A vehicle acceleration braking mapping table adaptive updating method and device based on recursive least squares and a medium

Abstract

The application provides a method, device and medium for updating acceleration braking mapping table in real time and adaptively during vehicle operation, to solve the problems of update lag, insufficient accuracy and inability to adapt online in the prior art. To achieve the above-mentioned purpose, the application comprises the following steps: S1: determining at least one target cell in the acceleration mapping table or the braking mapping table according to the real-time acquired vehicle speed and pedal position; S2: using the recursive least squares algorithm to update the parameters of the target cell according to the current vehicle state data; S3: maintaining the consistency of the updated mapping table to meet the preset physical constraint condition; S4: accuracy evaluation: the accuracy evaluation module calculates the RMSE before and after updating, and judges whether the effective threshold is met; S5: result output: if the update is valid, the new mapping table is output for vehicle control; if the update is invalid, the original mapping table is maintained, and the update failure log is recorded.

Description

Technical Field

[0001] This invention relates to the field of autonomous vehicle control technology, and in particular to an adaptive update method, apparatus and medium for vehicle acceleration and braking mapping table based on recursive least squares. Background Technology

[0002] In autonomous driving systems, the acceleration and braking map is the core lookup table for longitudinal vehicle control, and its accuracy directly affects the safety and comfort of vehicle operation. Existing map calibration methods are mostly static calibration or offline batch processing, which cannot achieve real-time, adaptive updates during vehicle operation, resulting in limited control accuracy and difficulty in adapting to individual vehicle differences and environmental changes. Summary of the Invention

[0003] The present invention aims to provide a method, apparatus and medium for real-time adaptive updating of the acceleration and braking mapping table during vehicle operation, so as to solve the problems of lagging updates, insufficient accuracy and inability to adapt online in the prior art.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: An adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares includes the following steps: S1: Based on the real-time vehicle speed and pedal position, determine at least one target cell in the acceleration mapping table or braking mapping table; S2: The recursive least squares algorithm is used to update the parameters of the target cell based on the current vehicle status data; S3: Maintain consistency of the updated mapping table to ensure it meets preset physical constraints; S4: Accuracy Assessment: The accuracy assessment module calculates the RMSE before and after the update and determines whether the effective threshold is met; S5: Output Results: If the update is valid, output the new mapping table for vehicle control; if invalid, maintain the original mapping table and record the update failure log.

[0005] Preferably, the recursive least squares algorithm in step S2 uses a forgetting factor, and the weight of historical data decays over time.

[0006] Preferably, step S2 includes at least one of the following update modes: Mode 1: Update cell by cell, performing recursive least squares update independently on each cell in the mapping table; Mode 2: Overall offset update, applying a uniform offset to the entire mapping table for updating; Mode 3: Four-cell surrounding update. Determine the four adjacent cells based on the current speed and pedal value, and simultaneously update them recursively by least squares.

[0007] Preferably, in the four-cell-surround update mode, the weight is calculated based on the normalized distance of the current data point among the four cells, and the update amount is allocated based on the weight.

[0008] Preferably, in the four-cell update pattern, an adaptive forgetting factor is used to prevent covariance matrix saturation.

[0009] Preferably, the physical constraints in step S3 include: For the acceleration map, at the same speed, the larger the pedal value, the greater the acceleration; at the same pedal value, the larger the speed, the smaller the acceleration; For the braking map, at the same speed, the larger the brake pedal value, the greater the deceleration; at the same pedal value, the larger the speed, the smaller the deceleration.

[0010] Preferably, the method further includes a step of evaluating the accuracy of the mapping table before and after the update, and determining whether the update is effective by calculating the root mean square error between the predicted acceleration and the actual acceleration.

[0011] An apparatus for implementing an adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares, comprising: The data input module is used to acquire vehicle speed and pedal position data; the recursive least squares update module is used to update the parameters of the target cell according to the selected update mode; the accuracy evaluation module is used to calculate the prediction error before and after the mapping table is updated and output the evaluation result; the output module is used to send the effectively updated mapping table to the vehicle control system.

[0012] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements an adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares.

[0013] The beneficial effects of this invention are: 1. High real-time performance: Enables online adaptive updates of the mapping table, with a response speed synchronized with the vehicle control cycle (10ms level), and can dynamically adapt to changes in vehicle status and environment; 2. Balancing accuracy and stability: The RLS algorithm with a forgetting factor prioritizes new data and combines three update modes to balance local accuracy and global efficiency. The four-cell mode also ensures the continuity of control characteristics. 3. Physically sound and reliable: The consistent maintenance mechanism strictly follows the vehicle's power output law to avoid control logic conflicts and improve driving safety and comfort; 4. High maintainability: The modular design decouples data acquisition, update calculation, and constraint verification, making it easier to locate faults and debug parameters, and reducing development and maintenance costs. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the update process according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention. Detailed Implementation

[0015] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings.

[0016] As shown in Figure 2, the device of one embodiment of the present invention consists of a data input module, a recursive least squares (RLS) update module, an accuracy evaluation module, and an output module. The modules communicate in real time via a data bus. The data input module continuously collects vehicle operating data, the RLS update module performs core parameter calculations, the accuracy evaluation module verifies the update effect, and finally, the output module sends the effectively updated mapping table to the vehicle control system.

[0017] The data input module acquires vehicle status data and performs preprocessing. Vehicle status data includes: longitudinal speed obtained through the vehicle speed sensor. The accelerator pedal position is obtained through the actuator state. and brake pedal position The steering angle (steer) is obtained through the steering sensor, and then transformed from the map coordinate system to the map coordinate system via coordinate transformation (TF). Obtain the vehicle's pitch angle using a coordinate system.

[0018] The update module uses a recursive least squares algorithm to update the mapping table. The system searches for the corresponding index position in the mapping table based on the current speed and pedal value. Search methods include exact match, nearest neighbor, and lower bound search, each used for different update modes.

[0019] The mapping table update employs a recursive least squares algorithm, supporting three update modes. The algorithm uses a forgetting factor, which causes the weight of historical data to decay over time, giving greater weight to new data. A larger forgetting factor value results in a slower response to new data, but better stability. In the four-cell update mode, an adaptive forgetting factor is used to prevent covariance matrix saturation.

[0020] This invention adopts the logic of "verification first, correction later": Verification logic: Traverse the mapping table by speed dimension and pedal dimension respectively, compare the values ​​of each cell with its adjacent cells, and verify whether the monotonicity constraint is satisfied; Correction method: If a cell violates the constraints, take the average value of the adjacent compliant cells as a reference value, and adjust the cell parameters through linear interpolation to ensure that the correction only affects the local area and does not destroy the overall control characteristics.

[0021] The specific process of one embodiment of the present invention is as follows: After system startup, the factory default mapping table and parameter configuration are loaded, including forgetting factor, update mode priority, RMSE threshold, etc., and the system enters a periodic cyclic update state, for example, with a period of 10ms, synchronized with the vehicle control cycle. Data trigger: When the data input module collects new vehicle status data or reaches the preset update cycle, the update process is triggered; Index location: The index lookup module determines the target cell by selecting the corresponding search method based on the current speed, pedal travel, and preset update mode; RLS Update: The RLS update module calls the algorithm of the corresponding mode, combines the forgetting factor with the current actual acceleration of the vehicle, and completes the update of the target cell parameters; Consistency verification: The consistency maintenance module performs physical constraint checks on the updated mapping table. If there are cells that violate the constraints, local corrections are performed. Accuracy assessment: The accuracy assessment module calculates the RMSE before and after the update to determine whether the effective threshold is met; Output results: If the update is valid, output the new mapping table for vehicle control; if invalid, maintain the original mapping table and record the update failure log.

[0022] The system can calculate the root mean square error of the prediction of the mapping table before and after the update in real time. If the error is significantly reduced after the update, the update is deemed effective and the new table can be used for vehicle control.

[0023] The detailed process of the RLS update algorithm is as follows: Figure 1 As shown, Taking the cell-by-cell update in Mode 1 as an example, the system first reads the current cell state, including offset, covariance, and predicted acceleration. Then, it calculates the design matrix, updates the covariance matrix, calculates the gain coefficient, and calculates the error. Update offset Finally, update the mapping table values. .

[0024] Mode 1 updates cell by cell, updating each cell in the mapping table independently. For each cell, the system calculates the covariance update and gain coefficient, then updates the offset based on the prediction error, and finally adds the offset to the original mapping table value to obtain the updated mapping table value.

[0025] Mode 2 is a global offset update, applying a uniform offset to the entire mapping table. The system calculates the global offset and applies the same offset to all cells; the update formula is as follows. .

[0026] Mode 3 involves updating the four cells surrounding the current data point simultaneously. The system calculates the normalized distance based on the current data point's position among these four cells, then uses the design vector and state vector to calculate the predicted value, performs an RLS update, and finally applies the updated offsets to the four cells. The updated mapping table must satisfy physical constraints to ensure its monotonicity. For the acceleration mapping table, at the same speed, a larger pedal value results in greater acceleration; at the same pedal value, a larger speed results in smaller acceleration. For the braking mapping table, at the same speed, a larger brake pedal value results in greater deceleration; at the same pedal value, a larger speed results in smaller deceleration. If the updated mapping table does not satisfy these constraints, the system performs consistency correction, fine-tuning the mapping table values ​​to ensure monotonicity.

[0027] The following is an example of a four-cell surrounding update pattern according to an embodiment of the present invention.

[0028] Assuming the current vehicle speed is 52 km / h, the accelerator pedal travel is 65%, the speed index interval of the mapping table is 10 km / h (50 km / h, 60 km / h), and the pedal travel index interval is 10% (60%, 70%), then the current data point falls within the area composed of four cells: (50 km / h, 60%), (50 km / h, 70%), (60 km / h, 60%), and (60 km / h, 70%).

[0029] Normalized distance calculation: Lateral (velocity dimension) normalized distance d1=(52-50) / (60-50)=0.2, longitudinal (pedal dimension) normalized distance d2=(65-60) / (70-60)=0.5; Weighting: The weights of the four cells are w1=(1-d1)(1-d2)=0.4, w2=(1-d1)(d2)=0.4, w3=d1(1-d2)=0.1, and w4=d1d2=0.1, with a sum of weights of 1. Synchronous Update: Based on the weight allocation of each cell, the RLS update amount is synchronously updated to update the state vector and covariance matrix of the four cells, ensuring the continuity of local area control characteristics.

[0030] RLS Algorithm Implementation Taking the cell-by-cell update mode as an example, the RLS algorithm iteration process is as follows: Initialization: The initial state vector of each cell is θ0=[0], the initial covariance matrix is ​​P0=αI (α is a constant greater than 1, default is 100), and the forgetting factor is λ=0.97; Data reading: Read the predicted acceleration a_map and state vector of the current cell. With covariance matrix The actual compensated acceleration a_comp collected by the data input module; Error calculation: Used for prediction error; Design matrix construction:

[0031] Based on a single-input single-output control model; Gain coefficient calculation: ; State vector update: ; Covariance matrix update: ; Mapping table update: Update the state vector θ k Convert the acceleration parameters to cell values ​​and write them back to the mapping table.

[0032] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. An adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares, characterized in that, Includes the following steps: S1: Based on the real-time vehicle speed and pedal position, determine at least one target cell in the acceleration mapping table or braking mapping table; S2: The recursive least squares algorithm is used to update the parameters of the target cell based on the current vehicle status data; S3: Maintain consistency of the updated mapping table to ensure it meets preset physical constraints; S4: Accuracy Assessment: The accuracy assessment module calculates the RMSE before and after the update and determines whether the effective threshold is met; S5: Output Results: If the update is valid, output the new mapping table for vehicle control; if invalid, maintain the original mapping table and record the update failure log.

2. The adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in claim 1, characterized in that, The recursive least squares algorithm described in step S2 uses a forgetting factor, where the weight of historical data decays over time.

3. The adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in claim 2, characterized in that, Step S2 includes at least one of the following update modes: Mode 1: Update cell by cell, performing recursive least squares update independently on each cell in the mapping table; Mode 2: Overall offset update, applying a uniform offset to the entire mapping table for updating; Mode 3: Four-cell surrounding update. Determine the four adjacent cells based on the current speed and pedal value, and simultaneously update them recursively by least squares.

4. The adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in claim 3, characterized in that, In the four-cell-around update mode, weights are calculated based on the normalized distance of the current data point among the four cells, and update amounts are allocated based on these weights.

5. The adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in claim 4, characterized in that, In the update pattern around the four cells, an adaptive forgetting factor is employed to prevent covariance matrix saturation.

6. The adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in claim 5, characterized in that, The physical constraints mentioned in step S3 include: For the acceleration map, at the same speed, the larger the pedal value, the greater the acceleration; at the same pedal value, the larger the speed, the smaller the acceleration. For the braking map, at the same speed, the larger the brake pedal value, the greater the deceleration; at the same pedal value, the larger the speed, the smaller the deceleration.

7. The adaptive update method for vehicle acceleration and braking mapping table based on recursive least squares as described in any of claims 1-6, characterized in that, It also includes a step of evaluating the accuracy of the mapping table before and after the update, and determining whether the update is effective by calculating the root mean square error between the predicted acceleration and the actual acceleration.

8. An apparatus for implementing the adaptive update method for a vehicle acceleration and braking mapping table based on recursive least squares as described in any one of claims 1-7. Its features are, It includes: a data input module for acquiring vehicle speed and pedal position data; a recursive least squares update module for updating the parameters of the target cell according to the selected update mode; an accuracy evaluation module for calculating the prediction error before and after the mapping table update and outputting the evaluation result; and an output module for sending the effectively updated mapping table to the vehicle control system.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images