A method and system for intelligent control of a pedal

Abstract

This application provides a method and system for intelligent control of a pedal, relating to the field of vehicle technology. The method includes: acquiring the user identifier of the target user and state information related to pedal control based on a detected target user; inputting the user identifier and state information into a deployed reinforcement learning agent, and determining the values ​​of multiple control parameters corresponding to the user identifier and state information through an established mapping strategy; controlling the pedal accordingly based on the determined values ​​of the multiple control parameters; evaluating the interaction information obtained during the target user's use of the pedal through a predefined reward function; and optimizing the mapping strategy specific to the target user and related to adjusting the control parameters based on the reward evaluation results. The aim is to improve user satisfaction with pedal use.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and specifically to a method and system for intelligent control of pedals. Background Technology

[0002] With the development of automotive technology, more and more car models have higher chassis, but a higher chassis can make it more inconvenient for users to get in and out of the vehicle. Currently, some vehicles are equipped with fixed pedals or manually controlled retractable pedals, but fixed pedals are prone to scraping the ground while the vehicle is in motion, and manually controlled retractable pedals require active operation from the user, which is not intelligent enough. In recent years, automatic pedals based on sensors (such as infrared or ultrasonic) have emerged, but these sensors can only detect distance and cannot accurately recognize user intentions and environmental context, leading to problems such as inaccurate timing of raising and lowering, false triggering, or delays. Summary of the Invention

[0003] In view of this, this application provides a method and system for intelligent control of pedals, aiming to solve or partially solve the problems existing in the prior art.

[0004] The first aspect of this application provides a method for intelligent control of a pedal, the method comprising: Based on the detected target user, obtain the user identifier of the target user, as well as the status information related to pedal control; The user identifier and the state information are input into the deployed reinforcement learning agent, and the values ​​of multiple control parameters corresponding to the user identifier and the state information are determined through the established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters related to the user identifier that adjust the corresponding basic control parameters. Based on the determined values ​​of the plurality of control parameters, the pedal is controlled accordingly; Based on the interaction information obtained during the target user's use of the pedal, the interaction information is evaluated for rewards using a predefined reward function; Based on the reward evaluation results, the mapping strategy that is specific to the target user and related to the adjustment and control parameters is optimized.

[0005] A second aspect of this application provides a system for intelligent control of a pedal, the system comprising: The data acquisition module is used to acquire the user identifier of the detected target user, as well as status information related to pedal control, based on the detected target user. The control parameter determination module is used to input the user identifier and the state information into the deployed reinforcement learning agent, and determine the values ​​of multiple control parameters corresponding to the user identifier and the state information through an established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters corresponding to the user identifier that adjust the values ​​of the corresponding basic control parameters. The control module is used to control the pedal according to the determined values ​​of the plurality of control parameters. The reward evaluation module is used to evaluate the interaction information obtained by the target user during the use of the pedal through a predefined reward function. The control optimization module is used to optimize the strategies related to the adjustment control parameters corresponding to the user identifier in the mapping strategy based on the reward evaluation results.

[0006] The method for intelligent control of pedals provided in this application has the following advantages: The method for intelligent control of pedals provided in this application introduces a reinforcement learning agent and incorporates a user identifier into the information input to the reinforcement learning agent. Based on continuous interaction with the target user during the pedal control process, the method adaptively optimizes the pedal control corresponding to the target user through a predefined reward function. This allows the reinforcement learning agent to execute a personalized pedal control strategy that is more satisfactory to the current target user for different target users. In other words, it controls the pedal with a strategy that better meets the needs of the corresponding target user, thereby improving the satisfaction of each pedal user.

[0007] To achieve this personalized pedal control scheme tailored to different target users and better suited to their needs, this application defines two types of control parameters for the reinforcement learning agent: basic control parameters corresponding to a specific state (e.g., the average speed at which the user approaches the vehicle, triggering pedal deployment when the user is 1 meter away from the vehicle); and adjusted control parameters optimized through a reward function, corresponding to a specific user. For example, the basic control parameter is to trigger pedal deployment when the user is 1 meter away from the vehicle, but after optimization through the reward function, the adjusted control parameter for user a is set by 0.5 meters, resulting in the actual control parameter for user a being 1.5 meters away from the vehicle; while for user b, after optimization through the reward function, the adjusted control parameter is reduced by 0.2 meters, resulting in the actual control parameter for user b being 0.8 meters away from the vehicle. Finally, the final control parameter value is determined based on the basic and adjusted control parameters for a user in a specific state. This design approach allows for continuous, user-specific optimization of pedal control for long-term users. It intelligently controls the pedal in a way that best suits each user's needs, thereby improving user satisfaction. For new users, while individual optimization of pedal control via a reward function is not yet possible, general intelligent control of the pedal can be achieved based on the mapping strategy for basic control parameters in a pre-trained reinforcement learning agent. In subsequent uses, the pedal control can also be optimized for individual user needs using a reward function, continuously improving user satisfaction for the same pedal. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating a method for intelligent control of a pedal according to one embodiment of this application; Figure 2 This is a schematic diagram illustrating a system for intelligent control of a pedal, as shown in one embodiment of this application. Detailed Implementation

[0010] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] refer to Figure 1 , Figure 1 This is a flowchart illustrating a method for intelligent control of a pedal, as shown in one embodiment of this application. Figure 1 As shown, the method includes: Step S01: Based on the detected target user, obtain the user identifier of the target user and the status information related to pedal control.

[0012] In this embodiment, the intelligent control method for pedals provided in this application requires no manual intervention from the user. Instead, it automatically monitors users and controls the pedals to extend and retract based on the monitoring results. Before application, users who need to use the vehicle's pedals for assistance are registered in the vehicle's infotainment system. This allows the system to identify users approaching the vehicle who require this assistance. These users could be the driver, frequent passengers, etc. An optional implementation involves registering the faces of these users in the vehicle's infotainment system and assigning a unique user identifier to each user—meaning any two users have different identifiers. These users registered in the system are referred to as target users. After registration, the vehicle uses a camera to capture images of users approaching it and matches their faces with the pre-registered target user faces to determine if the approaching user is indeed a target user. If the user approaching the vehicle is a target user pre-registered in the system, then it is determined that a target user who needs to use the vehicle's pedals is approaching the vehicle, and subsequent control actions are initiated.

[0013] In this embodiment, an attention mechanism is incorporated into user recognition, allowing the recognition network to focus on areas with prominent human features. During data training of the user recognition network model, this application uses dataset images under different lighting intensities to simulate complex lighting conditions such as nighttime, strong light, and shadows; and uses image datasets under adverse weather conditions such as rain, snow, and fog to simulate weather conditions; and randomly adds occlusions to improve detection robustness under partial occlusion. Furthermore, this application conducts multi-view training, including pedestrian images from various angles such as overhead and eye level. Through these optimizations and limitations, the user recognition network model is ensured to accurately identify the target user from multiple users approaching the vehicle.

[0014] In this embodiment, the core of this application is to adapt pedal control to individual differences and changes in habits over long-term use, so that each user using the same pedal can have a more satisfactory pedal control strategy. Therefore, a reinforcement learning agent (such as a Q-learning agent) trained based on reinforcement learning is introduced, along with a user identifier corresponding to a specific user. This reinforcement learning agent learns and dynamically optimizes key control parameters through continuous interaction with the user during the pedal control process, in order to achieve the highest user satisfaction for each user using the same pedal. That is, the reinforcement learning agent learns and dynamically optimizes key algorithm parameters through continuous interaction with the user and the vehicle to achieve higher user satisfaction.

[0015] In the field of reinforcement learning, the state space is the set of all possible states in a reinforcement learning environment, serving as the basis for the reinforcement learning agent's decision-making. Since the reinforcement learning agent in this application primarily controls pedal deployment based on determining the optimal timing for pedal deployment when the user approaches the vehicle, this application defines the state set input to the reinforcement learning agent as including at least the user's motion state. One optional implementation is to use the user's average speed as the user's approach speed as the user's motion state in the state set.

[0016] Because this application aims to adapt the pedal control of the reinforcement learning agent to individual differences and long-term usage habits for different users, it seeks to provide a more satisfactory pedal control strategy for each user using the same pedal. Therefore, this application defines the state set input to the reinforcement learning agent as including a user identifier. The input user identifier allows the reinforcement learning agent to know which target user is being targeted in determining the corresponding control action and to identify which target user's pedal control optimization is being performed. Thus, after inputting the state set consisting of the user's motion state and the user identifier into the reinforcement learning agent, the agent obtains the values ​​of basic control parameters based on the input motion state, and then obtains adjustable control parameters based on the input user identifier and motion state to adaptively adjust the values ​​of these basic control parameters according to the individual differences of the user (i.e., a specific target representation).

[0017] Based on this, this application defines a set of states. , where the state set s contains This indicates the user identifier obtained through facial recognition (different user identities are used to match different personalized strategies); The average speed at which the detected target user approaches the vehicle. The user identifier and the average speed at which the user approaches the vehicle are the minimum state parameters that the state set in this application must include.

[0018] As mentioned above, the reinforcement learning agent in this application primarily controls the pedal deployment based on determining an optimal timing when the user approaches the vehicle. Therefore, the action set of the reinforcement learning agent includes at least a trigger timing control parameter, and based on the determined value of this trigger timing control parameter, the pedal can be deployed at the corresponding moment.

[0019] Meanwhile, this application aims to adapt the pedal control of the reinforcement learning agent to individual differences and long-term usage habits, so that each user using the same pedal can have a more satisfactory pedal control strategy. Therefore, this application defines the trigger timing control parameters as having two parts. One part is the trigger timing basic control parameter, which is independent of a specific user (i.e., unaffected by user identification) and only related to the specific input state, which does not include the user identification in the state set. For example, in this application, when the state set includes the user's average speed approaching the vehicle and the user identification, the specific state only includes the user's average speed approaching the vehicle; in another subsequent embodiment of this application, when the state set includes the user's average speed approaching the vehicle, user height, vehicle tilt angle, light intensity, and user identification, the specific state includes the user's average speed approaching the vehicle, user height, vehicle tilt angle, and light intensity. The other part is the trigger timing adjustment control parameter, which is mainly related to the user identification and is an adjustment parameter for adaptively adjusting the value of the trigger timing basic control parameter.

[0020] One optional implementation uses the trigger distance as the trigger timing control parameter. That is, the trigger timing for pedal deployment is determined based on the detected distance between the target user and the vehicle. For example, if the trigger distance control parameter is 1 meter, the pedal deployment is triggered when the target user is detected to be 1 meter away from the vehicle. Since the trigger timing control parameter defined in this application includes two parts, when the trigger timing control parameter is a trigger distance control parameter, it includes a basic trigger distance control parameter and a trigger distance adjustment control parameter. For example, user identifier 'a' and the average speed 'b' of the target user corresponding to user identifier 'a' approaching the vehicle are input into a pre-trained reinforcement learning agent that has undergone optimization for a period of time. Based on the learned mapping strategy, the reinforcement learning agent obtains a basic trigger distance control parameter of 1 meter corresponding to the average speed 'b', and a trigger distance adjustment control parameter of 0.2 meters corresponding to user identifier 'a' and the average speed 'b'. Therefore, it is determined that the pedal deployment will be triggered when the target user corresponding to user identifier 'a' is currently 1.2 meters away from the vehicle. For a pre-trained reinforcement learning agent that has undergone optimization over a period of time, regardless of the user's identifier as input, as long as the specific state is the same (e.g., average speed b in this example), the basic control parameter for trigger distance output by the reinforcement learning agent will have the same value (e.g., 1 meter in this example). In the mapping strategy learned by the reinforcement learning agent, the basic control parameter for trigger distance is a universal control variable; its value is not affected by different user identifiers, but only by the specific state. Different user identifiers only affect the trigger distance adjustment control parameter. For example, if the target user with user identifier c approaches the vehicle at an average speed of b, based on the mapping strategy, the basic control parameter for trigger distance corresponding to this average speed b will also be 1 meter. However, the value of the trigger distance adjustment control parameter corresponding to user identifier c and this average speed b will be -0.1 meters. Therefore, it is determined that the target user corresponding to user identifier c triggers pedal deployment when currently 0.9 meters away from the vehicle.

[0021] In each embodiment and implementation of this application, the application targets the same vehicle's pedals used by the same user. Regardless of the number of target users, they all use the same vehicle's pedals, and the reinforcement learning agent is only deployed in the vehicle's infotainment system. For any user using the vehicle's pedals, the reinforcement learning agent will optimize pedal control based on the user's interaction information during pedal use, resulting in a more satisfactory pedal control strategy for that user.

[0022] In this embodiment, a pre-trained reinforcement learning agent that can be directly deployed is pre-trained. This deployed reinforcement learning agent has learned and established a mapping strategy from a predefined state space to a predefined action space. The state space includes all possible combinations of states, and the action space includes all possible combinations of actions (i.e., all possible control actions). That is, for any value of each state parameter in the state set, the deployed reinforcement learning agent can obtain a corresponding value for each action parameter in the action set. It should be noted that since the newly deployed and pre-trained reinforcement learning agent does not yet have interaction information with the user during pedal control, it has not yet performed targeted pedal control optimization for any user. Therefore, the mapping strategy learned and established by the initially deployed reinforcement learning agent does not yet include a strategy for adjusting the trigger distance control parameter. That is, when the initially deployed reinforcement learning agent controls the pedal, the trigger distance adjustment control parameter value for each target user in each state is 0. Only after the user's pedal control is optimized through a predefined reward function during subsequent pedal control processes, based on the interaction information with the user during pedal control, will a strategy for adjusting the trigger distance control parameter be established in the mapping strategy. In other words, only then will the mapping strategy contain a strategy for adjusting the trigger distance control parameter, and the value of the trigger distance adjustment control parameter will be non-zero in the mapping strategy.

[0023] In this embodiment, the subsequent control actions after activation are described. After detecting the target user, the user identifier of the target user that has been pre-recorded is obtained, as well as the status information related to pedal control. The status information related to pedal control includes at least the user's motion state, and in one optional implementation, the average speed at which the user approaches the vehicle.

[0024] Step S02: Input the user identifier and the state information into the deployed reinforcement learning agent, and determine the values ​​of multiple control parameters corresponding to the user identifier and the state information through the established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters related to the user identifier that adjust the corresponding basic control parameters.

[0025] In this embodiment, the user identifier and state information obtained in step S01 are input into a pre-trained reinforcement learning agent that has undergone optimization for a period of time. The mapping strategy established by the reinforcement learning agent determines the values ​​of multiple control parameters corresponding to the user identifier and the state information. These multiple control parameters include basic control parameters related to the state information, and adjustment control parameters, primarily related to the user identifier, used to adjust the corresponding basic control parameters.

[0026] When the basic control parameters include the trigger distance basic control parameter, and the adjustment control parameters include the trigger distance adjustment control parameter, the specific process for determining the values ​​of multiple control parameters based on the mapping strategy is as follows: Based on the state information obtained in step S01, the basic control parameters of the trigger distance corresponding to the state information are queried from the mapping strategy established by the reinforcement learning agent. Then, based on the user identifier and state information obtained in step S01, the trigger distance adjustment control parameters corresponding to both the user identifier and the state information are queried from the mapping strategy established by the reinforcement learning agent.

[0027] Step S03: Control the pedal accordingly based on the determined values ​​of the plurality of control parameters.

[0028] In this embodiment, the pedal is controlled accordingly based on the values ​​of multiple control parameters determined in step S02. When these multiple control parameters include a basic trigger distance control parameter and a trigger distance adjustment control parameter, the final trigger distance control parameter value corresponding to the target user is calculated based on the values ​​of the basic trigger distance control parameter and the trigger distance adjustment control parameter obtained in step S02. Based on this final trigger distance control parameter value, the pedal is deployed when the distance between the target user and the vehicle is equal to the value of this trigger distance control parameter. For example, if the basic trigger distance control parameter obtained in step S02 is 1 meter and the trigger distance adjustment control parameter is 0.2 meters, the final trigger distance control parameter value corresponding to the target user is calculated to be 1.2 meters. Therefore, the pedal is deployed when the target user is 1.2 meters away from the vehicle.

[0029] Step S04: Based on the interaction information obtained during the target user's use of the pedal, evaluate the interaction information using a predefined reward function.

[0030] In this embodiment, after the pedal begins its unfolding action, the entire process of the target user using the pedal is monitored to obtain corresponding interaction information. Then, based on a predefined reward function, the obtained interaction information is evaluated to assess the user's satisfaction throughout the pedal usage. For example, the time interval between the target user stepping on the pedal after it is fully unfolded is determined. If the interval is too long, it is not satisfactory because the triggering timing is inaccurate, and the pedal unfolds too early. The system also determines whether the pedal anti-pinch mechanism is triggered. If the anti-pinch mechanism is triggered, it is not satisfactory because a safety risk has arisen during use.

[0031] Step S05: Based on the reward evaluation results, optimize the mapping strategy that is specific to the target user and related to the adjustment control parameters.

[0032] In this embodiment, after obtaining the corresponding reward evaluation result through the reward evaluation in step S04, the strategy specific to the target user and related to the adjustment control parameters in the mapping strategy is optimized based on the reward evaluation result. That is, the optimization is not of the basic control parameters in the mapping strategy that are only affected by state information, but rather of the adjustment control parameters (such as the trigger distance adjustment parameters mentioned above) that are simultaneously affected by the user identifier and specific state of the target user. For example, previously, for a target user with user identifier 'a' under state information 'b', the trigger distance adjustment control parameter corresponding to user identifier 'a' and state information 'b' was found to be 0.2 meters through the mapping strategy. However, after optimizing the strategy specific to the target user with user identifier 'a' and related to the trigger distance adjustment control parameters in the mapping strategy, the current value of the trigger distance adjustment control parameter corresponding to user identifier 'a' and state information 'b' found through the mapping strategy becomes 0.1 meters. The optimization of the mapping strategy is completed directly on the vehicle during the time period when the pedals are not used.

[0033] The method for intelligent control of pedals provided in this application introduces a reinforcement learning agent and incorporates a user identifier into the information input to the reinforcement learning agent. Based on continuous interaction with the target user during the pedal control process, the method adaptively optimizes the pedal control corresponding to the target user through a predefined reward function. This allows the reinforcement learning agent to execute a personalized pedal control strategy that is more satisfactory to the current target user for different target users. In other words, it controls the pedal with a strategy that better meets the needs of the corresponding target user, thereby improving the satisfaction of each pedal user.

[0034] To achieve this personalized pedal control scheme tailored to different target users and better suited to their needs, this application defines two types of control parameters for the reinforcement learning agent: basic control parameters corresponding to a specific state (e.g., the average speed at which the user approaches the vehicle, triggering pedal deployment when the user is 1 meter away from the vehicle); and adjusted control parameters optimized through a reward function, corresponding to a specific user. For example, the basic control parameter is to trigger pedal deployment when the user is 1 meter away from the vehicle, but after optimization through the reward function, the adjusted control parameter for user a is set by 0.5 meters, resulting in the actual control parameter for user a being 1.5 meters away from the vehicle; while for user b, after optimization through the reward function, the adjusted control parameter is reduced by 0.2 meters, resulting in the actual control parameter for user b being 0.8 meters away from the vehicle. Finally, the final control parameter value is determined based on the basic and adjusted control parameters for a user in a specific state. This design approach allows for continuous, user-specific optimization of pedal control for long-term users. It intelligently controls the pedal in a way that best suits each user's needs, thereby improving user satisfaction. For new users, while individual optimization of pedal control via a reward function is not yet possible, general intelligent control of the pedal can be achieved based on the mapping strategy for basic control parameters in a pre-trained reinforcement learning agent. In subsequent uses, the pedal control can also be optimized for individual user needs using a reward function, continuously improving user satisfaction for the same pedal.

[0035] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. In this method, the basic control parameters include basic pedal height control parameters and basic trigger timing control parameters, and the adjustment control parameters include pedal height adjustment control parameters and trigger timing adjustment control parameters.

[0036] In this application, step S03 may include steps S03_1 to S03_5: Step S03_1: Based on the values ​​of the basic control parameters for triggering the timing corresponding to the first target user and the values ​​of the adjustment control parameters for triggering the timing, control the pedal to unfold. The first target user is the target user who triggers the pedal to unfold first.

[0037] In this embodiment, since users of different heights expect the pedal to be adjusted to different heights when using the pedal, in order to improve user satisfaction with pedal control, the predefined action set of this application may further include pedal height control parameters, which include basic pedal height control parameters and corresponding pedal height adjustment control parameters. The basic pedal height control parameters are independent of the specific user (i.e., unaffected by user identification) and are only related to the specific input state, which does not include the user identification in the state set. The pedal height adjustment control parameters are mainly related to the user identification and are adjustment parameters that adaptively adjust the value of the basic pedal height control parameters to accommodate individual differences.

[0038] Since the value of the pedal height control parameter is based on the target user's height, when the action set also includes the pedal height control parameter, the state set will include the target user's height information. That is, the acquired state information related to pedal control includes the user's average speed approaching the vehicle and the user's height. Because the target user's information is pre-registered in the vehicle's infotainment system, one possible implementation is to synchronously register each target user's height information, which can be directly read when a target user is detected. Alternatively, if the target user's height is not registered in the vehicle's infotainment system, the target user's height can be determined during facial recognition based on an image containing the target user's entire body.

[0039] In this embodiment, when the pedal control supports pedal height adjustment, the vehicle tilt angle affects the pedal height adjustment. For example, if a user boards the vehicle from the right side, and the road is tilted, causing the vehicle to tilt at a certain angle, and the vehicle is tilted to the left, the ground will provide some boarding height for the user. Therefore, in this scenario, the pedal height does not need to be adjusted too low. Therefore, to improve user satisfaction with pedal control, in one optional embodiment, this application can introduce the vehicle tilt angle into the state set, and simultaneously introduce illumination intensity into the state set, representing the confidence level affecting the visual algorithm. That is, in one optional embodiment, the state set of this application may include... In the state set s, This indicates the user's identity obtained through facial recognition; This indicates the height of the detected target user; The vehicle tilt angle is obtained from the IMU inertial navigation sensor; The average speed at which the target user approaches the vehicle; The light intensity represents the level of confidence that influences the visual algorithm.

[0040] In this embodiment, in actual pedal control scenarios, multiple target users who need to use the vehicle's pedals to assist in getting into the car are often detected during a single vehicle use. For example, other target users registered in the system may accompany the user as passengers when the user goes to drive. This application provides an implementation method for this scenario. In this implementation method, for any target user among all currently detected target users, the user identifier and the corresponding pedal control-related state information of the target user are obtained. When the state set includes the target user's user identifier, the target user's height, the vehicle tilt angle, the target user's average speed approaching the vehicle, and the light intensity, the pedal control-related state information of the target user includes the target user's user identifier, the target user's height, the vehicle tilt angle, the target user's average speed approaching the vehicle, and the current light intensity. Then, the target user's user identifier and the corresponding pedal control-related state information are input into a deployed reinforcement learning agent. Through an established mapping strategy, the values ​​of multiple control parameters unique to the target user corresponding to the user identifier and the state information are determined. Using the same implementation method, each currently detected target user can obtain the values ​​of its own multiple control parameters.

[0041] In this embodiment, for a scenario where multiple target users are detected who need to use the vehicle's pedals to assist in getting into the vehicle during a single use, after each detected target user has obtained the values ​​of multiple control parameters corresponding to them, the pedal control process in step S03 adopts another implementation method, as follows: Since each detected target user can obtain values ​​for multiple control parameters, each target user has a basic trigger timing control parameter (such as the basic trigger distance control parameter mentioned above) and a trigger timing adjustment control parameter (such as the trigger distance adjustment control parameter mentioned above). Based on the values ​​of the basic trigger timing control parameter and the trigger timing adjustment control parameter for the same target user, a trigger timing control parameter value corresponding to that target user can be obtained. This trigger timing control parameter value is used to determine when to trigger the pedal deployment. If the trigger timing control parameter is a trigger distance control parameter, then the trigger distance control parameter value corresponding to the target user is used to determine the specific distance between the target user and the vehicle when the pedal deployment is triggered. Through the same implementation method, each detected target user will have its own corresponding trigger distance control parameter value.

[0042] In scenarios where multiple target users approach and board the vehicle, only one pedal deployment trigger is needed. Therefore, as long as the distance between any one target user and the vehicle equals the trigger distance control parameter for that user, the pedal deployment is immediately triggered. This target user is the first one whose pedal deployment is triggered. Even if the distance between the remaining target users and the vehicle meets their respective pedal deployment trigger conditions, the pedal deployment will not be triggered. The pedal deployment trigger condition for a target user is that the distance between the target user and the vehicle during their approach equals the trigger distance control parameter for that user.

[0043] Step S03_2: Based on the values ​​of the basic control parameters for pedal height corresponding to the first target user and the values ​​of the pedal height adjustment control parameters, control the pedal to adjust to a height that matches the first target user.

[0044] In this embodiment, since the first target user is the first to trigger the pedal deployment, the pedal height is preferentially adjusted to match the height of the first target user after deployment. The specific height to be adjusted is determined by the values ​​of the basic control parameters and adjustment control parameters corresponding to the first target user's pedal height. For example, if the basic control parameter for the pedal height is 0.15 meters and the adjustment control parameter is 0.05 meters, the pedal height matching the first target user is adjusted to 0.2 meters above the ground. In this application, all basic control parameters are control parameters only related to the specific state and are not affected by the user identifier. However, all adjustment control parameters are mainly control parameters related to the user identifier; they are used to adjust the corresponding category of basic control parameters and are affected by the user identifier. Even if multiple target users have the same specific state, because their user identifiers are different, the adjustment control parameters corresponding to these target users in the same state will also be different.

[0045] Step S03_3: When the previous target user finishes using the pedal, determine the distance between the next target user and the vehicle.

[0046] In this embodiment, for a scenario where multiple target users are detected needing to use the vehicle's pedals to assist with boarding during a single vehicle use, the distance between the vehicle and the nearest target user is determined after the previous target user finishes using the pedals. One possible implementation for determining whether the previous target user has finished using the pedals is to determine if the target user has been seated on the seat corresponding to the boarding door; if so, the pedal use is considered complete. Alternatively, the pedal pressure sensor can be used to detect if the target user applied pressure to the pedal and then, if no further pressure is applied, the pedal use is considered complete.

[0047] Step S03_4: When the distance is greater than or equal to the set threshold, based on the value of the pedal height basic control parameter and the value of the pedal height adjustment control parameter corresponding to the target user, control the pedal to be adjusted to a height that matches the target user.

[0048] In this embodiment, if the distance between the next target user and the vehicle is greater than or equal to a preset threshold, it is determined that the next target user is still some distance away from the vehicle. At this time, based on the values ​​of the basic control parameters of the pedal height and the values ​​of the pedal height adjustment control parameters corresponding to the next target user, the pedal height matching the next target user is determined, and then the pedal is adjusted to match the pedal height matching the next target user, so as to ensure that the next target user has better satisfaction with using the pedal.

[0049] Step S03_5: If the distance is less than the set threshold, then intelligent control of the pedal is disabled.

[0050] In this embodiment, if the distance between the next target user and the vehicle is less than a preset threshold, it is determined that the next target user is already quite close to the vehicle. If the next target user has already sensed the approximate height of the pedal and stepped on it, but the pedal then makes an adaptive height adjustment, it could lead to the risk of the next target user missing their foot or losing balance. Therefore, if the distance between the next target user and the vehicle is less than the preset threshold, the pedal height will not be adjusted to match the next target user's height to avoid safety risks when the next target user uses the pedal. Although this may reduce user satisfaction to some extent, it effectively ensures safety, that is, maximizing user satisfaction with pedal use without sacrificing safety.

[0051] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. In this method for intelligent control of the pedal, step S04 may include: determining whether the pedal height has been adjusted for the current target user; if the pedal height has been adjusted for the current target user, evaluating the interaction information obtained during the target user's use of the pedal using a predefined reward function; if the pedal height has not been adjusted for the current target user, removing the interaction information during the target user's use of the pedal.

[0052] In this embodiment, regarding one implementation of step S03, which includes steps S03_1 to S03_5, among the multiple target users detected simultaneously, some may not have adjusted the pedal height for safety reasons. Therefore, the pedal control process for these target users sacrifices comfort to some extent, thus affecting user satisfaction. Consequently, no strategy optimization to improve user satisfaction is performed for such target users, because optimal pedal control was not originally considered for them, and the corresponding interaction information is not valuable for strategy optimization.

[0053] Specifically, before acquiring the interaction information of the current target user during pedal use, it is determined whether a pedal height adjustment matching the target user has been performed. If a matching pedal height adjustment has been performed for the current target user, the interaction information of the current target user during pedal use is acquired, and the interaction information is evaluated using a predefined reward function to optimize the mapping strategy specific to the current target user and related to the adjustment control parameters. If, for safety reasons, no pedal height adjustment has been performed for the current target user, the acquired interaction information of the current target user is discarded and not used for subsequent control strategy optimization.

[0054] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligently controlling a pedal. This method for intelligently controlling a pedal further includes: determining the priority of each target user among a plurality of detected target users.

[0055] In this embodiment, since pedal assistance is particularly important for the elderly and children, this application can also set corresponding priorities for user identifiers, with higher priorities for the elderly and children. For example, the user identifier for an elderly person in the family is set to the highest priority, and the user identifier for a child in the family is set to the second highest priority. After detecting multiple target users during a single vehicle use, the priority of each target user's user identifier is determined.

[0056] In this application, the method for intelligent control of a pedal provided in this application further includes determining the priority of each target user among multiple detected target users. Step S02 can be implemented in another way, including: inputting the user identifier and corresponding state information of the target user with the highest priority into the deployed reinforcement learning agent, and determining the values ​​of multiple control parameters corresponding to the user identifier and the state information through the established mapping strategy.

[0057] In this embodiment, after detecting multiple target users during a single vehicle use and determining the priority of each target user's user identifier, the user identifier of the highest priority target user and the corresponding state information of that target user are input into the deployed reinforcement learning agent. A mapping strategy is then established to determine the values ​​of multiple control parameters corresponding to that user identifier and state information. The pedals are then controlled only using the values ​​of these control parameters corresponding to that target user, prioritizing service to the highest priority target user.

[0058] This application provides another optional implementation method, applicable to scenarios where multiple target users are detected during a single vehicle use. In this implementation, priority is still assigned to target users, but priority is not considered when determining the control parameter values. Instead, for each currently detected target user, the values ​​of its corresponding multiple control parameters are processed. In this implementation, the method for triggering pedal deployment remains the same: as long as the distance between any target user and the vehicle equals the trigger distance control parameter value for that target user, the pedal is immediately deployed; this target user is the first target user to trigger pedal deployment. However, in this implementation, the initial pedal height is not matched with the first target user, but with the highest priority target user among all target users. That is, after the pedal is deployed, the pedal height is preferentially adjusted to match the height of the highest priority target user. The specific height to be adjusted is determined by the values ​​of the basic pedal height control parameter and the pedal height adjustment control parameter corresponding to the highest priority target user. Only after the highest priority target user finishes using the pedal can all subsequent target users perform the following steps: when the previous target user finishes using the pedal, determine the distance between the next target user and the vehicle; if the distance is greater than or equal to a set threshold, based on the values ​​of the pedal height basic control parameters and pedal height adjustment control parameters corresponding to the next target user, control the pedal to adjust to a height matching the next target user; if the distance is less than the set threshold, then disable intelligent control of the pedal.

[0059] In this embodiment, the action set of this application may further include a retraction delay control parameter, which includes a basic retraction delay control parameter and a retraction delay adjustment control parameter. The retraction delay control parameter is determined based on the basic and adjustment control parameters of a user in a specific state. The retraction delay control parameter can be selected as the delay period after all doors are detected as closed before retracting the pedal. For example, if user a in a specific state b has a basic retraction delay control parameter of 1 second and an adjustment control parameter of 0.5 seconds, then the final retraction delay control parameter is determined to be 1.5 seconds, meaning the pedal is retracted 1.5 seconds after all doors are detected as closed. When multiple target users are detected during a single vehicle use, the pedal retraction is controlled by the largest retraction delay control parameter value corresponding to all target users.

[0060] In one alternative implementation, the set of actions is as follows:

[0061] In action set a, This indicates the basic control parameters for triggering the pedal (this value determines the intelligent triggering timing of the pedal to a certain extent; for example, based on the user's current state, the triggering distance is determined to be 1m). This indicates the trigger timing adjustment control parameter (used to adjust the value of the basic trigger timing control parameter according to individual differences, and its value, together with the value of the basic trigger timing control parameter, determines the final intelligent trigger timing of the pedal. For example, based on the current user state, the trigger distance is determined to be 1m. Based on the mapping strategy optimized by the reward function, the value of the trigger timing adjustment control parameter corresponding to the user state and the user's identifier in the mapping strategy is 0.2m. Then, the pedal will be triggered when the user is 1.2m away from the vehicle). This indicates the basic control parameter for the retraction delay (this value determines the intelligent retraction timing of the pedal to a certain extent; for example, based on the user's current state, the retraction delay is determined to be 1 second after the door is closed). This refers to the retraction delay adjustment control parameter (used to adjust the value of the basic retraction delay control parameter according to individual differences, and its value, together with the value of the basic retraction delay control parameter, determines the final intelligent retraction timing of the pedal. For example, based on the current user state, the retraction delay is determined to be 1 second. Based on the mapping strategy optimized by the reward function, the retraction delay adjustment control parameter corresponding to the current state and the user's identifier in this mapping strategy is 0.5 seconds. Then, the pedal will retract 1.5 seconds after the door closes). This indicates the basic control parameter for pedal height (this value determines the height of the pedal after it is displayed to a certain extent; for example, based on the user's current state, the pedal height is determined to be 0.15 meters above the ground). This represents the pedal height adjustment control parameter (used to adjust the value of the basic pedal height control parameter according to individual differences, and its value, together with the value of the basic pedal height control parameter, determines the final pedal height. For example, based on the user's current state, the pedal height is determined to be 0.15 meters from the ground. Based on the mapping strategy optimized by the reward function, the value of the pedal height adjustment control parameter corresponding to this state and the user's identifier in this mapping strategy is 0.05 meters. Then, the final pedal height after deployment is 0.2 meters from the ground). The action set represents the adaptive adjustment of the core control parameters of the pedal control system, thereby accurately implementing personalized strategies for different users.

[0062] In this embodiment, the main application scenario of the intelligent pedal control method provided in this application is: the pedal unfolds when a user approaches the vehicle from a position far from the vehicle to get in. The above is only the main application scenario; this method can also be applied to the following scenarios: 1. For the deployment of the foot pedal when a user exits the vehicle, if there is a highest priority target user, the pedal height is adjusted to match that user after deployment. If there is no priority concept for target users in the vehicle, the average height of all foot pedals corresponding to all target users in the vehicle is taken, and the pedal height is adjusted to match this average after deployment. The foot pedal deployment in this scenario can be triggered by detecting a user opening the vehicle door.

[0063] 2. For retracting the pedal after all target users have boarded the vehicle, the pedal can be retracted after a certain delay after detecting that all doors are closed.

[0064] 3. For retracting the pedal after all target users have disembarked, the pedal can be retracted after a certain delay after detecting that all users have disembarked, or after detecting that all doors in the scenario have closed, the pedal can be retracted after a certain delay.

[0065] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. In this method for intelligent control of the pedal, the reward function is composed of at least one of a plurality of reward sub-functions, including a timing reward sub-function, a comfort reward sub-function, a safety reward sub-function, and an efficiency reward sub-function.

[0066] In this embodiment, the predefined reward function of this application evaluates user satisfaction with the intelligent control pedal from at least one of the following aspects: control timing, comfort, safety, and efficiency. For example, if the user arrives at the vehicle before the pedal is deployed, or if the user arrives at the vehicle long after the pedal is fully deployed, the control timing is poor; if the pedal's anti-pinch mechanism is frequently triggered, the safety is poor; if the user stumbles while using the pedal, the comfort is poor; if the cumulative time from when the pedal is fully deployed to when it is fully retracted is much longer than the cumulative time from when the user begins climbing the steps to when the user is fully on the vehicle, the efficiency is poor. For each evaluation aspect, this application defines a corresponding reward subfunction, including a timing reward subfunction, a comfort reward subfunction, a safety reward subfunction, and an efficiency reward subfunction.

[0067] In this embodiment, the reward function is a core design component of the reinforcement learning agent. It defines a behavioral policy for the agent, provides gradient directions for optimizing the pedal control policy, influences the agent's behavioral preferences through reward allocation, and reflects safety and other constraints through a penalty mechanism. When the predefined reward function in this application consists of multiple reward sub-functions, it can be a weighted sum of these sub-functions. Here, we take an example where the predefined reward function includes these four reward sub-functions, as shown below:

[0068] in, This indicates the timing reward subfunction. If the user steps on the pedal within the ideal time window after the pedal is deployed, a positive reward is given. If the user is ready to get on the vehicle but the pedal is not yet ready or is hastily deployed, a negative reward is given. The comfort reward subfunction detects whether the user gets into the car smoothly and without any bumps through in-vehicle sensors. It can also identify whether the user has positive / negative evaluative voices about the pedal system by obtaining microphone permission, and obtain a total value by combining ergonomics. If the smoothness value is high, a positive reward is given according to the proportion of the smoothness value; if the smoothness value is low, a negative reward is given according to the proportion of the smoothness value. The safety reward subfunction is given a negative reward if the anti-pinch mechanism of the pedal system is triggered due to incorrect timing of pedal deployment or incorrect personnel feature matching, posing a safety threat to the user. If no safety-related threat scenario occurs for a long time, a small positive reward is given. Represents the efficiency-based reward subfunction; The weighting coefficients, k, represent weights used to balance the importance of different rewards, with values ​​of 1, 2, 3, and 4. This reward function comprehensively considers factors such as the timing of acceleration and deceleration, user comfort, user safety, and system efficiency, enabling the reinforcement learning agent to adaptively control the pedals according to different users, effectively improving the user experience.

[0069] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. In this method for intelligent control of a pedal, when the reward function includes a timing reward sub-function, the interaction information includes first interaction information related to the timing of pedal deployment during the target user's use of the pedal. The first interaction information includes the time interval between the moment the pedal is fully deployed and the moment the target user steps onto the pedal. The timing reward sub-function is used to evaluate timing rewards based on the first interaction information; When the reward function includes a safety reward subfunction, the interaction information includes second interaction information related to safety during the target user's use of the pedal. The second interaction information includes the triggering information of the pedal anti-pinch mechanism and the safe running time information. The security reward sub-function is used to perform a security reward evaluation based on the second interaction information; When the reward function includes an efficiency reward sub-function, the interaction information includes third interaction information related to efficiency during the target user's use of the pedal. The third interaction information includes the cumulative time information from when the pedal is fully extended to when the pedal is fully retracted, and the cumulative time information from when the user starts climbing the steps to when the user is fully on the vehicle during the cumulative usage. The efficiency reward subfunction is used to evaluate efficiency rewards based on third interaction information.

[0070] In this embodiment, when the predefined reward function includes a timing reward subfunction, the acquired target user interaction information includes first interaction information related to the timing of pedal deployment during the target user's use of the pedal. This first interaction information includes the time interval between the moment the pedal is fully deployed and the moment the target user steps onto the pedal. This first interaction information is then provided to the timing reward subfunction for reward evaluation based on the trigger timing angle.

[0071] One possible implementation of the timing reward subfunction for evaluating the triggering timing based on the first interaction information is as follows: Define the time interval between the moment the pedal is fully deployed and the moment the target user steps on the pedal as . ; when If the pedal is not yet deployed or not fully deployed when the user arrives, it will cause a bad user experience. In this case, a negative reward value is given at the triggering time angle, such as a reward value of -2. when When the triggering timing is within the ideal time window, a positive reward value is given at the triggering angle. At the same time, the reward value increases as the linear deviation decreases, with a maximum value of 1. when At this point, although it is not the optimal solution, it is still within an acceptable range. Therefore, we uniformly set its reward value to a small positive value, such as 0.2. when If the pedal deploys too early, it wastes energy and is likely an incorrect decision about the user's intention. In this case, a negative reward value, such as -0.5, should be given.

[0072] In this implementation of the trigger timing evaluation, the expression for the timing reward subfunction is defined as follows:

[0073] In this embodiment, when the predefined reward function includes a safety reward subfunction, the acquired target user interaction information includes second interaction information related to safety during the target user's use of the pedal. This second interaction information includes the triggering information of the pedal anti-pinch mechanism and the safe running time information. This second interaction information will be provided to the safety reward subfunction for a reward evaluation from a safety perspective.

[0074] One possible implementation of the security reward subfunction based on the second interaction information for security assessment is as follows: The expression for the security reward subfunction is defined as follows:

[0075]

[0076] in, The safety penalty for pinching injury has an initial value of 0. Each time the anti-pinch mechanism is triggered, its value is reduced by 2. Represents a reward for long-term safe operation, awarded for each accumulated safe operating time. If the anti-pinch function is not triggered for 100 consecutive hours, a reward value of 0.5 will be given.

[0077] In this embodiment, when the predefined reward function includes an efficiency-based reward sub-function, the acquired target user interaction information includes third-level interaction information related to efficiency during the target user's use of the pedals. This third-level interaction information includes the cumulative time information from when the pedals are fully extended to when they are fully retracted, and the cumulative time information from when the user begins the step-climbing action to when the user is fully on the vehicle. This third-level interaction information will be provided to the efficiency-based reward sub-function for efficiency-based reward evaluation.

[0078] One possible implementation of the efficiency reward subfunction based on third-party interaction information for efficiency evaluation is as follows: The expression for the efficiency reward subfunction is defined as follows:

[0079] in, It is the cumulative time from when the pedal is fully extended to when the pedal is fully retracted during cumulative use; This refers to the cumulative time from when the user begins climbing the stairs to when the user is fully seated on the vehicle during the cumulative usage process. The larger the value, the more precise the timing of pedal execution, and the higher the efficiency of the intelligent pedal lifting system.

[0080] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. In this method for intelligent control of the pedal, when the reward function includes a comfort reward sub-function, based on the acquired corresponding interaction information, the comfort reward sub-function performs a comfort reward evaluation from a comfort perspective, including: Acquire fourth interactive information related to comfort during the target user's use of the pedal. The fourth interactive information includes subjective comfort interactive information and objective comfort interactive information. The subjective comfort interactive information includes the triggering information of the pedal anti-pinch mechanism and / or the target user's evaluation information during the use of the pedal and / or the target user's stumbling state information. The objective comfort interactive information includes the target user's knee angle range information and / or the hip angle range information. Based on the subjective comfort interaction information, the subjective comfort reward corresponding to the subjective comfort is determined through the subjective comfort reward subfunction in the comfort reward subfunction; Based on the objective comfort interaction information, the objective comfort reward corresponding to the objective comfort is determined through the objective comfort reward sub-function in the comfort reward sub-function; Based on the subjective comfort reward and the objective comfort reward, a comfort reward corresponding to the target user's comfort level is determined.

[0081] In this embodiment, regarding the reward evaluation from a comfort perspective, this application, in order to conduct a more accurate comfort evaluation, will assess comfort from both subjective and objective comfort perspectives, as detailed below: First, the fourth comfort-related interaction information related to the target user's use of the pedal is acquired. This fourth interaction information includes subjective comfort interaction information and objective comfort interaction information. The subjective comfort interaction information includes the triggering information of the pedal anti-pinch mechanism, and / or, the target user's evaluation information during pedal use, and / or, the target user's stumbling state information. The objective comfort interaction information includes the target user's knee angle range information, and / or, hip angle range information.

[0082] This document describes one embodiment of comfort reward assessment. In this embodiment, subjective comfort interaction information includes triggering information of the pedal anti-pinch mechanism, evaluation information of the target user during pedal use, and information on the target user's stumbling state. Simultaneously, objective comfort interaction information includes the target user's knee angle range and hip angle range. This embodiment will be described below.

[0083] Based on subjective comfort interaction information, the subjective comfort reward value corresponding to the subjective comfort level is determined through the subjective comfort reward subfunction within the comfort reward subfunction. The subjective comfort reward subfunction primarily evaluates the subjective comfort reward from the following perspectives: 1. If the pedal anti-pinch mechanism is triggered once, or if a negative evaluation from the target user is detected by the vehicle microphone, a corresponding negative reward value will be given in terms of subjective comfort reward; 2. Determine whether the target user experiences any stumbling or bumping while using the pedal. If such stumbling or bumping occurs, provide a corresponding negative reward value in the visual comfort reward.

[0084] In this embodiment, one possible implementation method for determining whether the target user trips or stumbles while using the pedal is as follows: The presence of a tripping hazard is determined based on the characteristics of the pressure sensor data. If the pressure sensor data shows a pressure center trajectory, and the pressure center moves smoothly along the user's foot's path from the door sill to the seat, then there is no tripping hazard. Conversely, if the pressure center's movement is interrupted or jerky along the same path, then a tripping hazard is confirmed.

[0085] If the pressure sensor data is characterized by pressure peak and duration, and a local pressure suddenly increases and then releases, then a bump is confirmed.

[0086] If the pressure sensor data is characterized by a rate of change in pressure distribution, and the pressure distribution changes continuously, then there are no obstructions. If the pressure distribution changes drastically, then there are obstructions.

[0087] Based on objective comfort interaction information, the value of the objective comfort reward corresponding to the objective comfort level is determined through the objective comfort reward subfunction within the comfort reward subfunction. The objective comfort reward subfunction primarily evaluates the objective comfort reward from the following perspectives: This application identifies the range of knee angles within a comfortable range when climbing hills or stairs. The hip joint is within a comfortable range of angles. .

[0088] Therefore, when the actual angle of the knee is within the corresponding comfort range mentioned above, a larger positive reward value is given in terms of objective comfort reward; when the actual angle of the knee is not within the corresponding comfort range mentioned above, a smaller positive reward value is given in terms of objective comfort reward.

[0089] When the actual angle of the hip bone is within the corresponding comfort range mentioned above, a larger positive reward value is given in terms of objective comfort reward; when the actual angle of the hip bone is not within the corresponding comfort range mentioned above, a smaller positive reward value is given in terms of objective comfort reward.

[0090] In this implementation of the comfort assessment, the expression for the comfort reward sub-function is defined as follows:

[0091]

[0092]

[0093]

[0094] in, This represents the reward subfunction for subjective comfort. This represents the objective comfort reward subfunction; the range of angles within which the knee is in a comfortable position. The angle range of the hip bone within a comfortable range When the actual angle of the knee is... exist If the angle is within the specified range, a positive reward of 1 is given; otherwise, a positive reward of 0.3 is given. exist If the value is within the range, give it a positive reward of 1; otherwise, give it a positive reward of 0.3. and Together, they form the comfort reward. .

[0095] In conjunction with the above embodiments, in one implementation, this application also provides a method for intelligent control of a pedal. This method for intelligent control of a pedal further includes: Step S001: Determine whether the target user gets on the vehicle within the preset time after the pedals are deployed.

[0096] In this embodiment, under the intelligent automatic control scenario of the pedal, there is a special false triggering scenario: when the target user merely approaches the vehicle without intending to get in, the pedal deployment may be falsely triggered. To solve this problem, this application provides the following implementation method: Determine whether the target user will board the vehicle within a preset time after the pedals are deployed. This preset time can be set according to the actual scenario, such as 20 seconds.

[0097] Step S002: If the target user does not get on the vehicle within a preset time after the pedal is deployed, the trajectory of the target user approaching the vehicle is determined to be a non-entry intention behavior trajectory.

[0098] In this embodiment, if the target user does not board the vehicle within a preset time after the pedal is deployed, it is determined that the target user is approaching the vehicle without the intention to board. At this time, the trajectory of the target user's approach to the vehicle is obtained and the trajectory is determined to be the target user's non-boarding intention behavior trajectory.

[0099] Step S003: Based on the acquired non-boarding intention behavior trajectories of the target user, construct the non-boarding intention standard behavior trajectory of the target user.

[0100] In this embodiment, after obtaining a certain number of non-boarding intention behavior trajectories of the same target user, a standard non-boarding intention behavior trajectory unique to that target user is constructed based on all such trajectories. This constructed standard non-boarding intention behavior trajectory unique to that target user will be used for subsequent determination of the target user's boarding intention.

[0101] Step S004: If the target user is detected, perform behavioral trajectory analysis on the target user based on the target user's non-boarding intention standard behavioral trajectory.

[0102] In this embodiment, when the target user is subsequently detected approaching the vehicle again, the trajectory of the target user is monitored, and the trajectory is compared with the target user's standard behavioral trajectory of non-entry intention.

[0103] Step S005: If the target user's current behavior trajectory is not a vehicle boarding intention behavior trajectory, disable intelligent control of the pedals.

[0104] In this embodiment, if the comparison result obtained through step S004 indicates that the current behavior trajectory of the target user is not a behavior trajectory indicating an intention to get on the vehicle, then the intelligent control of the pedals will not be activated to serve the target user.

[0105] Step S006: If the target user's current behavior trajectory is the intention to get on the vehicle, perform intelligent control of the pedals.

[0106] In this embodiment, if the comparison result obtained through step S004 indicates that the current behavior trajectory of the target user is not a non-boarding intention behavior trajectory, then the intelligent control of the pedal is activated to serve the target user. The specific intelligent control process is the same as the above implementation method, and will not be repeated here.

[0107] The method for intelligent control of the pedal provided in this application can bring at least the following beneficial effects: Adaptive interactive control: Visual data is analyzed by signal processing equipment to generate drive signals, which control the lifting and lowering response of the pedal (such as adjusting the height via an electric telescopic rod). This closed-loop control enables dynamic adjustment of the pedal, such as automatically matching the pedal height according to the user's position, improving the naturalness of the interaction and safety.

[0108] Multimodal data fusion and analysis: The system synchronously records user physiological data, spatial location, temporal information, and task performance parameters. Combined with visually captured motion data, it can comprehensively evaluate the interaction effect and provide a basis for algorithm optimization.

[0109] Enhanced safety: By visually detecting obstacles or abnormal conditions (such as user imbalance), the system can trigger the braking mechanism to brake immediately.

[0110] Human-machine collaboration efficiency is improved, human intervention is reduced, and the sense of intelligence is enhanced.

[0111] Based on the same inventive concept, this application provides a system for intelligent control of a pedal, such as... Figure 2 As shown, the system 200 includes: The data acquisition module 201 is used to acquire the user identifier of the target user and the status information related to pedal control based on the detected target user. The control parameter determination module 202 is used to input the user identifier and the state information into the deployed reinforcement learning agent, and determine the values ​​of multiple control parameters corresponding to the user identifier and the state information through the established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters corresponding to the user identifier to adjust the values ​​of the corresponding basic control parameters. The control module 203 is used to control the pedal according to the determined values ​​of the plurality of control parameters. The reward evaluation module 204 is used to evaluate the interaction information obtained by the target user during the use of the pedal through a predefined reward function. The control optimization module 205 is used to optimize the strategy related to the adjustment control parameters corresponding to the user identifier in the mapping strategy based on the reward evaluation results.

[0112] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily necessary for the embodiments of this application. As for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple; relevant details can be found in the descriptions of the method embodiments.

[0113] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0114] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0115] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] Although preferred embodiments of the present application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present application.

[0119] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0120] The above provides a detailed description of the method and system for intelligent control of pedals provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligent control of a pedal, characterized in that, The method includes: Based on the detected target user, obtain the user identifier of the target user, as well as the status information related to pedal control; The user identifier and the state information are input into the deployed reinforcement learning agent, and the values ​​of multiple control parameters corresponding to the user identifier and the state information are determined through the established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters related to the user identifier that adjust the corresponding basic control parameters. Based on the determined values ​​of the plurality of control parameters, the pedal is controlled accordingly; Based on the interaction information obtained during the target user's use of the pedal, the interaction information is evaluated for rewards using a predefined reward function; Based on the reward evaluation results, the mapping strategy that is specific to the target user and related to the adjustment and control parameters is optimized.

2. The method for intelligent control of a pedal according to claim 1, characterized in that, The basic control parameters include the pedal height basic control parameters and the trigger timing basic control parameters; the adjustment control parameters include the pedal height adjustment control parameters and the trigger timing adjustment control parameters. The step of controlling the pedal according to the determined values ​​of the plurality of control parameters includes: Based on the values ​​of the basic control parameters for triggering the timing corresponding to the first target user and the values ​​of the adjustment control parameters for triggering the timing, the pedal is controlled to unfold, and the first target user is the target user who triggers the pedal to unfold first. Based on the values ​​of the basic control parameters for pedal height corresponding to the first target user and the values ​​of the pedal height adjustment control parameters, the pedal is controlled to be adjusted to a height that matches the first target user. When the previous target user finishes using the pedal, determine the distance between the next target user and the vehicle; When the distance is greater than or equal to a set threshold, the pedal is adjusted to a height that matches the target user based on the values ​​of the pedal height basic control parameters and the pedal height adjustment control parameters corresponding to the target user. If the distance is less than a set threshold, intelligent control of the pedal is disabled.

3. The method for intelligent control of a pedal according to claim 1, characterized in that, Based on the interaction information obtained during the target user's use of the pedal, a reward evaluation is performed on the interaction information using a predefined reward function, including: Determine whether pedal height adjustment is available for the current target user; If the pedal height is adjusted for the target user, the interaction information obtained during the target user's use of the pedal is evaluated using a predefined reward function. If the target user does not adjust the pedal height, the interaction information during the target user's use of the pedal is removed.

4. The method for intelligent control of a pedal according to claim 1, characterized in that, The method further includes: Determine the priority of each target user among the detected multiple target users; The reinforcement learning agent, which inputs the user identifier and the state information into the deployed system, determines the values ​​of multiple control parameters corresponding to the user identifier and the state information through an established mapping strategy, including: The user identifier and corresponding state information of the highest priority target user are input into the deployed reinforcement learning agent, and the values ​​of multiple control parameters corresponding to the user identifier and the state information are determined through the established mapping strategy.

5. The method for intelligent control of a pedal according to claim 1, characterized in that, The reward function is composed of at least one of the following reward sub-functions, which include a timing reward sub-function, a comfort reward sub-function, a safety reward sub-function, and an efficiency reward sub-function.

6. The method for intelligent control of a pedal according to claim 5, characterized in that, When the reward function includes a timing reward subfunction, the interaction information includes first interaction information related to the timing of pedal deployment during the target user's use of the pedal. The first interaction information includes the time interval between the moment the pedal is fully deployed and the moment the target user steps onto the pedal. The timing reward sub-function is used to evaluate timing rewards based on the first interaction information; When the reward function includes a safety reward subfunction, the interaction information includes second interaction information related to safety during the target user's use of the pedal. The second interaction information includes the triggering information of the pedal anti-pinch mechanism and the safe running time information. The security reward sub-function is used to perform a security reward evaluation based on the second interaction information; When the reward function includes an efficiency reward sub-function, the interaction information includes third interaction information related to efficiency during the target user's use of the pedal. The third interaction information includes the cumulative time information from when the pedal is fully extended to when the pedal is fully retracted, and the cumulative time information from when the user starts climbing the steps to when the user is fully on the vehicle during the cumulative usage. The efficiency reward subfunction is used to evaluate efficiency rewards based on third interaction information.

7. The method for intelligent control of a pedal according to claim 5, characterized in that, When the reward function includes a comfort reward sub-function, based on the acquired corresponding interaction information, the comfort reward sub-function evaluates the comfort reward from a comfort perspective, including: Acquire fourth interactive information related to comfort during the target user's use of the pedal. The fourth interactive information includes subjective comfort interactive information and objective comfort interactive information. The subjective comfort interactive information includes the triggering information of the pedal anti-pinch mechanism and / or the target user's evaluation information during the use of the pedal and / or the target user's stumbling state information. The objective comfort interactive information includes the target user's knee angle range information and / or the hip angle range information. Based on the subjective comfort interaction information, the subjective comfort reward corresponding to the subjective comfort is determined through the subjective comfort reward subfunction in the comfort reward subfunction; Based on the objective comfort interaction information, the objective comfort reward corresponding to the objective comfort is determined through the objective comfort reward sub-function in the comfort reward sub-function; Based on the subjective comfort reward and the objective comfort reward, a comfort reward corresponding to the target user's comfort level is determined.

8. The method for intelligent control of a pedal according to claim 1, characterized in that, The method further includes: Determine whether the target user boards the vehicle within a preset time after the pedals are deployed; If the target user does not get on the vehicle within a preset time after the pedal is deployed, the trajectory of the target user approaching the vehicle is determined to be a non-intent to get on the vehicle trajectory. Based on the acquired non-boarding intention behavior trajectories of the target user, construct the target user's non-boarding intention standard behavior trajectory; Upon detecting the target user, behavioral trajectory analysis is performed on the target user based on the target user's non-boarding intention standard behavioral trajectory. If the target user's current behavior trajectory is not a behavior trajectory indicating an intention to board the vehicle, intelligent control of the pedals shall be prohibited. If the target user's current behavior trajectory is the same as the intention to get on the vehicle, intelligent control of the pedals is executed.

9. A method for intelligent control of a pedal according to claim 2, characterized in that, The status information includes the target user's height, the target user's average speed as they approach the vehicle, the light intensity, and the vehicle's tilt angle; the basic control parameters include basic retraction delay control parameters, and the adjustment control parameters include retraction delay adjustment control parameters.

10. A system for intelligently raising and lowering a pedal based on an intelligent agent, characterized in that, The system includes: The data acquisition module is used to acquire the user identifier of the detected target user, as well as status information related to pedal control, based on the detected target user. The control parameter determination module is used to input the user identifier and the state information into the deployed reinforcement learning agent, and determine the values ​​of multiple control parameters corresponding to the user identifier and the state information through an established mapping strategy. The multiple control parameters include basic control parameters related to the state information and adjustment control parameters corresponding to the user identifier that adjust the values ​​of the corresponding basic control parameters. The control module is used to control the pedal according to the determined values ​​of the plurality of control parameters. The reward evaluation module is used to evaluate the interaction information obtained by the target user during the use of the pedal through a predefined reward function. The control optimization module is used to optimize the strategies related to the adjustment control parameters corresponding to the user identifier in the mapping strategy based on the reward evaluation results.

Images