unified scalar evaluation benchmark for shortest path search. The reference path is a geometric connection consisting of a series of discrete spatial coordinate points, connecting the vehicle's current topological starting point to the target endpoint parking space, with the minimum globally equivalent spatial cost. At this stage, the path has not yet been subject to time constraints.

[0026] Specifically, in a large, high-density underground parking lot with a multi-layered physical structure, traditional path optimization algorithms typically use the shortest geometric distance as the objective function, which easily leads to congestion by concentrating a large number of vehicles onto the same main physical road. Existing technologies attempt to use weighted multi-objective cost functions, but this violates the principle in physics that different dimensions cannot be directly added. To eliminate the dimensional conflict in existing technologies, this invention abstracts the parking lot as a directed topological graph. For any search connected edge in the topological graph, the system calculates its spatial cost according to the following corresponding mathematical process.

[0027] First, the system extracts the geometric distance from the current node to the target storage location. The formula for calculating the absolute Euclidean geometric length of a connected edge is: .

[0028] in, This represents the geometric distance between connected edges, measured in meters. This represents the physical horizontal, vertical, and coordinate systems of the starting node in the global three-dimensional coordinate system, with the dimension being meters. This represents the physical horizontal, vertical, and longitudinal coordinates of the termination node in the global three-dimensional coordinate system, with the dimension in meters.

[0029] Next, the system extracts the traffic flow density around the current node and the maximum design density of the road segment, calculates the density ratio, and multiplies the density ratio by a preset distance constant to obtain the density cost. The calculation formula is as follows: .

[0030] in, This represents the density cost, with the dimension of meters. This indicates the traffic density of the current road segment, measured in vehicles per square meter. This represents the maximum traffic density limit under the physical design limit of the road section, with the dimension being vehicles per square meter. Dividing the two results in a dimensionless congestion penalty coefficient. This represents a preset baseline distance constant, measured in meters.

[0031] Among them, the preset reference distance constant The value determination process is as follows: Obtain 15,000 valid historical parking trajectory data points from the target parking lot over the past 30 days. Extract the total time sampled for all vehicles actually traversing the congested road segment when the physical design maximum traffic density limit is reached. Combine this with the set average cruising speed to convert the time consumption into equivalent mileage under non-congested conditions. After removing outlier samples due to traffic accidents, calculate the difference between the remaining valid equivalent mileage samples and the actual physical length of the road segment. Calculate the mean of the normal distribution of this difference set and round it up to the tens place. Finally, based on the historical mean data, calculate a value of 100 meters as an example. The constant value of .

[0032] The system then multiplies the queuing time by the set average vehicle speed to obtain the time cost. The formula is: .

[0033] in, This represents the cost of time, measured in meters. This represents the expected queuing time, measured in seconds. This represents the set average cruising speed, measured in meters per second.

[0034] Among them, the set average cruising speed The value determination process is as follows: 50,000 monitoring data points of manually driven vehicles were obtained from the target parking lot over the past 6 months (excluding holidays). Instantaneous speed samples of vehicles on straight sections and regular curves were extracted. Abnormally low-speed samples below 0.5 m / s caused by searching for parking spaces and speeding samples exceeding the speed limit were removed. Statistical analysis was performed on the remaining valid speed samples to calculate the mean of their probability density distribution. This was then weighted by a 0.8-fold adjustment based on the autonomous driving safety redundancy coefficient. Finally, the calculation result was rounded down to the nearest integer. Based on the above limited historical data and processing steps, an example was calculated to yield 2 m / s as... The constant values ​​that can be taken.

[0035] This step converts the time spent waiting with the vehicle stationary into the equivalent spatial distance that can be covered while maintaining an average speed.

[0036] Finally, the system performs a scalar superposition of the geometric distance, the density cost, and the time cost to obtain the spatial cost, and selects the path with the minimum spatial cost as the reference path. The spatial cost reconstruction formula is: .

[0037] in, This represents the total spatial cost after comprehensive mapping of the road segment, with the unit of measurement being meters.

[0038] The system calls a graph search algorithm, which iteratively searches within a single scalar space after dimensionality reduction, using spatial cost as the cumulative cost, and outputs the reference path with the highest overall circulation efficiency.

[0039] Step 2: When the vehicle is stationary, obtain the physical baseline formed by the two positioning points of the vehicle body, use the pitch angle to perform horizontal projection compensation on the physical baseline, and combine the compensated effective baseline with the plane coordinates of the positioning points to solve for the initial heading.

[0040] This step involves concepts such as dual positioning points, physical baseline, pitch angle, horizontal projection compensation, effective baseline, and initial heading. Dual positioning points are two independent spatial coordinate ranging nodes fixed at the front and rear centers of symmetry of the vehicle's rigid body structure. The physical baseline is a straight line segment connecting the two dual positioning points, with a length that is a constant scalar of the rigid body. The pitch angle is the angle between the vehicle's longitudinal axis within its longitudinal plane of symmetry and the absolute horizontal plane. Horizontal projection compensation is the process of mapping a three-dimensional spatial line segment with a tilt angle to a two-dimensional absolute horizontal plane, using trigonometric functions to eliminate linear compression errors caused by elevation differences. The effective baseline is the equivalent mapped scalar length of the compensated physical baseline in the two-dimensional horizontal coordinate system. The initial heading is the vehicle's yaw angle relative to the coordinate axes in the global two-dimensional coordinate system. When the vehicle speed is zero, single-point positioning is mathematically unobservable and must rely on the spatial vector formed by the two points for static boundary constraint calculation.

[0041] When a vehicle enters the underground parking lot entrance gate from an open road, the system must immediately initialize the indoor positioning coordinate system due to the loss of satellite navigation signals. At the instant the vehicle comes to a complete stop, the system confirms the vehicle is stationary and establishes an external ranging communication link. The system acquires two positioning points located at the front and rear center axes of the vehicle, respectively, and determines the physical straight-line distance between the two positioning points as the physical baseline. If the vehicle is parked on a sloping entrance ramp, directly reading the planar coordinates to calculate the heading will introduce nonlinear coordinate matching distortion. Therefore, the system acquires the pitch angle output by the vehicle status monitoring module and extracts the cosine value of the pitch angle. Subsequently, the system multiplies the physical baseline by the cosine value to obtain the effective baseline after horizontal projection compensation. The calculation formula for this process is: .

[0042] in, This represents the effective baseline length after compensation, in meters. Indicates the length of the physical baseline, with the dimension of meters; It represents the vehicle's longitudinal pitch angle, measured in radians.

[0043] Finally, the system obtains the difference in planar coordinates between the two positioning points in the global coordinate system, and calculates the initial heading using the arctangent function in conjunction with the effective baseline. The system obtains the global planar coordinates of the positioning points by ranging with external observation base stations. The formula for calculating the absolute initial heading is: .

[0044] in, This represents the initial heading angle of the vehicle in the global two-dimensional absolute coordinate system, with the dimension of radians; and These represent the longitudinal coordinates of the front and rear positioning points on the absolute horizontal plane, respectively, with the dimension of meters; and These represent the horizontal coordinates of the front and rear positioning points on the absolute horizontal plane, respectively, with the dimension of meters; Through this closed-loop calculation, the system directly completes the initial pose calibration, including the precise yaw angle, at the stationary moment using the principle of geometric projection, thus satisfying the observability conditions of the environment at the moment of system startup.

[0045] Step 3: Construct a spatiotemporal corridor based on the initial heading and the reference path. When the path space overlaps, calculate the arrival time limit based on the departure time of the preceding vehicle, apply phase constraints with time redundancy to the following vehicles, and generate a cruise curve based on the phase constraints.

[0046] During this step, the system needs to process multiple dynamic parameters with rigorous physical and geometric definitions. The spatiotemporal corridor refers to the three-dimensional geometric envelope space formed by adding a time dimension (T-axis) to the reference path in a two-dimensional plane. It not only spatially defines the vehicle's trajectory but also strictly limits the specific physical position the vehicle must occupy at a specific point in time. Path space overlap refers to the physical state where the spatiotemporal corridors of multiple vehicles geometrically intersect at the intersection nodes of the topology map. The exit time of the preceding vehicle refers to the absolute time point at which the rear of the physical body of the vehicle that entered the overlapping area completely leaves the intersection area. The arrival time limit is the earliest absolute time point at which the system forcibly designates subsequent vehicles to enter the physical conflict area to avoid physical collisions. Temporal redundancy is a fixed interval constant reserved on the time axis to absorb system-level control errors, communication delays, and vehicle dynamic response lags. Phase constraint refers to the forced translational restriction imposed on the motion state of subsequent vehicles on the time axis, that is, delaying the originally expected arrival time of the vehicle. The cruise curve is a distribution profile that satisfies the above phase constraints and shows the continuous change of the vehicle's longitudinal target speed over time.

[0047] At an unlit intersection or T-junction within an underground parking garage, multiple vehicles simultaneously traveling along their respective reference paths are highly susceptible to deadlocks or collisions at the physical intersection. Traditional passive obstacle avoidance systems typically allow vehicles to reach the intersection's line-of-sight before relying on sensors to detect oncoming vehicles and trigger emergency braking. This approach not only disrupts traffic flow but also greatly increases the risk of rear-end collisions. This invention constructs a spatiotemporal corridor to resolve conflicts directly in the time domain before vehicles even reach the conflict area.

[0048] Specifically, the system first uses the precise initial heading obtained in step 2, combined with the vehicle's own length and width physical dimensions, to perform an integral mapping along the reference path planned in step 1. The system then converts the two-dimensional spatial coordinates... Extended to a three-dimensional coordinate system The system generates a 3D spatiotemporal tunnel containing a timeline. Next, it performs a spatial traversal on the global topology map, extracting overlapping areas where the spatiotemporal tunnels of multiple vehicles physically intersect at topology nodes. Once an overlapping area is detected, the system immediately sorts the vehicles involved in the conflict along their timelines, distinguishing between preceding and subsequent vehicles.

[0049] Subsequently, the system uses the current velocity vector and remaining path length of the preceding vehicle to predict the expected end time when the preceding vehicle completely exits the overlapping area. After obtaining this expected end time, the system no longer uses spatial distance as a safety threshold, but instead adds a preset safety time redundancy interval to it, determining this as the starting time limit for the subsequent vehicle to be allowed to enter the overlapping area. The underlying mathematical logic of this process is represented by inequality constraints on the time axis, and the relevant physical model formula is:

[0050] In the above formula, It represents the calculated arrival time limit, with the physical dimension of seconds (s), and represents the earliest absolute timestamp when the front end of the following vehicle is allowed to cross the physical boundary of the overlapping area.

[0051] It represents the predicted departure time of the preceding vehicle, with the physical dimension of seconds (s), and is the expected absolute timestamp of the rear end face of the preceding vehicle completely leaving the physical boundary of the overlapping area.

[0052] This represents the system's set time redundancy interval constant, with its physical dimension being seconds (s). In this embodiment, it is specifically set to 2.0 seconds.

[0053] Among them, the preset safety time redundancy interval constant The value determination process is as follows: 5000 historical trajectory data points from multiple vehicle merging events at parking lot intersections within the past 6 months are retrieved. The actual time difference between the moment the preceding vehicle completely exits the node boundary and the moment the following vehicle's front end just reaches the node boundary is extracted during each collision-free alternating passage. Invalid samples causing sudden braking due to pedestrians crossing randomly are removed. Gaussian distribution fitting is applied to the remaining 4800 valid time difference samples. A time difference value of 1.5 seconds, satisfying the upper limit of the 99.7% confidence interval, is selected. Then, a 0.5-second system communication network delay and underlying mechanical actuator response compensation time are directly added to this value, ultimately resulting in a value of 2.0 seconds. The safety redundancy constraint threshold.

[0054] The physical significance of this formula lies in establishing a time isolation interval, which strictly transforms the spatial avoidance game in the traditional two-dimensional plane into phase separation on a one-dimensional time axis.

[0055] After determining the arrival time limit, the system uses this initial time limit as the time-forced anchor point for phase constraints. Starting from the current physical position and time of the following vehicle, and ending at the physical boundary and arrival time limit of the overlapping area at the intersection, the system reverse-engineers the longitudinal deceleration profile of the following vehicle before entering the overlapping area. The system generates the cruise curve by solving a two-point boundary value problem with terminal time constraints and maximum deceleration constraints. The functional expression for this process is:

[0056] Specifically, the inverse kinematic trajectory generation function A uniformly accelerated kinematic model is used for construction. The system first calculates the theoretical constant acceleration requirement that satisfies both time and space constraints. The formula is: ,in This represents the vehicle's current instantaneous speed.

[0057] Subsequently, the system applies dynamic boundary constraints to truncate the theoretical constant acceleration requirement: when When the velocity profile is deemed physically feasible, the generated cruise curve formula is output as follows: ; If it exceeds the boundary, then... or A piecewise polynomial fitting is triggered by the boundary. This specific model ensures that abstract phase constraints can be directly transformed into control sequences executable by the chassis motor.

[0058] In the above formula, This represents the cruise curve generated by the system for subsequent vehicles, i.e., their various future moments. The longitudinal target velocity should be followed, with the dimension of meters per second (m / s).

[0059] This represents the inverse kinematic trajectory generation function, which is constructed based on the principle of polynomial fitting or minimization.

[0060] It represents the remaining arc length distance from the current centroid position of the following vehicle to the starting point of the intersection and overlap area, with the dimension of meters (m).

[0061] It represents the current absolute time of the system, measured in seconds (s).

[0062] It represents the remaining time difference from the current moment until the boundary must be reached, and its dimension is seconds (s).

[0063] and These represent the maximum driving acceleration and maximum braking deceleration allowed by vehicle dynamics, respectively, with the dimension of meters per square second (m / s²).

[0064] By solving the constraints of the aforementioned physical equations, the system-issued cruise curve forces subsequent vehicles to begin a gentle deceleration maneuver on the straightaway phase, a considerable distance from the intersection, to eliminate any unnecessary time lag. When the subsequent vehicle arrives at the intersection using this cruise curve, its arrival time precisely satisfies... The requirements are met. At this point, the preceding vehicle has completely left the conflict area and left a sufficient physical safety distance, so that the following vehicle does not need to stop and wait at the intersection, realizing efficient and continuous alternating passage of multiple vehicles in narrow topological nodes.

[0065] Step 4: Estimate the adhesion coefficient of the current road surface, dynamically adjust the repulsion weight of the potential field method according to the adhesion coefficient, and embed the repulsion field corresponding to the repulsion weight into the prediction algorithm to perform path tracking and local obstacle avoidance along the cruise curve.

[0066] During this step, the control system needs to handle the interaction between vehicle microdynamics and sudden local environmental changes. The coefficient of adhesion is a physical quantity reflecting the maximum available friction between the tire and the current road surface; its value directly determines the physical limits of vehicle acceleration, braking, and steering. The repulsion weight is a dimensionless gain constant set in the artificial potential field method, used to define the strength of the spatial repulsion tendency of obstacles on the vehicle. The predictive algorithm refers to model predictive control, an optimization control strategy based on a vehicle dynamics model. It outputs the current control command by continuously solving for the minimum value of the objective cost function within a finite time window. Path tracking refers to the process of controlling the steering angle and driving force to make the vehicle's actual physical trajectory conform to the cruise curve generated in step 3. Local obstacle avoidance refers to the action of superimposing instantaneous steering and braking biases within the allowable range of the underlying physical limits during path tracking, in order to avoid dynamic obstacles (such as pedestrians and illegally parked vehicles) not present in the static planning.

[0067] Specifically, in traditional automatic parking systems, path tracking and local obstacle avoidance are usually two independent control modules. When an obstacle is detected, the system often uses a fixed safety distance threshold or a fixed potential field repulsion weight to trigger obstacle avoidance. However, underground parking lots are generally paved with epoxy resin flooring, and the adhesion coefficient of such surfaces decreases significantly when wet. If a vehicle is traveling at a high speed on a low-adhesion surface, and the system still uses the fixed repulsion weight under dry conditions, the late intervention of the repulsion force can easily cause the vehicle to instantly output steering or braking commands to avoid obstacles, exceeding the current friction limit of the tires (i.e., exceeding the Gaum friction circle), leading to vehicle lock-up and skidding, resulting in physical loss of control. This invention achieves deep algebraic coupling between the road surface physical state and the upper-level control algorithm through real-time observation of the underlying dynamic parameters.

[0068] In practice, the multi-constraint trajectory tracking execution terminal first collects the drive wheel speed and actual longitudinal vehicle speed output by the vehicle chassis sensors. Based on these two physical quantities, the system calculates the real-time tire slip ratio and uses this as the basis for estimating the current road surface adhesion coefficient. Subsequently, the system introduces nonlinear dynamic adjustment logic, determining that the repulsive force weight of the potential field method is inversely proportional to the estimated adhesion coefficient and directly proportional to the vehicle's current actual longitudinal speed, thereby calculating the dynamic repulsive force weight. The system assigns this repulsive force weight to obstacles, generating a penalty constraint term (i.e., repulsive force field) for dynamic obstacles. Finally, the system directly substitutes this penalty constraint term into the quadratic programming solver of the model predictive control algorithm, simultaneously performing joint optimization of trajectory tracking error and obstacle avoidance penalty within the solver, and outputting flexible steering and braking commands that satisfy vehicle dynamics constraints.

[0069] The underlying physical mechanism and mathematical derivation of this control process include the following equations. First is the observation equation for the adhesion coefficient:

[0070] In the above formula, This represents the estimated current road surface adhesion coefficient, which is dimensionless and has a physical effective range of 0 to 1.

[0071] This represents the angular velocity of the drive wheel collected by the wheel speed sensor, measured in radians per second (rad / s).

[0072] This represents the rolling radius of the wheel, measured in meters (m).

[0073] This represents the actual longitudinal speed of the vehicle calculated by a high-precision inertial navigation system, expressed in meters per second (m / s).

[0074] The fraction in parentheses calculates the real-time longitudinal slip ratio of the wheel.

[0075] This represents the observer mapping function based on nonlinear Kalman filtering, which dynamically infers the road surface adhesion state through slip ratio.

[0076] Based on this, the system calculates the dynamic repulsion weight:

[0077] In this formula, This represents the repulsive force weight after dynamic adjustment, and is dimensionless.

[0078] This represents the system's preset fundamental repulsion constant, which is dimensionless.

[0079] Among them, the preset basic repulsion constant The process of obtaining the value is as follows: 2000 historical extreme obstacle avoidance test data of similar vehicles on a dry epoxy resin floor were retrieved. Critical steering and braking state parameters were extracted at different initial vehicle speeds, ensuring the vehicle just barely avoided ABS triggering and did not cross the lateral safety physical boundary. A multiple regression analysis matrix was established for vehicle speed, adhesion coefficient, and the required maximum lateral repulsion force. The least squares method was used to identify the parameters of this matrix. After removing outliers with a fitting residual greater than 10%, the mean value of the basic repulsion force gain coefficient that maintains vehicle dynamic stability and prevents loss of control was calculated and rounded up. The final calculated value was taken as 15.0. The calibration value.

[0080] The physical meaning of this formula constructs an adaptive feedback between obstacle avoidance behavior and vehicle motion state: when the vehicle speed... The higher the coefficient of friction, the better. The lower the value (e.g., when encountering a puddle), the higher the repulsive force weight calculated by the system. It will be significantly enlarged proportionally.

[0081] This amplification effect enables the system to perceive the repulsion of obstacles from a greater physical distance, thereby extending the action time and distance of obstacle avoidance actions and avoiding the output of drastic commands by the underlying actuators.

[0082] Finally, the system completes control domain fusion in the optimization objective cost function of model predictive control:

[0083] In the cost function formula, This represents the objective cost function value that needs to be minimized by the quadratic programming solver.

[0084] and These represent the prediction step size and control step size of the model predictive control, respectively.

[0085] This indicates the future prediction based on the vehicle's kinematics model. The state vector of the step (including position, heading angle, etc.).

[0086] This represents the target state vector corresponding to the reference cruise curve.

[0087] The penalty weight matrix represents the state tracking error.

[0088] Indicates the future number The control vector of the step (including front wheel steering angle and driving force).

[0089] This represents the penalty weight matrix for controlling energy consumption.

[0090] This represents the current state calculated using the traditional artificial potential field method. and obstacles The reference repulsive potential energy field function between them.

[0091] Specifically, the physical meaning of this objective cost function formula lies in breaking down the system barriers between tracking and obstacle avoidance. The first two terms of the formula require the system to conform as closely as possible to the cruise curve generated in step 3 and to have smooth control actions. The third term of the formula, as a hard penalty term, requires the system to stay away from obstacles. When the solver seeks the minimum value within the polygonal constraint space of the vehicle adhesion limit, due to the dynamic repulsion weights... The physical limits of the road surface have been internalized, and the optimal control output of the system is... This manifests naturally as: on slippery surfaces, early and gentle steering to evade, while on high-traction surfaces, agile avoidance maneuvers are permitted at closer distances. Based on the first principles of control theory, this ensures the physical feasibility and safety of the underlying execution modules under complex operating conditions.

[0092] Step 5: When the communication interruption exceeds the threshold, freeze the path tracking command, use local grid data and dead reckoning to guide the vehicle to slow down and drive to the nearest idle node and park.

[0093] During this step, the control system needs to handle the extreme condition of physical link failure in cross-node communication within the vehicle network. Communication interruption refers to the physical state where periodic heartbeat data packets based on a precise time protocol between the multi-constraint trajectory tracking execution terminal and the global spatiotemporal phase modulation center cease updating due to physical obstruction or radio frequency signal attenuation. Local grid data refers to a topological matrix that discretizes the continuous space within a certain physical radius around the vehicle into a two-dimensional grid, and uses binary or probabilistic numerical values ​​to mark whether each grid is occupied by static obstacles or road boundaries. Dead reckoning is an autonomous relative positioning physical mechanism that cuts off dependence on external positioning signals. It deduces the current relative position coordinates by continuously integrating internal sensor kinematic parameters from a known initial position. An idle node refers to a safe physical space in the local grid data that is statically marked as passable and is not topologically related to a major road intersection.

[0094] Specifically, in large underground parking lot engineering environments, intermittent interruptions in cross-node communication links are inevitable due to the obstruction of reinforced concrete load-bearing walls. If a vehicle continues to execute an outdated spatiotemporal corridor path after losing global instructions, it will face the risk of colliding with newly appearing dynamic obstacles. If a vehicle triggers emergency braking directly on the main road at the moment of communication interruption, it not only violates the smoothness requirements of vehicle dynamics but also causes following vehicles to rear-end each other due to insufficient braking, thereby triggering widespread congestion in the topological road network. This invention achieves a smooth physical transition from a globally connected state to a single-vehicle degraded state by constructing an underlying topology fallback mechanism.

[0095] Specifically, the multi-constraint trajectory tracking execution terminal continuously monitors the communication heartbeat packet status in the underlying operating system and records the time span of consecutive heartbeat packet loss. When the time span exceeds the set threshold (e.g., 150 milliseconds), the process for determining the set communication interruption threshold is as follows: extract the heartbeat interaction logs of 1 million cross-node data packets from the V2X micro base stations deployed in the parking lot over the past year, filter out all network fluctuation segments that have experienced signal attenuation and physical obstruction, calculate the probability density distribution of the data packet arrival time interval (Jitter), locate the critical time difference inflection point that causes the vehicle's underlying control bus to determine timeout and trigger abnormal actuator jitter, remove long-tail abnormal data caused by base station power outages, take the 99th percentile value of the network fluctuation time interval distribution (142 milliseconds), and perform engineering rounding and alignment to finally obtain 150 milliseconds as the set threshold for triggering the underlying topology fallback safety mechanism, and the system determines that a communication interruption has occurred.

[0096] At this point, the multi-constraint trajectory tracking execution terminal forcibly cuts off the original global path tracking control and freezes the path tracking instructions that could not be updated due to the network outage. Simultaneously, the system immediately retrieves the local raster data of the topology environment, which was last validly issued and cached before the communication interruption, from the vehicle's memory.

[0097] After the active power output command is cut off, the vehicle loses its external positioning reference. The multi-constraint trajectory tracking execution terminal then relies solely on the chassis's original sensors to accumulate relative position parameters and perform the dead reckoning. Its core mathematical derivation depends on the discrete kinematic state transition equations under nonholonomic constraints of the vehicle: ; ; .

[0098] In the above formula, Indicates the current number The dead reckoning state vector for each discrete calculation cycle contains the vehicle's relative lateral coordinates, relative longitudinal coordinates, and yaw angle in the local grid map system.

[0099] This represents the dead reckoning state vector of the previous calculation cycle. It represents the instantaneous longitudinal linear velocity of the previous cycle, which is measured purely physically by the vehicle wheel-end encoder, and is expressed in meters per second (m / s).

[0100] It represents the yaw angle of the previous cycle, with the dimension of radians (rad).

[0101] This represents the time integration step of the discrete control system for a multi-constraint trajectory tracking execution terminal, with the dimension of seconds (s).

[0102] It represents the instantaneous yaw rate of the previous cycle, physically measured by the vehicle-mounted high-precision inertial measurement unit (gyroscope), with the dimension of radians per second (rad / s).

[0103] The physical significance of this kinematic formula lies in the fact that, in an isolated state cut off from external observation information, geometric calculus is performed using only the original physical relative parameters of the vehicle chassis. Although this accumulation will produce drift errors over long periods, its accuracy is sufficient to support the underlying relative displacement calculations during the short time window of communication interruption.

[0104] The multi-constraint trajectory tracking execution terminal performs spatial matching between the coordinates updated in real time based on the dead reckoning and the extracted local grid data. By running a breadth-first search algorithm within the local grid, the system quickly identifies the nearest available node (e.g., a safe blind spot between two load-bearing pillars) to the vehicle's current position. Subsequently, the multi-constraint trajectory tracking execution terminal controls the vehicle to slow down to a speed below the safety limit (e.g., limited to 1.0 m / s) and proceed to the available node. Once the vehicle's geometric center enters the boundary of the available node, the system instructs the chassis controller to reduce the wheel speed to zero and trigger the electromechanical parking brake calipers, bringing the vehicle into a safe stationary waiting state until the cross-node communication link is restored. This step constitutes the underlying safety defense line of the entire cooperative control system.

[0105] The process for determining the safe speed limit is as follows: extract historical driving data of all vehicles in the parking lot under extreme conditions such as narrow passages, low-adhesion road surfaces, and communication interruptions. Combine this with the vehicle emergency braking distance model to calculate the maximum speed that ensures safe braking when an unpredictable obstacle is detected ahead. At the same time, refer to the speed limit requirements of relevant domestic and international AVP regulations for vehicles in automatic reverse mode. After comprehensive consideration, the safe speed limit is set to 1.0 m / s.

[0106] Step 6: After communication is restored and the vehicle reaches the target parking space, reverse into the parking space and send a locking signal.

[0107] In this step, the control logic switches from macro-level scheduling back to micro-level physical alignment, involving pose alignment, the parking search area, nonholonomic constraints, reversing into parking space, and locking signals. Pose alignment is the transition process when a vehicle recovers from its dead reckoning state to its global positioning state, eliminating accumulated drift errors and remapping the local coordinate system to the global absolute coordinate system. The parking search area is a specific geometric space directly in front of the target physical parking space. Nonholonomic constraints are wheeled vehicle dynamics limitations based on the Ackermann steering principle, meaning the vehicle cannot generate purely lateral translational speeds. Reversing into parking space is the physical movement process within a confined physical space, following nonholonomic constraints, by continuously adjusting the front wheel steering angle and longitudinal speed to align the vehicle with the target parking space and align its heading. The locking signal is a state machine switching instruction at the resource scheduling level, indicating that the physical space has been occupied and the system must release the reservation status of the parking space.

[0108] Specifically, after the vehicle leaves the communication blind spot, the system monitors the communication link status and confirms that the system's dead reckoning coordinates are aligned with the absolute coordinates. The system smoothly guides the drifted internal coordinates to the true global coordinates by introducing external absolute observations. After coordinate alignment, the system controls the vehicle to drive to the parking search area in front of the target parking space. Upon entering this area, the system activates the vehicle's end-sensing components to identify the physical parking space boundaries. Instead of relying on a macroscopic topology map, the system directly extracts geometric features from the physical parking space lines and adjacent vehicles on the ground to construct a local environmental polar coordinate grid.

[0109] Subsequently, the system plans the trajectory based on a locally nonholonomic constraint kinematics algorithm to perform the reversing into the parking space. Within the locally confined space, the system constructs a kinematic model with nonholonomic constraints:

[0110]

[0111]

[0112] in, and The longitudinal and lateral velocity components of the rear axle center of the vehicle in a local two-dimensional coordinate system are expressed in meters per second. It represents the yaw rate of a vehicle, measured in radians per second. This represents the longitudinal linear velocity of the vehicle's rear axle center, measured in meters per second. This represents the vehicle's current heading angle, measured in radians. It represents the actual physical steering angle of the vehicle's front wheels, measured in radians. This represents the constant wheelbase between the front and rear axles of a vehicle, measured in meters.

[0113] The system adjusts and Calculate the parking trajectory that satisfies the terminal state boundary conditions.

[0114] The system solves the extreme value problem of the reversing trajectory that minimizes the pose error, guiding the vehicle to reverse into the parking space until the parking termination boundary conditions are met. The boundary condition formula is as follows: ,as well as .

[0115] in, and This indicates the real-time coordinates of the vehicle's rear axle center during the parking process, expressed in meters. and This indicates the coordinates of the preset stop point of the target storage location, expressed in meters. This represents the vehicle's real-time heading angle, measured in radians. The angle of the normal to the target storage location is expressed in radians. This represents the maximum allowable physical error tolerance for positional alignment in the system, expressed in meters. Among them, the maximum allowable physical error tolerance for position alignment in the system. The value determination process is as follows: 3000 historical automatic parking tests of the same model of autonomous vehicles were collected in standard physical parking spaces. The actual physical distance from the outermost point of the vehicle's outline to the parking space boundary line after each parking maneuver was extracted. Anti-scratch margin was calculated based on the maximum difference between the national standard parking space width and the vehicle width. Failed samples due to occasional sensor malfunctions that caused deviations from the parking space line were removed. The positional deviation of the remaining successful parking samples was fitted with a normal distribution. The upper limit of the 95% confidence interval, which ensures normal vehicle door opening and does not affect adjacent parking spaces, was selected and rounded down to the millimeter level. Based on this historical data processing, a value of 0.05 meters was obtained as the [value]. The set threshold.

[0116] This represents the maximum allowable physical error tolerance for heading alignment in radians.

[0117] Among them, the maximum allowable heading alignment physical error tolerance of the system The value determination process is as follows: Retrieve the aforementioned 3000 historical automatic parking test data, extract the actual angle sample between the vehicle's longitudinal axis centerline and the longitudinal boundary line of the standard physical parking space after each vehicle comes to a complete stop, combine this with the critical angle geometric model of vehicles with different aspect ratios intruding into adjacent parking spaces at extreme deflection angles, eliminate outlier samples caused by slippage or control divergence, calculate the root mean square error of the remaining effective angle deviation sample set, and set a safety margin of 1.5. Round the obtained theoretical tolerance threshold down to the percentile. Based on the aforementioned limited historical data and processing steps, a value of 0.02 radians is calculated as... The set threshold.

[0118] When the vehicle's kinematic state meets the aforementioned boundary conditions, it indicates that the vehicle is properly parked in the target parking space. The system confirms that the vehicle has stopped and the parking system is activated, and sends the locking signal to the database via the restored communication link. Upon receiving the instruction, the database updates the parking space attribute status, thereby declaring that the vehicle's collaborative parking cycle has achieved a closed-loop control at both the underlying physical and upper-level logical levels.

[0119] In some optional embodiments, in a large parking lot with a building area exceeding 20,000 square meters and including a three-story underground structure, when a vehicle equipped with an automated parking system initiates a parking request at the entrance, the path planning unit first invokes spatial cost dimensionality reduction mapping logic to filter the optimal reference path. Assume that while traversing the topology map, the system is calculating the connected edge cost from topology node 1 (global coordinates 0, 0, 0) to topology node 2 (global coordinates 50, 0, -5). The system first calculates the geometric distance between the two points: The distance is measured in meters. The sensing device then reports that the current traffic density on this road segment is 0.08 vehicles per square meter, while the known physical design maximum density limit is 0.1 vehicles per square meter. The preset distance constant is 100 meters. The system calculates the density cost:

[0120] The density cost is also reduced to meters. Meanwhile, the model predicts a 15-second queuing time at this node due to oncoming traffic. The system's average cruising speed is set at 2 meters per second. Therefore, the time cost is:

[0121] The calculated time cost is converted to meters. Finally, the path planning unit superimposes the three dimensionality-reduced scalars to obtain the total space cost of the connected edge:

[0122] By comparing the cumulative spatial cost of each candidate path across the entire map, the system selects the path with the smallest total cost as the reference path for distribution, thereby avoiding the introduction of vehicles into areas with short physical distances but high congestion levels at the source.

[0123] When the vehicle receives the reference path and prepares to start from the entrance gate, it stops precisely on a downhill slope. At this point, the vehicle is stationary. To obtain an accurate initial heading at zero speed, the heading initialization unit calls upon the two fixed positioning points of the vehicle body for attitude determination calculation. The physical baseline constants formed by the front and rear positioning points of the vehicle are known. Meters, the current longitudinal pitch angle output by the vehicle-mounted sensor. Radius. The system first performs horizontal projection compensation to calculate the effective baseline:

[0124] The effective baseline is in meters. Next, via an external ranging communication link, the system obtains the horizontal coordinates of the forward positioning point as (10.0, 5.0) and the horizontal coordinates of the rear positioning point as (7.1, 4.5), both in meters. The system substitutes the coordinate difference into the attitude determination formula to solve for the initial heading:

[0125] The calculated initial heading is in radians. This calculation eliminates the 0.036-meter baseline projection compression error caused by the three-dimensional slope elevation difference, enabling the vehicle to have observable attitude boundary conditions at the moment of start-up.

[0126] When a vehicle starts and travels along a reference path to an intersection without traffic lights, the utility tunnel scheduling unit detects that the trajectory of this vehicle (the following vehicle) overlaps with the path of another vehicle crossing the intersection (the preceding vehicle) within the constructed spatiotemporal utility tunnel. The system predicts the expected end time when the preceding vehicle completely exits the overlapping area. Seconds, combined with the set time redundancy The time limit for subsequent vehicles to enter the area is calculated in seconds.

[0127] Assuming the current time Seconds, the remaining distance of the following vehicle from the start of the intersection area. meters, current instantaneous vehicle speed Meters per second. The system extracts the remaining time difference as 12.0 seconds and calculates the theoretical constant deceleration requirement:

[0128] The unit of this deceleration is meters per square second. After confirming that this value is within the vehicle dynamics boundary, the gallery scheduling unit then applies phase constraints and generates a cruise curve:

[0129] The vehicle begins to execute this gentle braking profile to eliminate the excess time difference, ensuring it arrives at the intersection at 102.0 seconds, achieving a smooth, alternating passage without stopping.

[0130] After passing the intersection, the vehicle entered a straight section of road with a wet surface, where an obstacle suddenly appeared on the side. The system's chassis sensors collected the drive wheel speeds in real time. Radius per second, wheel rolling radius meters, while the actual longitudinal speed at this time Meters per second. The trajectory tracking unit calculates the current real-time wheel slip rate:

[0131] The system estimates the current road surface adhesion coefficient based on this slip ratio. The system's pre-defined fundamental repulsion constant is known. The system dynamically adjusts the repulsive force weight:

[0132] Subsequently, the trajectory tracking unit embeds the repulsive field corresponding to the repulsive weight of 112.5 into the objective cost function of the quadratic programming solver of the prediction algorithm:

[0133] Under the combined effects of low adhesion and high speed, the solver begins to output gentle steering and slight deceleration control vectors at a distance from the obstacle, avoiding tire lock-up and sideslip caused by emergency obstacle avoidance on wet and slippery roads, and achieving path tracking and local obstacle avoidance along the cruise curve.

[0134] When the vehicle continued driving behind a thick load-bearing wall, the continuous loss of communication heartbeat packets reached 200 milliseconds, exceeding the system's set communication interruption threshold of 150 milliseconds. The degradation and rollback unit determined that the communication interruption exceeded the threshold, immediately froze the path tracking instructions, retrieved the local raster data cached before the network outage, and began executing dead reckoning. Assuming the vehicle state vector from the previous calculation cycle... With a horizontal coordinate of 50.0 meters, a vertical coordinate of 20.0 meters, and a yaw angle of 1.5 radians, the longitudinal linear velocity of the previous cycle... meters per second, yaw rate radians per second, discrete integral step size Seconds. The system calculates the kinematic integral increment of the lateral coordinate:

[0135] Vertical coordinate kinematic integral increment:

[0136] Yaw angle kinematic integral increment:

[0137] The coordinates calculated incrementally are in meters, and the angles are in radians. The updated current state vector. The dead reckoning value is (50.0014, 20.0198, 1.502). Based on this dead reckoning, the vehicle traveled 4.5 meters at a speed of less than 1.0 meters per second while the communication was interrupted, and finally came to a smooth stop in the nearest available node matched by the local grid data, where it parked and waited.

[0138] After communication is restored, the vehicle restarts from the idle node and enters the area in front of the target storage location. The entry locking unit establishes a set of nonholonomic constraint kinematic equations in the local coordinate system:

[0139]

[0140]

[0141] Of these, 2.8 meters is the vehicle's wheelbase. The longitudinal linear velocity is expressed in meters per second. This refers to the front wheel steering angle, in radians. The system continuously adjusts these two parameters to drive the vehicle in reverse into a parking space. As the reversing maneuver nears completion, the system monitors the rear axle center coordinates as (100.03, 49.98) and the current heading angle as 1.56 radians; the target stopping point coordinates are known to be (100.0, 50.0), and the target normal orientation angle is 1.57 radians. The system then performs a terminal position error check.

[0142] The position error is in meters and meets the set tolerance of less than or equal to 0.05 meters. Simultaneously, terminal heading error verification is performed.

[0143] The heading error is measured in radians and meets the set tolerance of less than or equal to 0.02 radians. At this point, the system confirms that the vehicle is properly parked in the target parking space, engages the parking brake, and sends a locking signal to the cloud. The database then updates the status of the parking space, thus completing the collaborative control closed loop in this fully automated valet parking scenario.

[0144] 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 AVP parking lot space reservation and automatic parking cooperative control method, characterized in that, include: In response to a parking request, the system obtains the traffic flow density and queuing time, maps the traffic flow density and queuing time to a spatial cost in a reduced dimension, and plans a reference path based on the spatial cost. When the vehicle is stationary, the physical baseline formed by the two positioning points of the vehicle body is obtained. The horizontal projection compensation of the physical baseline is performed using the pitch angle. The initial heading is solved by combining the compensated effective baseline and the plane coordinates of the positioning points. A spatiotemporal corridor is constructed based on the initial heading and the reference path. When the path space overlap is detected, the arrival time limit is calculated based on the departure time of the preceding vehicle. A phase constraint with time redundancy is applied to the subsequent vehicle, and a cruise curve is generated based on the phase constraint. Estimate the adhesion coefficient of the current road surface, dynamically adjust the repulsive force weight of the potential field method according to the adhesion coefficient, and embed the repulsive force field corresponding to the repulsive force weight into the prediction algorithm to perform path tracking and local obstacle avoidance along the cruise curve. When the communication interruption exceeds the threshold, the path tracking command is frozen, and the vehicle is guided to slow down and drive to the nearest available node and park using local grid data and dead reckoning. Once communication is restored and the vehicle reaches the target storage location, it will reverse into the storage space and send a locking signal.

2. The AVP parking space reservation and automatic parking cooperative control method according to claim 1, characterized in that, The process of responding to a parking request involves obtaining traffic density and queuing time, mapping the traffic density and queuing time to a spatial cost using a reduced-dimensionality method, and planning a reference path based on the spatial cost, including: Extract the geometric distance from the current node to the target storage location; Extract the traffic flow density around the current node and the maximum design density of the road segment, and calculate the density ratio; Multiply the density ratio by a preset distance constant to obtain the density cost; Multiply the queuing time by the set average vehicle speed to obtain the time cost; The geometric distance, density cost, and time cost are superimposed by a scalar to obtain the spatial cost, and the path with the minimum spatial cost is selected as the reference path.

3. The AVP parking space reservation and automatic parking cooperative control method according to claim 1, characterized in that, When the vehicle is stationary, a physical baseline formed by two positioning points of the vehicle body is obtained. The physical baseline is then horizontally projected and compensated using the pitch angle. The initial heading is then calculated by combining the compensated effective baseline with the planar coordinates of the positioning points, including: Confirm that the vehicle is stationary and establish an external ranging communication link; Two positioning points located at the front center axle and rear center axle of the vehicle are obtained respectively, and the physical straight-line distance between the two positioning points is determined as the physical baseline. Obtain the pitch angle output by the vehicle status monitoring module, and extract the cosine value of the pitch angle; Multiply the physical baseline by the cosine value to obtain the effective baseline after horizontal projection compensation; Obtain the difference in planar coordinates between the two positioning points in the global coordinate system, and solve for the initial heading using the arctangent function in conjunction with the effective baseline.

4. The AVP parking space reservation and automatic parking cooperative control method according to claim 1, characterized in that, The process of constructing a spatiotemporal corridor based on the initial heading and the reference path, and when path spatial overlap is detected, calculating the arrival time limit based on the departure time of the preceding vehicle, applying phase constraints with time redundancy to the subsequent vehicles, and generating a cruise curve based on the phase constraints, includes: Based on the initial heading, the vehicle's physical boundary is mapped along the reference path to generate the three-dimensional spatiotemporal corridor containing a time axis; Extract the overlapping regions where the spatiotemporal tunnels of multiple vehicles physically intersect at the topological nodes; Predict the expected end time when the preceding vehicle has completely driven out of the overlapping area; The expected end time plus the safety redundancy interval is determined as the starting time limit for the subsequent vehicles to be allowed to enter the overlapping area. Using the starting time limit as the time-forced anchor point for phase constraints, the longitudinal deceleration profile of the subsequent vehicle before entering the overlapping area is reverse-engineered to generate the cruise curve.

5. The AVP parking space reservation and automatic parking cooperative control method according to claim 1, characterized in that, The process of estimating the adhesion coefficient of the current road surface, dynamically adjusting the repulsive force weight of the potential field method based on the adhesion coefficient, and embedding the repulsive force field corresponding to the repulsive force weight into the prediction algorithm, along with path tracking and local obstacle avoidance along the cruise curve, includes: Collect the vehicle's drive wheel speed and actual longitudinal vehicle speed; The real-time slip ratio is calculated based on the drive wheel speed and the actual longitudinal vehicle speed, and the adhesion coefficient of the road surface is estimated. The repulsion weight is determined to be inversely proportional to the adhesion coefficient and directly proportional to the actual longitudinal vehicle speed, and the dynamic repulsion weight is calculated. The repulsive weights are applied to the potential field method to generate penalty constraint terms for dynamic obstacles; Substituting the penalty constraint into the quadratic programming solver of the model predictive control algorithm, the algorithm outputs flexible steering and braking commands that track the cruise curve and avoid obstacles.

6. The AVP parking space reservation and automatic parking cooperative control method according to claim 1, characterized in that, When the determination of communication interruption exceeds the threshold, the path tracking command is frozen, and the vehicle is guided to slow down and park at the nearest available node using local grid data and dead reckoning, including: Monitor the status of communication heartbeat packets and record the time span of consecutive lost heartbeat packets; When the time span exceeds the set threshold, the original global path tracking control is forcibly cut off; Extract the local raster data of the topology environment last effectively cached before the communication interruption; The active power output command is cut off, and dead reckoning is performed solely based on the accumulation of relative position parameters; Based on the dead reckoning results and the local grid data, the nearest available node is matched, and the vehicle is controlled to travel to the available node at a speed lower than the safety limit and the parking brake is triggered.

7. An AVP parking space reservation and automatic parking cooperative control system, characterized in that... The method for implementing the AVP parking space reservation and automatic parking cooperative control method as described in any one of claims 1 to 6 includes: The path planning unit is used to respond to parking requests, obtain traffic flow density and queuing time, reduce the traffic flow density and queuing time to a spatial cost, and plan a reference path based on the spatial cost. The heading initialization unit is used to obtain the physical baseline formed by the two positioning points of the vehicle body when the vehicle is stationary, perform horizontal projection compensation on the physical baseline using the pitch angle, and solve the initial heading by combining the compensated effective baseline and the plane coordinates of the positioning points. The utility tunnel scheduling unit is used to construct a spatiotemporal utility tunnel based on the initial heading and the reference path. When the path space overlap is detected, the unit calculates the arrival time limit based on the departure time of the preceding vehicle, applies a phase constraint with time redundancy to the following vehicle, and generates a cruise curve based on the phase constraint. The trajectory tracking unit is used to estimate the adhesion coefficient of the current road surface, dynamically adjust the repulsive force weight of the potential field method according to the adhesion coefficient, and embed the repulsive force field corresponding to the repulsive force weight into the prediction algorithm to perform path tracking and local obstacle avoidance along the cruise curve. The downgrade rollback unit is used to freeze path tracking commands when the communication interruption exceeds the threshold. It uses local grid data and dead reckoning to guide the vehicle to slow down and drive to the nearest idle node and park. The in-warehouse locking unit is used to reverse into the warehouse and send a locking signal after communication is restored and the vehicle arrives at the target warehouse location.

8. The AVP parking space reservation and automatic parking cooperative control system according to claim 7, characterized in that, It includes a state synchronization mechanism, uses a precise time protocol to achieve global clock alignment, and broadcasts a synchronous ranging signal based on the two-way time-of-flight ranging principle.

9. The AVP parking space reservation and automatic parking collaborative control system according to claim 7, characterized in that, The physical baseline is composed of positioning points located at the front center axle and the rear center axle of the vehicle, respectively, and the line connecting the two positioning points is parallel to the longitudinal symmetry plane of the vehicle.

10. The AVP parking space reservation and automatic parking cooperative control system according to claim 7, characterized in that, It includes a parking space database, and the parking space attributes updated in real time include at least: parking space geometry, ground slope, and charging pile power parameters.

Images