Optimization of vehicle performance supporting vehicle control

Abstract

Optimization of vehicle performance for vehicle control is supported. A system for interacting with a vehicle user includes an optimization module configured to determine a control strategy for traversing a selected route based on a probabilistic prediction of actuator inputs, a monitoring module configured to monitor vehicle motion and driver inputs during operation of the vehicle on the selected route, and a prediction module configured to predict the vehicle motion and driver inputs during operation of the vehicle on the selected route. And a controller configured to predict an optimal local path for traversing a section of the selected route and to predict a set of optimal actuator control actions. The system also includes an interface module configured to present a set of suggested actuator commands to a vehicle user based on the set of optimal actuator control actions, the control system is configured to control the vehicle to execute the set of suggested actuator commands based on the vehicle user accepting the set of suggested actuator commands.

Description

[0001] introduce

[0002] This disclosure relates to the field of vehicle control. More particularly, this disclosure relates to systems and methods for optimizing user control of vehicles.

[0003] Vehicles are increasingly equipped with sensors and perception devices that improve vehicle control systems and driver awareness, thereby enabling autonomous control and / or driver support. In many situations, such as high-performance driving on the track, drivers are constantly seeking ways to improve overall performance. Given the numerous variables that can affect performance and the countless potential emergencies that can occur, systems that support drivers in improving their overall performance are desirable. Summary of the Invention

[0004] In one exemplary embodiment, a system for interacting with a vehicle user includes an optimization module configured to determine a control strategy for traversing a selected route based on probabilistic predictions of actuator inputs; a monitoring module configured to monitor vehicle motion and vehicle user inputs during vehicle operation on the selected route; and a prediction module configured to predict the optimal local path for traversing a segment of the selected route and to predict an optimal set of actuator control actions. The system also includes an interface module configured to present a suggested set of actuator commands to the vehicle user based on the optimal set of actuator control actions, and a control system configured to control the vehicle to execute the suggested set of actuator commands based on the vehicle user's acceptance of the suggested set of actuator commands.

[0005] In addition to one or more features described herein, the system includes a data fusion module configured to select a detection method for detecting each of a plurality of emergency events, the detection method specifying a combination of one or more types of detection data.

[0006] In addition to one or more features described in this paper, the detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0007] In addition to one or more features described herein, the control strategy is determined based on a driver model and a vehicle model, and specifies actuator input values ​​over time that maximize the vehicle's performance while traversing the selected route.

[0008] In addition to one or more features described in this paper, the control strategy is determined based on a surrogate model constructed from the driver model.

[0009] In addition to one or more features described in this paper, the prediction module is configured to probabilistically predict the best local path over multiple consecutive time horizons based on the control policy and the probability of encountering an emergency during each consecutive time horizon.

[0010] In addition to one or more features described herein, the selected route is a pre-planned route along a road with known parameters, and the control strategy is determined to optimize the actuator input to minimize the amount of time required to traverse the pre-planned route.

[0011] In another exemplary embodiment, a method for interacting with a vehicle user includes determining a control strategy for traversing a selected route by an optimization module based on probabilistic predictions of actuator inputs, monitoring vehicle motion and vehicle user inputs during vehicle operation on the selected route, and predicting, by a prediction module, an optimal local path for traversing a segment of the selected route, and predicting a set of optimal actuator control actions. The method also includes presenting a set of suggested actuator commands to the vehicle user based on the set of optimal actuator control actions, and controlling the vehicle to execute the set of suggested actuator commands based on the vehicle user's acceptance of the set of suggested actuator commands.

[0012] In addition to one or more features described herein, the method includes selecting a detection method for detecting each of a plurality of emergency events, the detection method specifying a combination of one or more types of detection data.

[0013] In addition to one or more features described in this paper, the detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0014] In addition to one or more features described herein, the control strategy is determined based on a driver model and a vehicle model, and specifies actuator input values ​​over time that maximize the vehicle's performance while traversing the selected route.

[0015] In addition to one or more features described in this paper, the control strategy is determined based on an agent model constructed from the driver model.

[0016] In addition to one or more features described in this paper, predicting the best local path involves probabilistically predicting the best local path over multiple consecutive time ranges based on control strategies and the probability of encountering an emergency event in each consecutive time range.

[0017] In addition to one or more features described herein, the selected route is a pre-planned route along a road with known parameters, and the control strategy is determined to optimize the actuator input to minimize the amount of time required to traverse the pre-planned route.

[0018] In yet another exemplary embodiment, a vehicle system includes a memory having computer-readable instructions and a processing device for executing the computer-readable instructions. The computer-readable instruction control processing device determines a control strategy for traversing a selected route based on probabilistic predictions of actuator inputs, monitors vehicle motion and vehicle user input during vehicle operation on the selected route, predicts an optimal local path for traversing a segment of the selected route, predicts an optimal set of actuator control actions, presents a set of suggested actuator commands to the vehicle user based on the set of optimal actuator control actions, and controls the vehicle to execute the set of suggested actuator commands based on the vehicle user's acceptance of the set of suggested actuator commands.

[0019] In addition to one or more features described herein, the vehicle system includes a data fusion module configured to select a detection method for detecting each of a plurality of emergency events, the detection method specifying a combination of one or more types of detection data.

[0020] In addition to one or more features described in this paper, the detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0021] In addition to one or more features described herein, the control strategy is determined based on a driver model and a vehicle model, and specifies actuator input values ​​over time that maximize vehicle performance while traversing the selected route.

[0022] In addition to one or more features described in this paper, the control strategy is determined based on an agent model constructed from the driver model.

[0023] In addition to one or more features described in this paper, predicting the optimal local path involves probabilistically predicting the optimal local path across multiple consecutive time ranges based on control strategies and the probability of encountering an emergency at each consecutive time range.

[0024] A system for interacting with a vehicle user includes: an optimization module configured to determine a control strategy for traversing a selected route based on probabilistic predictions of actuator inputs; a monitoring module configured to monitor vehicle motion and vehicle user inputs during vehicle operation on the selected route; a prediction module configured to predict an optimal local path for traversing a segment of the selected route and to predict an optimal set of actuator control actions; an interface module configured to present a set of suggested actuator commands to the vehicle user based on the optimal set of actuator control actions; and a control system configured to control the vehicle to execute the suggested set of actuator commands based on the vehicle user's acceptance of the suggested set of actuator commands.

[0025] It further includes a data fusion module configured to select a detection method for detecting each of multiple emergency events, wherein the detection method specifies a combination of one or more types of detection data.

[0026] The detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0027] The control strategy is determined based on the driver model and the vehicle model, and the control strategy specifies the actuator input values ​​over time, which maximize the vehicle's performance when traversing the selected route.

[0028] Among them, the control strategy is determined based on the agent model constructed from the driver model.

[0029] The prediction module is configured to probabilistically predict the optimal local path across multiple consecutive time ranges based on the control strategy and the probability of encountering an emergency during each consecutive time range.

[0030] The selected route is a pre-planned route along a road with known parameters, and the control strategy is determined to optimize the actuator input to minimize the amount of time required to traverse the pre-planned route.

[0031] A method for interacting with a vehicle user includes: determining a control strategy for traversing a selected route by an optimization module based on probabilistic predictions of actuator inputs; monitoring vehicle motion and vehicle user inputs during vehicle operation on the selected route; predicting an optimal local path for traversing a segment of the selected route by a prediction module, and predicting a set of optimal actuator control actions; presenting a set of suggested actuator commands to the vehicle user based on the set of optimal actuator control actions; and controlling the vehicle to execute the set of suggested actuator commands based on the vehicle user's acceptance of the set of suggested actuator commands.

[0032] It further includes selecting a detection method for detecting each of the multiple emergency events, wherein the detection method specifies a combination of one or more types of detection data.

[0033] The detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0034] The control strategy is determined based on a driver model and a vehicle model, and the control strategy specifies the actuator input values ​​over time, which maximize the vehicle's performance while traversing the selected route.

[0035] Among them, the control strategy is determined based on the agent model constructed from the driver model.

[0036] The prediction of the optimal local path includes probabilistically predicting the optimal local path across multiple consecutive time ranges based on the control strategy and the probability of encountering an emergency event in each consecutive time range.

[0037] The selected route is a pre-planned route along a road with known parameters, and the control strategy is determined to optimize the actuator input to minimize the amount of time required to traverse the pre-planned route.

[0038] A vehicle system includes: a memory having computer-readable instructions; and a processing device for executing the computer-readable instructions, the computer-readable instructions controlling the processing device to perform: determining a control strategy for traversing a selected route based on probabilistic predictions of actuator inputs; monitoring vehicle motion and vehicle user inputs during vehicle operation on the selected route; predicting an optimal local path for traversing a segment of the selected route and predicting a set of optimal actuator control actions; presenting a set of suggested actuator commands to a vehicle user based on the set of optimal actuator control actions; and controlling the vehicle to execute the set of suggested actuator commands based on the vehicle user's acceptance of the set of suggested actuator commands.

[0039] It further includes a data fusion module configured to select a detection method for detecting each of multiple emergency events, wherein the detection method specifies a combination of one or more types of detection data.

[0040] The detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

[0041] The control strategy is determined based on a driver model and a vehicle model, and the control strategy specifies the actuator input values ​​over time, which maximize vehicle performance while traversing the selected route.

[0042] Among them, the control strategy is determined based on the agent model constructed from the driver model.

[0043] The prediction of the optimal local path includes probabilistically predicting the optimal local path across multiple consecutive time ranges based on the control strategy and the probability of encountering an emergency event in each consecutive time range.

[0044] The above-described features and advantages, as well as other features and advantages of this disclosure, will become readily apparent from the following detailed description when considered in conjunction with the accompanying drawings. Attached Figure Description

[0045] Other features, advantages, and details appear by way of example only in the following detailed description, which refers to the accompanying drawings, in which:

[0046] Figure 1This is a top view of a motor vehicle including aspects of user interaction and prediction systems according to an exemplary embodiment;

[0047] Figure 2 A prediction and user interaction system according to an exemplary embodiment is described;

[0048] Figure 3 It is a flowchart depicting aspect a of the prediction, control, and user interaction method according to an exemplary embodiment;

[0049] Figure 4 A data fusion process for determining a detection method for an emergency event, according to an exemplary embodiment, is described;

[0050] Figure 5 An example of a decision tree according to an exemplary embodiment is depicted;

[0051] Figure 6 Aspects of a simplified vehicle model according to exemplary embodiments are depicted; and

[0052] Figure 7 A computer system according to an exemplary embodiment is depicted; Detailed Implementation

[0053] The following description is exemplary in nature and is not intended to limit this disclosure, its application, or use. It should be understood that throughout the drawings, corresponding reference numerals indicate the same or corresponding parts and features.

[0054] According to one or more exemplary embodiments, methods and systems are provided for supporting cooperative operation of a vehicle. One embodiment of the system is configured to cooperate or coordinate with a driver by monitoring vehicle operation during route traversal and determining an optimal set of controls (actuator inputs) predicted to optimize the vehicle's performance during the traversal route. The optimal set of controls is presented to the driver as a recommendation to improve performance. The driver has the option to accept the recommendation either by directly inputting actuator commands (e.g., controlling steering and throttle) or by providing instructions to the system.

[0055] The embodiments described herein present several advantages. For example, the embodiments provide enhanced vehicle responsiveness and improved driver performance. The embodiments provide drivers with improved real-time information, increasing their understanding of the vehicle environment and presenting opportunities to enhance driver performance (e.g., increased lap times). The embodiments also provide predictive control for rapid and controlled responses to potential emergencies. Integration from various sources (data fusion) can be used to predict and assess the likelihood of emergencies, thereby enhancing the ability to respond to such emergencies.

[0056] For example, when driving on a racetrack, a driver may be unsure which path to follow (e.g., when cornering) and / or what speed to choose. The embodiments provide information to the driver that allows them to maximize lap times. Furthermore, the embodiments assist the driver in responding to events to maximize performance. This type of information can be presented to the driver quickly and intuitively, enabling them to make decisions more efficiently.

[0057] Figure 1 An embodiment of a motor vehicle 10 is shown, which includes a body 12 that at least partially defines a passenger compartment 14. The body 12 also supports various vehicle subsystems, including a propulsion system 16 and other subsystems that support the functions of the propulsion system 16 and other vehicle components, such as a braking subsystem, a suspension system, a steering subsystem, and, if the vehicle is a hybrid electric vehicle, a fuel injection subsystem, an exhaust subsystem, and other subsystems.

[0058] Vehicle 10 may be an internal combustion engine vehicle, an electric vehicle (EV), or a hybrid vehicle. In an embodiment, vehicle 10 is a hybrid vehicle that includes an internal combustion engine system 18 and at least one electric motor assembly. In an embodiment, propulsion system 16 includes an electric motor 20 and may include one or more additional motors, such as motor 32, located at various locations. Vehicle 10 may be a fully electric vehicle with one or more electric motors.

[0059] Vehicle 10 also includes various control devices for controlling aspects of vehicle operation. Such devices include, for example, an engine 18 and motors 20 and 32, a steering wheel 34, an accelerator pedal 36, a front brake 38 connected to the front wheels 26, and a rear brake 40 connected to the rear wheels 30. Other actuators may include a front downforce actuator 42 and / or a rear downforce actuator 44 (e.g., if vehicle 10 is a high-performance vehicle). The control devices and actuators are controllable via one or more control units, which are uniformly represented by a controller 46. Various sensors are included for monitoring vehicle motion and operation, such as torque sensors, pedal position sensors, brake sensors, inertial measurement units (IMUs), and other sensors.

[0060] The vehicle also includes an environmental monitoring system for detecting and monitoring the environment around the vehicle. This monitoring system includes, for example, one or more optical cameras 50 configured to capture images, which can be still images and / or video images. Additional devices or sensors may be included, such as one or more radar components 52 included in the vehicle 10. The monitoring system is not so limited and may include other types of sensors, such as lidar and infrared sensors.

[0061] Vehicle 10, the monitoring system, the controller 46, and other vehicle systems include or are connected to an onboard computer system 54, which includes one or more processing devices 56 and a user interface 58. The user interface 58 may include a touchscreen, a voice recognition system, and / or various buttons for allowing users to interact with features of the vehicle. The user interface 58 may be configured to interact with a user or driver via visual communication (e.g., text and / or graphical displays), tactile communication or alarms (e.g., vibration), and / or auditory communication. The onboard computer system 54 may also include or communicate with devices for monitoring the user or driver, such as internal cameras and image analysis components. Such devices may be incorporated into a driver monitoring system (DMS).

[0062] User interface 58 may include various types of displays and / or other devices that can interact with and / or deliver information to the user. For example, vehicle 10 may include displays (e.g., console screens, full-view mirrors, or FDMs). In one embodiment, vehicle 10 includes one or more head-up displays (HUDs). Other devices that may be incorporated include indicator lights, haptic devices, interior lights, auditory communication devices, and others. Haptic devices (haptic interfaces) include, for example, vibrating devices in the vehicle steering wheel and / or seats. Various displays, haptic devices, lights, and auditory devices are configured to be used in various combinations to present information and suggestions to the user.

[0063] Figure 2 An embodiment of a predictive and user interaction system 60 is depicted, configured to collaborate with a user or driver during vehicle operation to optimize or enhance performance. All or part of system 60 may be incorporated into vehicle computer system 54, controller 46, and / or any other suitable processing equipment or control system.

[0064] System 60 includes a perception and estimation module 62 configured to collect detection data from an environmental monitoring system. Examples of collected data include camera images, radar / LiDAR images, object and feature detection using image processing and / or machine learning, and others.

[0065] The data fusion module 64 is configured to evaluate the collected data and determine detection methods for detecting emergencies or events. In an embodiment, the data fusion module 64 includes a reliability assessment module 66 for determining the reliability of various detection modalities (e.g., camera, radar, weather data, Global Positioning System (GPS) data, etc.). A probability calculation module 68 may be included for calculating the probability of an emergency, as discussed further herein.

[0066] An “emergency” is an event, feature, object, or condition that may occur while a vehicle is traveling along a given route and requires some real-time action or deviation to respond to it. Examples of emergencies include objects such as vehicles or pedestrians in the path of vehicle 10, changes in road conditions, weather conditions, and others.

[0067] The global optimization module 70 is configured to generate the optimal control strategy for traversing the planned route, and the prediction module 72 is used to predict the optimal actuator inputs and vehicle control in response to emergency situations. The prediction module can predict the optimal local path for segments of the planned route (box 74), provide braking control predictions (box 76), downforce command predictions (78), and / or other predictions for actuator inputs or commands to optimally respond to emergency events.

[0068] Control devices such as controller 46 are configured to operate one or more actuators based on user input. The actuators(s) may be operated based on direct user input, such as steering or engaging the accelerator pedal. Alternatively or additionally, controller 46 may operate the actuators(s) in response to user interaction with user interface 80. For example, controller 46 may operate actuators 42 and 44(s). Figure 1 ) for downforce control (e.g., via downforce control module 47) and / or limited slip differential (LSD) control (e.g., via LSD control module 49).

[0069] Figure 3 Method 100 is described, aspects of which can be performed offline (e.g., global path optimization) or online (i.e., during vehicle operation). Method 100 is discussed in conjunction with boxes 101-106. Method 100 is not limited to the number or order of its steps, as some steps represented by boxes 101-106 may be performed in a different order than described below, or fewer than all steps may be performed. For example, if the selected route is not a pre-planned route along a road (e.g., a racetrack) with known parameters, global optimization as discussed herein can be omitted.

[0070] Combination Figure 1 The discussion covers method 100 for vehicles and processing systems, where the processing system may be, for example, a computer system 54, a controller 46, or a combination thereof. Also referenced is... Figure 2 Method 100 will be discussed in conjunction with the prediction and user interaction system 60. For illustrative purposes, aspects of method 100 will be discussed in conjunction with vehicle 10 and the prediction and user interaction system 60. Note that method 100 is not so limited and can be performed by any suitable processing device or system or combination of processing devices.

[0071] At box 101, the perception and estimation module 62 collects detection data. Detection data includes, for example, camera images, radar data, lidar data, GPS data, map data, temperature, and / or any other suitable type of detection data.

[0072] At box 102, the data fusion module 64 combines different types of detection data and uses the combined data to determine the detection methods for different emergency events.

[0073] Generally, different detection data sources and different algorithms may be appropriate for different events and conditions. For example, detecting objects using image and / or radar data may be most appropriate, while detecting weather conditions using GPS or location data (optionally utilizing image data) may be most appropriate. Furthermore, conditions can affect the reliability of various modalities. For example, image data is more reliable in clear weather conditions. GPS data may also have varying degrees of reliability because location and / or weather conditions can affect signal reliability.

[0074] In this embodiment, the data fusion module 64 utilizes a methodology that considers the probability P of an emergency event occurring and the reliability R of different types of detection data in determining an appropriate emergency detection algorithm and method. The probability P can be determined by the probability calculation module 68, and the reliability R can be determined by the reliability assessment module 66.

[0075] Figure 4 An example of this methodology is depicted. In this example, the data fusion module 64 considers a number of N potential emergency events (events E1-E2). N ) and its correlation probability (P1-P N ).

[0076] In this example, "Method A" refers to event detection using camera data C, and "Method B" refers to event detection using LiDAR data L. Method A and / or Method B may include one or more additional types of detection data, such as map data M. In this example, P1 refers to the probability of an obstacle or object entering the road (event E1), and P2 refers to the probability of a vehicle encountering ice or slippery conditions (event E2).

[0077] For example, for event E1, P1 is calculated as follows:

[0078]

[0079] Where c is a constant, P 1A It is the probability of detecting event E1 using method A, and P 1B R is the probability of detecting event E1 using method B. 1AThe reliability of the data collected using method A (determined based on weather conditions, visibility, etc.), and R 1B This refers to the reliability of the data collected using method B.

[0080] If P1 is high enough, the emergency detection method for detecting event E1 is a combination of method A and method B (camera and LiDAR data). If P1 is not high enough (i.e., it does not meet or exceeds the selected threshold), then method A or method B is selected based on which reliability R... 1A Or R 1B Higher. For the remaining events E2 to E N Similarly, the detection method is determined. Note that real-time data can be collected during operation and used to update the probability and reliability scores as appropriate.

[0081] Refer again Figure 3 At box 103, the global optimization module 70 determines the optimal control strategy for traversing the selected route. This route can be a pre-planned route (e.g., a planned route specifying a particular road to be traveled, or a racetrack or other included road).

[0082] In the following description, the selected route is a pre-planned route, and global optimization is performed to specify the vehicle dynamics and actuator inputs that lead to the optimal vehicle path and optimal actuator input. Optimization can be performed offline (e.g., before operating the vehicle) or online during operation.

[0083] The global optimization module 70 is not limited to pre-planned or pre-selected routes. For example, if the vehicle is not traveling on a pre-planned route, the optimization strategy can be executed in real time over a relatively small time frame to locally optimize performance.

[0084] In an embodiment, if the expected road properties (e.g., direction, banking, road surface) are known and the route is known (e.g., the route is a single lap of the track), then global optimization is performed to optimize vehicle handling on the route or single lap (single lap timing optimization).

[0085] Lap timing optimization can be performed by determining the optimal vehicle dynamics and actuator control for consecutive track segments or time windows. For example, determining the optimal pedal and steering inputs (and other actuator commands, such as downforce control) for each segment of a track or other known route. Examples of segments include straight sections and different curved sections (each segment may have its own parameters, such as path, turning radius, incline, road gradient, etc.).

[0086] In this embodiment, optimization is performed based on a driver model and a high-fidelity vehicle model. This optimization determines the optimal input for a route with known profiles and attributes, including track layout, incline and gradient features, road boundaries, and others.

[0087] The driver model simulates the actuator inputs of a human driver and can be represented by the following:

[0088] f(x) = αx 2 +βx+γ,

[0089] Where x represents vehicle position, vehicle speed, actuator state, or other variables related to position and / or dynamics, and α, β, and Y are coefficients. For example, for braking speed control, the braking input A... x，brake It can be expressed by the following formula:

[0090] A x，brake =αV x 2 +βV x +γ,

[0091] Where V x It is the braking speed, subject to maximum and minimum constraints. Any number of functions can exist for each actuator input. Furthermore, the function for a given input can be any suitable polynomial.

[0092] The vehicle model simulates the vehicle's response to driver input or actuator commands. Specifically, it simulates the vehicle's dynamic response to different inputs (e.g., changes in speed, direction, yaw, etc.). In the example, the vehicle model includes elements representing the body, wheels, tires, powertrain, suspension, brakes, steering, and others.

[0093] For example, by using a selected machine learning technique, such as deep learning or reinforcement learning, the function and coefficients for each actuator input are learned.

[0094] To reduce the computational costs associated with this optimization, in this embodiment, a surrogate model is used to evaluate the objective in each optimization cycle (i.e., each single-cycle optimization or optimization for a track segment).

[0095] Proxy models mimic the behavior of vehicle and driver models while being computationally cheaper. Proxy optimization attempts to find the global minimum of an objective function (e.g., lap time) using a relatively small amount of objective function evaluation.

[0096] Agent optimization balances two objectives: exploration and speed. Exploration is performed to find the global minimum. Speed ​​is related to the number of evaluations of the objective function.

[0097] Proxy optimization alternates between Proxy construction and minimization. The Proxy model is constructed by evaluating the objective function (e.g., a driver model) at multiple points and interpolating radial basis functions at these points.

[0098] The search for the minimum involves sampling random points (e.g., thousands of points) within the constraints of an objective function f(x). The objective function f(x) is evaluated to derive its value at one or more sample points, and the values ​​are interpolated to create a surrogate model.

[0099] For a given point (incumbent), the merit function is evaluated based on the value s(x) of the surrogate model at the incumbent and the distance to each point. The distance, denoted as D(x), is defined as the difference between the surrogate value at point x and the value of the objective function at that point.

[0100] The value function is determined by sampling points around the current incumbent and calculating the value function based on the distance D(x) relative to point x.

[0101] Select the optimal point (i.e., the point with the highest value function) and evaluate the objective function at the optimal point. Then, adjust the surrogate model by updating the surrogate model at the optimal point using the evaluation results of the objective function.

[0102] In the embodiment, the value function (denoted as f) meerit The following is a weighted combination of the scaled surrogate model S(x) and the distance D(x):

[0103] f merit (x)=wS(x)+(1-w)D(x),

[0104] Here, w is the weight value. A larger w value indicates greater importance to S(x), causing the surrogate optimization algorithm to minimize the surrogate model. A smaller w value indicates greater importance to points with larger D(x) values, leading to the search of new regions of the surrogate model.

[0105] The proportional proxy model S(x) is as follows, based on the proxy value s(x) at point x:

[0106] S(x)=(s(x)-s min ) / (s max -s min ),

[0107] Where s max and s min These are the maximum and minimum values ​​among the sample points.

[0108] The distance D(x) is defined based on the set of evaluation points, denoted by x. j d, where j = 1...k. ij d is defined as the distance from sample point i to evaluation point k of the driver model. min and d max It is the distance d among all evaluation and sample points. ij The minimum and maximum values.

[0109] D(x) can be expressed by the following formula:

[0110] D(x)=(d max -d(x)) / (d max -d min ),

[0111] Where d(x) is the minimum distance between point x on the surrogate model and evaluation point k.

[0112] At box 104, a probabilistic prediction method is used, taking into account environmental uncertainties, and local path optimization is performed. Prediction module 72 performs local predictions during vehicle operation (e.g., when the vehicle traverses a section of the track).

[0113] Typically, prediction module 72 determines the probability of a given emergency event occurring (i.e., the event or situation that the vehicle will encounter if it does not take actions that deviate from the planned path).

[0114] refer to Figure 5 In one embodiment, prediction module 72 uses a prediction tree to assess the probability of an emergency event (or any of the potential emergency events) occurring for each of a plurality of prediction windows. The prediction windows are continuously updated as the vehicle is operated.

[0115] The prediction tree shows nominal and emergency prediction branches with associated probabilities. For example, as... Figure 5 As shown, prediction tree 90 includes nodes or time steps, and branches representing the nominal state (no emergency, the vehicle can continue as normal). Prediction tree 90 also includes branches for emergency situations. Any number of prediction windows, branches, and nodes can be generated, and is therefore not limited to... Figure 5 Those shown.

[0116] As shown, the prediction tree 90 includes nodes 0, 1, N1, N1+1, N2, and N3, each node having an associated location state x. i The subscript i includes subscript n, which indicates a nominal state, and subscript e, which indicates an emergency. The number of subscripts is based on the number of possible emergencies. Figure 5In the example, a first emergency event (e.g., a potential pop-up obstacle) exists in a first time range 91 including nodes N1 and N1+1, and a second emergency event (e.g., ice) exists in a second time range 92 including nodes N2 and N3.

[0117] For example, x n,n This refers to the location state when no emergency event is predicted, and x e,n This refers to the location state when the first emergency event is predicted to occur (i.e., with a high probability, or a probability higher than a threshold). n,e This refers to the location state when the first emergency event is not predicted to occur, but the second emergency event is predicted to occur. e,e This refers to the location state when both emergency events are predicted to occur. Each node associated with an emergency event also includes the associated control action u.

[0118] exist Figure 5 In the example, if the position state at the beginning of time range 91 is not predicted to lead to an emergency (position state is x) n,n If the prediction tree 90 proceeds to node H1+1 without branches, then the prediction tree 90 will continue without branches. However, if the position state at the beginning of time range 91 is predicted to lead to an emergency event (position state is x), then the prediction tree 90 will continue without branches. e,n If the prediction tree 90 proceeds to node H1+1 via a branch, then the prediction tree 90 will continue using the branch. For example, in... Figure 5 As shown, prediction tree 90 can similarly branch in time range 92.

[0119] During operation, prediction module 72 determines the probability of events within a given time window. If the probability of an emergency event occurring is higher than a threshold (or more likely than the nominal condition), prediction module 72 determines a recommended control action u. When prediction module 72 processes decision tree 90, multiple emergency events may exist within more than one time window. In such instances, the recommended control actions can combine multiple actions to resolve the emergency event.

[0120] In this embodiment, a vehicle prediction model is used to describe the vehicle dynamics (e.g., speed, heading, yaw, etc.) and position state of vehicle 10. The vehicle prediction model also optimizes control actions based on various constraints.

[0121] In this embodiment, the vehicle prediction model uses Figure 6 The planar bicycle model shown is a control-oriented model. The planar bicycle model includes two velocity states (longitudinal velocity v). x and lateral velocity v y ), yaw rate r, position status 112 and 114, heading deviation Δψ and lateral deviation e.

[0122] Given the desired longitudinal velocity v along the path x Furthermore, a longitudinal controller is used to calculate the required torque for following the desired longitudinal velocity. The lateral dynamics of the vehicle can be described by the following equations for changes in lateral acceleration and yaw rate:

[0123] in:

[0124]

[0125] In the above formula, m is the vehicle mass, and I... z This is the vehicle's moment of inertia, g is the acceleration due to gravity, and l... f and l r θ is the distance from the front and rear axles to the vehicle's center of gravity (CG). r It is the road slope and Φ r It refers to the road's angle of inclination. F yf and F yr These refer to the front and rear tire forces, respectively.

[0126] Tire forces can be determined using a brush tire model, which captures the decrease in lateral force caused by the applied longitudinal force. The tire force F on the tire. y It can be expressed by the following formula:

[0127] in:

[0128]

[0129] C a It refers to tire cornering stiffness, F z α is the normal load on the tire, μ is the sideslip angle, μ is the road friction force, and ξ is the derating factor.

[0130] The derating factor represents the remaining lateral force capacity based on the friction limit circle as follows:

[0131]

[0132] Where F x It is a longitudinal force.

[0133] Front tire sideslip angle α f and rear tire sideslip angle α r Determined based on the following formula:

[0134]

[0135] The nonlinear tire force (Fyr) of the rear axle is linearized at each time step around the vehicle's current operating point. The front lateral force (F...) yf The yaw rate is treated as a control input, rather than the steering angle δ, and is then mapped to the steering angle. The lateral dynamics are then written as affine functions of the velocity and yaw states:

[0136]

[0137] in, and These are the lateral force, rear slip angle, and cornering stiffness of the rear axle, respectively. The positional states of the heading deviation Δψ and lateral deviation e are relative to a given curvature κ(s) and road slope θ. r The paths of (s) and road inclination angle φ(s) are local, which specifies the equation of motion as:

[0138] Δψ=r-uκ(s)cos(φ r cos(θ) r ),

[0139]

[0140] The resulting continuous-time vehicle model includes vehicle dynamics and position state, and can be represented as follows:

[0141]

[0142] Where x is a state vector with multiple vehicle states, and u is the control action.

[0143] For example, x and u are represented as follows:

[0144] x = [v y r Δψ e],

[0145] u = F yf .

[0146] The coefficients of the vehicle model are:

[0147]

[0148]

[0149] Then the continuous model is discretized as follows:

[0150] x(k+1)=Akx(k)+B k u(k)+d k .

[0151] The overall system model (i.e., the driver model and the vehicle model) is expanded to include the entire prediction range. For example, for three potential emergency events within the prediction window, the variable position state is denoted as x. i,i,i For k time steps, the discretized system model can be represented as:

[0152]

[0153] The optimization of control actions can be achieved using the objective function J. e To execute the objective function J e Defined as:

[0154]

[0155]

[0156] In the objective function J e In this context, subscript i includes subscript n indicating the nominal state at a given time step, and subscript e indicating an emergency or situation associated with a control action.

[0157] In the objective function J e In the middle, sr i，i,i It is the slack variable for the yaw rate constraint of each predicted path, ssβ i,i，i These are the slack variables for the sideslip constraints of each predicted path, and se i,i，i P is the slack variable for the lateral deviation constraint of each prediction path. i，i，i Q is the probability of each predicted path. i,i,i M is the tracking weight for each predicted path, and M i，i,i It is the weight of the proximity of the control action for each predicted path. These are the weights of the yaw rate constraint slack variables for each predicted path. These are the weights of the sideslip constraint relaxation variables, and These are the weights of the slack variables for the lateral deviation constraint.

[0158] Objective function J e It is subject to discrete-time models and various constraints. These constraints may include control action constraints, slew rate constraints, yaw rate stability soft constraints, sideslip stability soft constraints, and lateral deviation soft constraints (collision avoidance).

[0159] For example, slack variables are constrained to be non-negative:

[0160]

[0161] The objective function may also be subject to the equality constraint regarding positional states as follows:

[0162]

[0163] The objective function can be further constrained by the equality constraint of control actions as follows:

[0164]

[0165] Additional constraints can be applied, including yaw rate stability constraints, sideslip stability constraints, and / or lateral deviation constraints. The yaw rate stability constraints are as follows:

[0166] r min ≤r≤r max ,in:

[0167]

[0168]

[0169] The sideslip stability constraints are as follows:

[0170]

[0171] Where α r，sat This is the saturated (maximum) backslip angle. The lateral deviation constraint is as follows:

[0172] e min (k)≤e(k)≤e max (k).

[0173] Refer again Figure 3 At box 105, the optimized control action is presented to the driver as a suggested actuator control or actuator input. This suggestion can take any suitable form, such as visual, auditory, tactile, or a combination thereof.

[0174] For example, interface 80 could be a real-time advisory module that presents information and suggestions to the driver. This module can be configured to provide an augmented reality display or screen display indicating obstacles and the surrounding environment, (e.g., from global optimization) a predicted optimal path, and a predicted optimal actuator command (suggested control action). A haptic system can provide directional information and / or warnings. The magnitude of the haptic signal can be controlled based on, for example, the predicted urgency or imminence of an event.

[0175] For example, an augmented reality display projected onto the windshield can provide information about obstacles or other emergencies and their probability, the vehicle's local location, current actuator commands, potential changes in road gradient, inclination, and curvature, among other things.

[0176] The display can provide textual and / or visual suggestions for the optimal route (e.g., route and speed profiles). These suggestions assist the driver in making decisions and finding the best way to traverse road sections or turns and / or react to emergencies. Depending on the situation, different options may be given to the driver. For example, if an obstacle is predicted, multiple suggestions may be presented. The driver can decide which option to take based on, for example, their preferred driving style. A prediction of the downforce level to be applied can be presented to inform the driver and reduce any conflict between the driver and the downforce actuator.

[0177] At box 106, vehicle 10 is controlled if the driver accepts the suggested actuator control. For example, the driver may interact with the interface by engaging a touchscreen, giving verbal or nonverbal approval, or otherwise indicating agreement to the suggested actuator control. In another example, the driver may accept the suggestion by directly controlling one or more actuators as suggested (e.g., by engaging the brake, engaging the accelerator pedal, steering, etc.).

[0178] Figure 7 The illustration shows aspects of an embodiment of a computer system 140 that can perform various aspects of the embodiments described herein. The computer system 140 includes at least one processing device 142, which typically includes one or more processors for performing aspects of the image acquisition and analysis methods described herein.

[0179] The components of computer system 140 include processing device 142 (such as one or more processors or processing units), memory 144, and bus 146 coupling various system components, including system memory 144, to processing device 142. System memory 144 may be a non-transitory computer-readable medium and may include a variety of computer system-readable media. Such media may be any available media accessible by processing device 142 and include volatile and non-volatile media, as well as removable and non-removable media.

[0180] For example, system memory 144 includes non-volatile memory 148 such as a hard disk drive, and may also include volatile memory 150 such as random access memory (RAM) and / or cache memory. Computer system 140 may further include other removable / non-removable, volatile / non-volatile computer system storage media.

[0181] System memory 144 may include at least one program product having (e.g., at least one) collection of program modules configured to implement the functionality of the embodiments described herein. For example, system memory 144 stores various program modules that typically implement the functionality and / or methodology of the embodiments described herein. One or more modules 152 may be included to perform the functionality discussed herein. System 140 is not so limited, as it may include other modules. As used herein, the term "module" refers to a processing circuitry system that may include application-specific integrated circuits (ASICs), electronic circuitry, processors (shared, dedicated, or grouped) and memories executing one or more software or firmware programs, combinational logic circuitry, and / or other suitable components providing the described functionality.

[0182] The processing device 142 can also communicate with one or more external devices 156, such as a keyboard, a pointing device, and / or any device that enables the processing device 142 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Communication with various devices can occur via input / output (I / O) interfaces 164 and 165.

[0183] Processing device 142 can also communicate with one or more networks 166 via network adapter 168, such as a local area network (LAN), a general wide area network (WAN), a bus network, and / or a public network (e.g., the Internet). It should be understood that, although not shown, other hardware and / or software components may be used in conjunction with computer system 140. Examples include, but are not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, and data archiving storage systems.

[0184] The terms “a” and “an” do not indicate a limitation of quantity, but rather indicate the presence of at least one of the referenced items. The term “or” means “and / or” unless the context clearly indicates otherwise. References to “aspect” throughout the specification mean that a particular element described in connection with that aspect (e.g., a feature, structure, step, or characteristic) is included in at least one aspect described herein and may or may not be present in other aspects. Furthermore, it will be understood that the described elements may be combined in any suitable manner across various aspects.

[0185] When a component, such as a layer, film, region, or substrate, is referred to as being "on" another component, it can be directly on the other component, or there may be intermediate components. Conversely, when a component is referred to as being "directly on another component," there are no intermediate components.

[0186] Unless otherwise specified herein, all test standards are the most recently effective standards up to the date of filing of this application, or, if priority is claimed, the filing date of the earliest priority application in which the test standard appeared.

[0187] Unless otherwise defined, the technical and scientific terms used herein have the same meaning as commonly understood by one skilled in the art to which this disclosure pertains.

[0188] While the foregoing disclosure has been described with reference to exemplary embodiments, those skilled in the art will understand that various changes can be made and equivalents can be substituted for its elements without departing from its scope. Furthermore, many modifications can be made to adapt particular situations or materials to the teachings of this disclosure without departing from its essential scope. Therefore, this disclosure is intended to be limited to the specific embodiments disclosed, but will include all embodiments falling within its scope.

Claims

1. A system for interacting with vehicle users, comprising: The optimization module is configured to determine the control strategy for traversing the selected route based on probabilistic predictions of the actuator inputs; The monitoring module is configured to monitor vehicle movement and vehicle user input during vehicle operation on a selected route; The prediction module is configured to predict the optimal local path for traversing the selected route segment and to predict the optimal set of actuator control actions. The interface module is configured to present a set of suggested actuator commands to the vehicle user based on the set of optimal actuator control actions; and The control system is configured to control the vehicle to execute a set of suggested actuator commands based on a set of actuator commands accepted by the vehicle user.

2. The system of claim 1, further comprising a data fusion module configured to select a detection method for detecting each of a plurality of emergency events, the detection method specifying a combination of one or more types of detection data.

3. The system according to claim 2, wherein, The detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

4. The system according to claim 1, wherein, A control strategy is determined based on a driver model and a vehicle model, and the control strategy specifies actuator input values ​​over time that maximize the vehicle's performance while traversing a selected route.

5. The system according to claim 4, wherein, The control strategy is determined based on the agent model constructed from the driver model.

6. The system according to claim 1, wherein, The prediction module is configured to probabilistically predict the best local path over multiple consecutive time ranges based on the control strategy and the probability of encountering an emergency during each consecutive time range.

7. The system according to claim 1, wherein, The selected route is a pre-planned route along a road with known parameters, and the control strategy is determined to optimize the actuator input to minimize the amount of time required to traverse the pre-planned route.

8. A method for interacting with a vehicle user, comprising: The optimization module determines the control strategy for traversing the selected route based on probability predictions of the actuator input; Monitor vehicle movement and user input while the vehicle is operating on the selected route; The prediction module predicts the optimal local path for traversing the selected route segment and the optimal set of actuator control actions. A set of suggested actuator commands is presented to the vehicle user based on the set of optimal actuator control actions; and The vehicle is controlled to execute the set of suggested actuator commands based on the set of suggested actuator commands accepted by the vehicle user.

9. The method of claim 8, further comprising selecting a detection method for detecting each of a plurality of emergency events, the detection method specifying a combination of one or more types of detection data.

10. The method according to claim 9, wherein, The detection method is selected based on the probability of detecting an emergency event by the selected detection method and the reliability of the selected detection method.

Images