Vehicle motion control device, in-vehicle system, and vehicle control method
The vehicle motion control device generates trajectories that reflect the driver's characteristics using a trained model, addressing the lack of driver-specific trajectory generation in existing technologies, thereby improving safety and comfort.
Patent Information
- Application Number
- PCT/JP2026/008807
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-14
- Filing Date
- 2026-03-06
- Publication Date
- 2026-11-19
AI Technical Summary
Existing vehicle control technologies fail to generate travel trajectories that reflect the driving characteristics of the driver.
A vehicle motion control device that includes a data acquisition unit, an area determination unit, a trajectory generation parameter prediction unit, and a trajectory generation unit, which utilize a trained model to generate trajectories based on vehicle and environmental data, incorporating the driver's driving characteristics.
Generates vehicle trajectories that accurately reflect the driver's driving style, enhancing safety and comfort by aligning vehicle behavior with the driver's preferences.
Smart Images

Figure JP2026008807_19112026_PF_FP_ABST
Abstract
Description
Vehicle motion control device, in-vehicle system, and vehicle control method
[0001] The present invention relates to a vehicle motion control device, an in-vehicle system, and a vehicle control method.
[0002] As one type of vehicle control technology represented by driving assistance and automated driving, a technology is known that generates a travel trajectory including data such as a travel route and travel speed that serve as the travel targets for the vehicle, and controls the powertrain, brakes, steering and the like so that the vehicle travels along the travel trajectory. As methods for generating a travel trajectory, for example, the techniques disclosed in Patent Document 1 and Patent Document 2 are known. Patent Document 1 discloses a vehicle control device including: a recognition unit that recognizes objects present around the vehicle; a generation unit that generates one or a plurality of target trajectories along which the vehicle should travel, based on the objects recognized by the recognition unit; and a driving control unit that automatically controls driving of the vehicle based on the target trajectories generated by the generation unit. The generation unit calculates a travelable region that is a region in which the vehicle can travel, based on the state of the objects recognized by the recognition unit, excludes target trajectories that exist outside the calculated travelable region from the generated one or more target trajectories, and the driving control unit automatically controls driving of the vehicle based on the target trajectories remaining without being excluded by the generation unit. Patent Document 2 discloses an automated driving device provided with a control program that inputs external environment information and vehicle information and outputs target control values for the vehicle, wherein the control program has a first program that generates a first target control amount based on a dynamically changing algorithm, and a second program that generates a second target control amount based on a prescribed algorithm.
[0003] Japanese Unexamined Patent Application Publication No. 2021-160531, Japanese Unexamined Patent Application Publication No. 2018-24286
[0004] In the inventions described in Patent Document 1 and Patent Document 2, it is not possible to generate a travel trajectory for the host vehicle that reflects the driving characteristics of the driver.
[0005] A vehicle motion control device according to a first aspect of the present invention includes: a data acquisition unit that acquires vehicle data, which is data relating to the vehicle itself, and external data, which is data relating to the surrounding environment of the vehicle, including a drivable area, which is an area on which the vehicle can travel; an area determination unit that determines a first drivable area, which is an area on which the vehicle will travel, with the direction of travel of the vehicle being forward, and at least one of a second drivable area adjacent to the rear of the first drivable area or a third drivable area adjacent to the front of the first drivable area; a trajectory generation parameter prediction unit that inputs the vehicle data, the first drivable area, and at least one of the second drivable area or the third drivable area into a trained model and outputs trajectory generation parameters used to generate the trajectory on which the vehicle will travel; and a trajectory generation unit that inputs the trajectory generation parameters and generates the trajectory on which the vehicle will travel. A second aspect of the present invention provides an in-vehicle system comprising the above-described vehicle motion control device, a powertrain system for controlling the driving force of the vehicle, a brake system for controlling the braking force of the vehicle, and a steering system for controlling the steering angle of the wheels of the vehicle, wherein the vehicle motion control device controls the powertrain system, the brake system, and the steering system so that the vehicle travels along the trajectory generated by the trajectory generation unit.A third aspect of the present invention is a vehicle control method performed by a computer, comprising: an information acquisition step of acquiring vehicle data which is data relating to the vehicle itself and external data which is data relating to the surrounding environment of the vehicle, including vehicle data which is data relating to the vehicle itself and a drivable area which is an area in which the vehicle itself can travel; an area determination step of determining a first drivable area which is an area for generating a trajectory on which the vehicle itself travels, with the direction of travel of the vehicle itself being forward, and at least one of a second drivable area adjacent to the rear of the first drivable area or a third drivable area adjacent to the front of the first drivable area; a trajectory generation parameter prediction step of inputting the vehicle data, the first drivable area, and at least one of the second drivable area or the third drivable area into a trained model and outputting trajectory generation parameters used to generate a trajectory on which the vehicle itself travels; and a trajectory generation step of generating a trajectory on which the vehicle itself travels using the trajectory generation parameters.
[0006] According to the present invention, it is possible to generate a vehicle trajectory that reflects the driver's driving characteristics.
[0007] Functional block diagram of the in-vehicle system Hardware configuration diagram of the vehicle motion control device Detailed configuration diagram of the track planning unit Flowchart showing the processing of the track planning unit Figure 4 shows an example of operation of Figure 2 Figure 3 shows an example of operation of Figure 4 Figure 4 shows an example of operation of Figure 4 Configuration diagram of the operation management unit in the second embodiment Configuration diagram of the in-vehicle system in the third embodiment Configuration diagram of the HMI unit in the fourth embodiment
[0008] Embodiments of the present invention will be described below with reference to the drawings. The following description and drawings are illustrative for explaining the present invention, and have been omitted and simplified as appropriate for clarity of explanation. The present invention can be carried out in various other forms. Unless otherwise specified, each component may be singular or plural. The position, size, shape, and range of each component shown in the drawings may not represent the actual position, size, shape, and range in order to facilitate understanding of the invention. Therefore, the present invention is not necessarily limited to the position, size, shape, and range disclosed in the drawings.
[0009] —First Embodiment— The first embodiment of the vehicle motion control device, the in-vehicle system, and the vehicle control method will be described below with reference to Figures 1 to 5.
[0010] <Functional Block Diagram> Figure 1 is a functional block diagram of the in-vehicle system 1. This in-vehicle system 1 is installed in the vehicle V1 and is a system for performing vehicle control such as driver assistance and autonomous driving. Hereafter, the people riding in the vehicle V1 will be referred to as the "driver" or "occupants". As shown in Figure 1, the in-vehicle system 1 includes an external communication device 11, a GNSS (Global Navigation Satellite System) 12, a map data storage unit 13, a sensor 14, an HMI (Human Machine Interface) unit 15, a vehicle motion control device 2, a powertrain system 3, a brake system 4, and a steering system 5. The vehicle motion control device 2 also includes an operation management unit 21, a driving trajectory planning unit 22, and a driving control unit 23. The following describes each component in order.
[0011] <Data Input Sources for Vehicle Motion Control Device 2> The following describes the external communication device 11, GNSS 12, map data storage unit 13, sensor 14, and HMI unit 15, which are devices that input data to the vehicle motion control device 2. The external communication device 11 enables vehicle-to-vehicle communication between the vehicle V1 and other vehicles, or vehicle-to-infrastructure communication between the vehicle V1 and roadside units, via wireless communication. The external communication device 11 transmits and receives data related to the vehicle V1, data on the surrounding environment of the vehicle V1, and data on the wide-area network. The GNSS 12 receives radio waves transmitted from artificial satellites such as quasi-zenith satellites and GPS (Global Positioning System) satellites, and acquires data such as the position of the vehicle V1.
[0012] The map data storage unit 13 stores general road data used in navigation systems, road data including road width and curve data such as road curvature, data such as road surface conditions and traffic conditions, and data such as the driving status of other vehicles, including vehicle and surrounding environment data. The data stored in the map data storage unit 13 may be updated sequentially with data acquired via vehicle-to-vehicle communication or vehicle-to-infrastructure communication through the external communication device 11.
[0013] Sensor 14 is an external environment recognition sensor that detects data such as the vehicle V1 and the surrounding environment, such as an image sensor, millimeter-wave radar, or lidar. Sensor 14 may also include sensors that detect data such as driver operation, vehicle V1 speed, acceleration, jerk, angular velocity, and wheel steering angle. The data of the vehicle and surrounding environment detected by the external environment recognition sensor in Sensor 14 includes data on various objects present around the vehicle V1, such as obstacles, signs, lane boundaries, lane outlines, buildings, pedestrians, bicycles, and other vehicles. A sign, which is a type of environmental element, is, for example, a "bicycle lane" sign. Sensor 14 may also have a function to recognize lane boundaries and lane outlines based on the difference between the white lines in the image data captured by the image sensor and the brightness of the road surface, and to output the position and shape of the lane boundaries and lane outlines.
[0014] The HMI unit 15 is a device that combines input and output. The HMI unit 15 accepts user input operations such as selecting a driving mode and setting a destination. The HMI unit 15 also presents data that the user needs from the data acquired by the external communication device 11, GNSS 12, and sensor 14, as well as the data recorded in the map data storage unit 13. Information can be presented to the user using screen display or audio output. Furthermore, the HMI unit 15 notifies the user of warnings to alert them.
[0015] <Vehicle Motion Control Device 2> Figure 2 is a hardware configuration diagram of the vehicle motion control device 2. The vehicle motion control device 2 includes an arithmetic unit 24 such as a CPU (Central Processing Unit), a main memory 25 such as semiconductor memory, an auxiliary memory 26, and a communication device 27. The vehicle motion control device 2 is an ECU (Electronic Control Unit) that controls the vehicle V1. The vehicle motion control device 2 realizes various functions such as the operation management unit 21 by executing a program loaded into the main memory 25 with the arithmetic unit 24. In this embodiment, the operation management unit 21, the driving trajectory planning unit 22, and the driving control unit 23 are described as different functional blocks, but at least some of these may be configured as a single functional block. Furthermore, the vehicle motion control device 2 may be realized by the cooperation of multiple hardware components, for example, multiple ECUs, or it may be realized by the same hardware component along with other functions. For example, a higher-level controller may realize various functions of these units.
[0016] The vehicle's driving modes controlled by the vehicle motion control device 2 include, for example, comfort mode, economy mode, sport mode, shortest time mode to minimize travel time, and shortest distance mode to minimize travel distance. The vehicle's driving mode may be set arbitrarily by the user, set in advance by the user, or set by the operation management unit 21 based on the driving condition data described later. As a result, the vehicle motion control device 2 sets the speed, acceleration, jerk of the vehicle V1, the speed and distance between the vehicle V1 and the preceding vehicle traveling in front of it, and so on.
[0017] <Operation Management Unit 21> The operation management unit 21 generates data related to the behavior of the vehicle V1. This data includes the position data of the vehicle V1, data on various objects present around the vehicle V1 (such as the position and speed of the vehicle and surrounding environment), longitudinal acceleration, longitudinal jerk, lateral acceleration, yaw rate, and lateral jerk. The operation management unit 21 generates data related to the behavior of the vehicle V1 based on data acquired by the external communication device 11, GNSS 12, and sensors 14, and map data recorded in the map data storage unit 13.
[0018] Furthermore, the operation management unit 21 periodically transmits the generated location data of its own vehicle V1, data on various objects, and data on the behavior of its own vehicle V1 to other vehicles and roadside units via the external communication device 11, and simultaneously transmits it to the map data storage unit 13. The map data storage unit 13 sequentially updates the stored map data using the acquired location data of its own vehicle V1, data on various objects, and data on the behavior of the vehicle. In addition, the operation management unit 21 sets route data indicating the route from the vehicle's current position to the destination based on this location data of its own vehicle V1, data on various objects, data on the vehicle's behavior, and data acquired by the HMI unit 15 (for example, driving mode and destination). This route data is represented on the map described later and is used to determine the vehicle's driving trajectory. The data generated by the operation management unit 21 and the data set by the operation management unit 21 will be referred to as "driving status data" below.
[0019] <Driving Control Unit 23> The driving control unit 23 sets target driving force, target braking force, target steering angle, etc., so that the vehicle follows the driving trajectory output from the driving trajectory planning unit 22. By setting these, the driving control unit 23 controls the powertrain system 3, brake system 4, and steering system 5.
[0020] <Controlled objects of the driving control unit 23> The powertrain system 3 controls the driving force generated by the internal combustion engine, electric motor, etc., based on the driver's operation or the target driving force output from the driving control unit 23. The brake system 4 controls the braking force generated by the brake caliper, etc., based on the driver's operation or the target braking force output from the driving control unit 23. The steering system 5 controls the steering angle of the wheels based on the driver's operation or the target steering angle output from the driving control unit 23.
[0021] <Trajectory Planning Unit 22> Figure 3 is a detailed configuration diagram of the Trajectory Planning Unit 22, and Figure 4 is a flowchart showing the processing of the Trajectory Planning Unit 22. The Trajectory Planning Unit 22 comprises a data acquisition unit 22a, a region determination unit 22b, a trajectory generation parameter prediction unit 22c, a trajectory generation unit 22d, and a data output unit 22e. The trajectory generation parameter prediction unit 22c includes a trained model M.
[0022] The processing of each component will be explained with reference to the flowchart shown in Figure 4. First, in step S301, the data acquisition unit 22a acquires vehicle data and external environment data. Here, vehicle data refers to data related to the specifications and position of the vehicle V1. This vehicle data is obtained from the operation management unit 21. External environment data refers to data related to the surrounding environment of the vehicle V1, including data that can identify the drivable area of the vehicle V1.
[0023] In the following step S302, the region determination unit 22b takes the vehicle data and external data acquired in step S301 as input and determines the first drivable region Av1 and at least one of the second drivable region Av2 and the third drivable region Av3. That is, the region determination unit 22b may determine three regions: the first drivable region Av1, the second drivable region Av2, and the third drivable region Av3; or it may determine two regions: the first drivable region Av1 and the second drivable region Av2; or it may determine two regions: the first drivable region Av1 and the third drivable region Av3. The second drivable region Av2 is the region adjacent to the rear of the first drivable region Av1, with the direction of travel of the vehicle V1 being forward. The third drivable region Av3 is the region adjacent to the front of the first drivable region Av1, with the direction of travel of the vehicle V1 being forward.
[0024] The region determination unit 22b determines the first drivable region Av1 based on the range that the sensor 14 mounted on the vehicle V1 can recognize. For example, if the range that the sensor 14 can recognize is 100m, the region determination unit 22b defines the first drivable region Av1 as the range up to 50m in front of the vehicle V1. The first drivable region Av1 is the region where the driving trajectory of the vehicle V1 is generated, and also the region where the prediction accuracy of the trajectory generation parameters when the trained model M performs training is evaluated.
[0025] The region determination unit 22b, for example, sets the second drivable region Av2 to the range from the current position of the vehicle V1 a predetermined time ago, with the calculation cycle of the trajectory planning unit being a predetermined time. The region determination unit 22b may also set the second drivable region Av2 to include, for example, half of the range from the current position of the vehicle V1 a predetermined time ago, as well as a predetermined range on the side of the second drivable region Av2 closer to the vehicle V1, in addition to the range from the current position of the vehicle V1 a predetermined time ago.
[0026] The region determination unit 22b determines the third drivable region Av3 based on the range that the sensor 14 mounted on the vehicle V1 can recognize and the first drivable region Av1. For example, the region determination unit 22b defines the third drivable region Av3 as the region obtained by excluding the first drivable region Av1 from the maximum range that the sensor 14 mounted on the vehicle V1 can recognize.
[0027] The second drivable region Av2 and the third drivable region Av3 are regions where data can be obtained that determines the factors that led the vehicle V1 from the past to its current position Bs, and the factors that determine the direction of travel of the vehicle V1 from its current position Bs to its future position. In other words, these regions are useful for improving the prediction accuracy of trajectory generation parameters by the trained model M. Therefore, it is desirable for the region determination unit 22b to determine both the second drivable region Av2 and the third drivable region Av3, but it is also acceptable to determine only one of them.
[0028] In the subsequent step S303, the trajectory generation parameter prediction unit 22c predicts the trajectory generation parameters used to generate the trajectory on which the vehicle V1 travels, using the trained model M. The input to this trained model M includes the vehicle data acquired in step S301, the first drivable region Av1 determined in step S302, and at least one of the second drivable region Av2 or the third drivable region Av3.
[0029] The trajectory generation parameters output by the trained model M include data such as the range of physical quantities including at least one of the vehicle V1's velocity, acceleration, and jerk; a margin width Mw that specifies the distance from the left and right edges of the drivable area Av to the vehicle V1's trajectory; and the endpoint position Be, which is the position of the vehicle V1 at the boundary between the first drivable area Av1 and the third drivable area Av3. Furthermore, the aforementioned range of physical quantities, margin width Mw, and endpoint position Be are predicted for at least one of the road shape, road surface conditions, and obstacles present around the vehicle V1, which are included in the external data.
[0030] The trained model M may be input with the first drivable region Av1, the second drivable region Av2, and the third drivable region Av3, and the position Bs1 of the vehicle at the boundary between the first drivable region Av1 and the second drivable region Av2 may also be output as a trajectory generation parameter, in addition to the aforementioned endpoint position Be. Alternatively, the trained model M may be input with the boundary position of the first drivable region Av1 and the second drivable region Av2, and the first drivable region Av1 and the third drivable region Av3, instead of the second drivable region Av2 itself, and output as a trajectory generation parameter including the aforementioned endpoint position Be. In this case, the boundary position of the first drivable region Av1 and the second drivable region Av2 is input assuming the current position of the vehicle V1.
[0031] The trained model M is a model that takes data measured using GNSS 12 and sensors 14, such as the driving behavior of a vehicle driven by a driver, as input, and is trained to output trajectory generation parameters. This trained model M is pre-trained, for example, based on machine learning. In the case of supervised learning, the training data is, for example, a dataset in which the range of the correct physical quantities that the trained model M should output, the correct margin width Mw, and the correct endpoint position Be are labeled for the driving behavior of a vehicle driven manually by a driver and the surrounding environment of the vehicle. Furthermore, the trained model M may be a DNN (Deep Neural Network), CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), or a combination of these.
[0032] Furthermore, the trained model M may be a nominal model that has learned the driving behavior of vehicles driven by multiple drivers, and additional training on the driving behavior of the vehicle V1 driven by its own driver may be added to this nominal model. It is also desirable that the trained model M be a general-purpose model that can be applied to various road shapes and road surface conditions. However, in order to improve the prediction accuracy of trajectory generation parameters, it is also possible to have multiple models for situations with little temporal change, such as daytime and nighttime, or general roads and highways, and to use them according to the situation. In addition, in the case of supervised learning, it is desirable to exclude at least the driving behavior when the driver violates traffic rules or drives in a manner contrary to driving ethics from the training data.
[0033] In the following step S304, the trajectory generation unit 22d calculates the target path W and target speed Ve of the vehicle V1 based on the vehicle data and external data acquired in step S301, the first drivable area Av1 determined in step S302, at least one of the second drivable area Av2 or the third drivable area Av3, and the trajectory generation parameters predicted in step S303.
[0034] The trajectory generation unit 22d is, for example, a model that uses an optimization method to find the control variables that bring the vehicle as close as possible to the target for the given problem. The trajectory generation unit 22d may also be a rule-based model, a Bayesian estimation model, a maximum likelihood estimation model, or a combination of these. Furthermore, the trajectory generation unit 22d may be a model that connects the current position Bs of the vehicle V1 and the endpoint position Be with a multidimensional curve.
[0035] In step S305, the data output unit 22e outputs data of the vehicle V1's trajectory, consisting of the target path W and target speed Ve generated in step S304, to the driving control unit 23. This allows the driving control unit 23 to control the vehicle V1 with an appropriate target path W and target speed Ve, taking into account the relationship between moving objects and environmental elements around the vehicle.
[0036] <Example of processing by the trajectory planning unit 22> Figure 5 is a diagram illustrating the processing of steps S302 to S304, in other words, the first example of operation. In this example, a pedestrian P is present in the direction of travel of the vehicle V1. In this figure, the vehicle V1 is traveling on a straight path sandwiched between the first outer line Lm1 and the second outer line Lm2, and the pedestrian P is walking in the direction of arrow D near the second outer line Lm2 in front of the vehicle. This figure shows the target path W and target speed Ve within the drivable area Av of the vehicle V1 in this situation. The drivable area Av in this example is the area sandwiched between the first outer line Lm1 and the second outer line Lm2. The method for setting the first drivable area Av1 to the third drivable area Av3 is as described above. The first drivable area Av1 to the third drivable area Av3 shown in this figure are examples, and in reality they may be wider or narrower than these.
[0037] In this environment, the trajectory generation unit 22d generates a target path W that passes alongside the pedestrian P within the drivable area Av. The target speed Ve, margin width Mw, and endpoint position Be at this time are determined according to the path and speed learned by the trained model M. For example, if the model has learned to pass straight alongside the pedestrian P at a constant speed, the trajectory generation parameter prediction unit 22c outputs a prediction with speed Lim1 and an endpoint position Be1. The trajectory generation unit 22d then generates the target path W1 and target speed Ve1 based on these predictions.
[0038] Furthermore, if the learning process involves a route that moves away from pedestrian P, or if the vehicle speed is set to a speed Lim2 lower than speed Lim1 when the distance from pedestrian P is from distance Tr1 to distance Tr2, the processing of the trajectory generation parameter prediction unit 22c and the trajectory generation unit 22d is as follows: The trajectory generation parameter prediction unit 22c outputs a prediction that sets the speed from distance Tr1 to distance Tr2 to Lim2 and the margin width to Mm, and a prediction that sets the endpoint position to Be2. The trajectory generation unit 22d generates a target route W2 and a target speed Ve2 within the drivable area Av based on that prediction.
[0039] When parameters optimized only for specific scenes are used, only target speeds with speed limits of Lim1 or Lim2 can be generated. However, by generating a driving trajectory that matches human perception based on the driver's learning results from their normal driving, the comfort of the vehicle's occupants (V1) is not compromised in various scenes. Furthermore, when a low target speed (Ve2) is generated, a highly safe driving experience can be achieved that is less likely to cause a sense of danger to pedestrians (P).
[0040] Furthermore, in emergencies such as avoiding a collision with a pedestrian who suddenly steps into the road, drivers may need to brake suddenly or steer abruptly. Including such clearly abnormal driving behaviors in the training data would degrade the quality of the trained model M. Therefore, it is desirable to exclude clearly abnormal driving behaviors from the training data to prevent degradation of the trained model M's quality.
[0041] As explained above, according to this embodiment, when both environmental elements such as the outer edge of the roadway and moving objects such as pedestrians are detected, an appropriate driving trajectory can be generated that matches the driver's driving characteristics by using trajectory generation parameters such as speed, acceleration, and jerk output from a trained model M that has been trained using measurement data from manual driving.
[0042] <Second Operation Example> Figure 6 shows a second operation example. In this example, the vehicle V1 is traveling along an S-shaped curve enclosed by the first outer line Lm1 and the second outer line Lm2. The vehicle V1 is in a situation where it needs to decelerate before entering the curve. The target path W is shown in the figure, and the longitudinal acceleration obtained by differentiating the target velocity Ve with respect to time is shown at the bottom of Figure 6.
[0043] The trajectory generation unit 22d generates a target path W that passes through the drivable region Av. The position of the target path W in the drivable region Av depends on the data and learning method used to train the trained model M. The target path W shown in Figure 6 is just one example; the curvature of the target path W may be smaller than shown, or the target path W may pass through the center of the drivable region Av.
[0044] The target speed Ve is determined according to the deceleration start timing learned by the learned model M. For example, when learning is performed using data with a large number of instances of starting deceleration from a distance Tr3 close to the curve entrance, the following third acceleration target Lg3 is output by the learned model M. The third acceleration target Lg3 starts deceleration when the vehicle travels the distance Tr3 from the current position Bs. The trajectory generation unit 22d generates a target speed based on this third acceleration target Lg3.
[0045] Further, when learning is performed using data with a large number of instances of starting deceleration from a distance Tr4 that is closer to the host vehicle V1 than the distance Tr3, the following fourth acceleration target Lg4 is output by the learned model M. The fourth acceleration target Lg4 starts deceleration when the vehicle travels the distance Tr4 from the current position Bs. Then, the trajectory generation unit 22d generates a target speed based on this fourth acceleration target Lg4. By generating such a travel trajectory, it is possible to generate a travel trajectory with a high sense of security that matches the driving characteristics of the driver and the like.
[0046] <Third Operation Example> FIG. 7 is a diagram showing the third operation example. The host vehicle V1 in this example is traveling on an S-shaped curve surrounded by a first outer line Lm1 and a second outer line Lm2. FIG. 7 illustrates the target path W within the travelable area Av of the host vehicle V1 in a situation where the vehicle is traveling on the S-shaped curve. The difference from FIG. 6 is that it can be slightly recognized from the shape of the third travelable area Av3 that FIG. 7 shows a right curve. Under this environment, the trajectory generation unit 22d generates the target path W that passes through the inside of the travelable area Av. Note that the position where the target path W passes through the travelable area Av is determined by the content learned by the learned model M. For example, when a path that goes around the outside as much as possible at the entrance of a curve is learned, the target path W having the end point position Be shown in FIG. 7 is generated. By generating such a travel trajectory, it is possible to generate a travel trajectory with a high sense of security that matches the driving characteristics of the driver and the like.
[0047] <Fourth Operation Example> FIG. 8 is a diagram showing the fourth operation example. The host vehicle V1 in this example is traveling on the right lane side of a two-lane road on one side having a first outer line Lm1, a third boundary line Lm3, and a fourth center line Lm4. In addition, a preceding vehicle V2 is traveling ahead of the host vehicle V1 in the traveling direction. In this situation, the host vehicle V1 changes lanes to the left lane in the traveling direction to overtake the preceding vehicle V2. The broken line indicates the first path W1, and the solid line indicates the second path W2.
[0048] The trajectory generation unit 22d generates a target route for overtaking the preceding vehicle V2 within the drivable area Av. For example, when learning is performed using data that frequently passes near the preceding vehicle V2, the trajectory generation parameter prediction unit 22c outputs a trajectory generation parameter whose end point position is the first end point Be1 close to the preceding vehicle V2.
[0049] Further, for example, when learning is performed using data that frequently passes through positions distant from the preceding vehicle V2, the trajectory generation parameter prediction unit 22c outputs a trajectory generation parameter whose end point position is the second end point Be2 far from the preceding vehicle V2. The trajectory generation unit 22d generates the first path W1 and the second path W2 based on the output trajectory generation parameters. By generating such a travel trajectory, it is possible to generate a travel trajectory with a high sense of security that matches the driving characteristics of the driver and the like.
[0050] According to the second embodiment described above, the following effects can be obtained: (1) The in-vehicle system 1 includes a data acquisition unit 22a that acquires data relating to the vehicle V1, which is vehicle data, and external data relating to the surrounding environment of the vehicle V1, which includes a drivable area, which is an area on which the vehicle V1 can travel; an area determination unit 22b that determines a first drivable area Av1, which is an area on which the vehicle V1 will travel, with the direction of travel of the vehicle V1 being forward, and at least one of a second drivable area Av2 adjacent to the rear of the first drivable area Av1 or a third drivable area adjacent to the front of the first drivable area Av1; an area determination unit 22b that determines a first drivable area Av1, which is an area on which the vehicle will travel, and at least one of a second drivable area Av2 adjacent to the rear of the first drivable area Av1 or a third drivable area Av3 adjacent to the front of the first drivable area Av1; an area generation parameter prediction unit 22c that inputs the vehicle data, the first drivable area Av1, and at least one of the second drivable area Av2 or the third drivable area Av3 into a learned model M and outputs an area generation parameter used to generate an area on which the vehicle V1 will travel; and an area generation unit 22d that inputs the area generation parameter and generates an area on which the vehicle V1 will travel. Therefore, by using a pre-trained model M that has learned the driver's manual driving tendencies, it is possible to generate a vehicle trajectory that reflects the driver's driving characteristics.
[0051] (2) The trajectory generation parameter prediction unit 22c inputs the first drivable area Av1, the second drivable area Av2, and the third drivable area Av3 to the trained model M, and the trained model M outputs trajectory generation parameters including the position Bs1 of the vehicle at the boundary between the first drivable area Av1 and the second drivable area Av2, and the endpoint position Be, which is the position of the vehicle at the boundary between the first drivable area Av1 and the third drivable area Av3. Therefore, a trajectory can be generated using the position Bs1 and the endpoint position Be.
[0052] (3) The trajectory generation parameter prediction unit 22c inputs the first drivable area Av1 and the third drivable area Av3, and the position of the vehicle at the boundary between the first drivable area Av1 and the second drivable area Av2 into the trained model, and outputs trajectory generation parameters under the condition that the position of the vehicle at the boundary between the first drivable area Av1 and the second drivable area Av2 is assumed to be the current position of the vehicle V1. Therefore, a trajectory can be generated using the current position of the vehicle V1, which is assumed to be the endpoint position Be.
[0053] (4) The region determination unit 22b determines the third drivable region Av3 based on the range that the sensor 14 mounted on the vehicle V1 can recognize and the first drivable region Av1. For example, the region determination unit 22b determines the third drivable region Av3 to be the range obtained by subtracting the first drivable region Av1 from the range that the sensor 14 can recognize.
[0054] (5) The region determination unit 22b determines the second drivable region Av2 based on the second drivable region Av2 and the first drivable region Av1 from a predetermined time prior, and the first drivable region Av1.
[0055] (6) The trajectory generation parameter prediction unit 22c outputs trajectory generation parameters that include a margin width Mw which specifies the distance from the left and right edges of the drivable area to the trajectory on which the vehicle will travel.
[0056] (7) The trajectory generation parameter prediction unit 22c outputs trajectory generation parameters that include at least one of the vehicle's speed, acceleration, and jerk. Therefore, it can generate a trajectory that also takes speed into account.
[0057] (8) The trained model M is generated by learning using data acquired by the data acquisition unit during manual operation of the vehicle. Therefore, the trained model M reflects the driver's driving characteristics, and by using this trained model M, it is possible to generate a vehicle trajectory that reflects the driver's driving characteristics.
[0058] (9) The orbit generation unit 22d includes a model based on at least one of the following methods: rule-based, machine learning, optimization, Bayesian estimation, and maximum likelihood estimation.
[0059] (10) The in-vehicle system 1 includes a vehicle motion control device 2, a powertrain system 3 that controls the driving force of the vehicle V1, a brake system 4 that controls the braking force of the vehicle V1, and a steering system 5 that controls the steering angle of the wheels of the vehicle V1. The vehicle motion control device 2 controls the powertrain system 3, the brake system 4, and the steering system 5 so that the vehicle travels along the trajectory generated by the trajectory generation unit 22d.
[0060] —Second Embodiment— A second embodiment of the vehicle motion control device, in-vehicle system, and vehicle control method will be described with reference to Figure 9. In the following description, the same reference numerals are used for components that are the same as in the first embodiment, and the differences will be mainly explained. Points that are not specifically described are the same as in the first embodiment. This embodiment differs from the first embodiment mainly in that it changes the driving mode.
[0061] Figure 9 is a diagram showing the configuration of the operation management unit 21 in the second embodiment. The operation management unit 21 comprises a data acquisition unit 21a, a vehicle control unit 21b, and a data output unit 21c. The operation of the operation management unit 21 is as follows: First, the data acquisition unit 21a acquires data. The vehicle control unit 21b uses this data to select an operation mode. The data output unit 21c outputs the data acquired by the data acquisition unit 21a and the operation mode selected by the vehicle control unit 21b to the track planning unit 22 and the driving control unit 23.
[0062] The data acquisition unit 21a acquires data on obstacles present around the vehicle V1, including input from the user to the HMI unit 15 and data on obstacles present around the vehicle V1 acquired by the sensor 14. The data acquisition unit 21a may also estimate obstacles present around the vehicle V1 using the position of the vehicle V1 output by the GNSS 12 and map data stored in the map data storage unit 13.
[0063] The data acquired by the data acquisition unit 21a includes, for example, data used to identify the driving mode. The driving mode is, for example, one of the following: automatic driving mode, manual driving mode, switching mode, and support mode. Automatic driving mode is a mode in which the vehicle motion control device 2 controls the vehicle V1. Manual driving mode is a mode in which the occupant controls the vehicle V1 and the vehicle motion control device 2 does not participate in the control of the vehicle V1. Switching mode is a mode that controls the switching between manual driving by the occupant and the aforementioned automatic driving mode. Support mode is a mode that supports manual driving of the vehicle V1 by the occupant.
[0064] The vehicle control unit 21b selects an operating mode using the data acquired by the data acquisition unit 21a. For example, if the occupant selects automatic driving control in the HMI unit 15, and automatic driving is possible based on the conditions of the vehicle V1 and the surrounding environment, the vehicle control unit 21b selects the automatic driving mode. Otherwise, the vehicle control unit 21b selects manual driving or manual driving support control.
[0065] The vehicle control unit 21b can use the obstacle data acquired by the sensor 14 to understand the surrounding environment. Furthermore, the vehicle control unit 21b may combine the position of the vehicle V1 output by the GNSS 12 with the map data stored in the map data storage unit 13 to estimate obstacles present around the vehicle V1 obtained from the map data. The driving control unit 23 controls the vehicle V1 based on the operating mode selected by the vehicle control unit 21b and the driving trajectory generated by the trajectory generation unit 22d. However, the driving control unit 23 does not operate when the operating mode is manual driving mode.
[0066] According to the second embodiment described above, the following effects can be obtained. (11) The operation management unit 21 includes a vehicle control unit 21b that performs at least one of the following: automatic driving control of the vehicle based on the track generated by the track generation unit 22d, switching control between automatic driving and manual driving of the vehicle, and support control for manual driving by the driver of the vehicle. Therefore, the vehicle control unit 21b of the operation management unit 21 can achieve safe and unobtrusive vehicle control by selecting an appropriate operating mode according to the conditions of the vehicle V1 and the surrounding environment.
[0067] —Third Embodiment— A fourth embodiment of the vehicle motion control device, in-vehicle system, and vehicle control method will be described with reference to Figure 10. In the following description, the same reference numerals are used for components that are the same as in the first embodiment, and the differences will be mainly explained. Points that are not specifically described are the same as in the first embodiment. This embodiment differs from the first embodiment mainly in that learning is performed outside the in-vehicle system.
[0068] Figure 10 is a configuration diagram of the in-vehicle system 1B in the third embodiment. In this embodiment, the in-vehicle system 1B communicates with at least an external learning processing unit 6 via an external communication device 11. The external learning processing unit 6 is a computing device that communicates with the vehicle motion control device 2 via the external communication device 11. The external learning processing unit 6 may be a single physical server, or it may be a so-called cloud server realized by multiple computing resources connected to a network.
[0069] The external learning processing unit 6 acquires data about the vehicle V1 and the surrounding environment via the external communication device 11, and uses this data to train a pre-trained model M. This training may be for generating a pre-trained model M, or it may be additional training for a pre-trained model M already used in the vehicle motion control device 2. The external learning processing unit 6 then outputs the pre-trained model M to the vehicle motion control device 2 via the external communication device 11. The vehicle motion control device 2, upon receiving this output, uses the received pre-trained model M from then on.
[0070] Furthermore, the external learning processing unit 6 may output a trained model, which has been trained based on data such as the driving behavior of multiple vehicles or the driving behavior of multiple drivers, as a nominal model to the vehicle motion control device 2 as a trained model M. Alternatively, the external learning processing unit 6 may output a model, which has been further trained using data such as the driving behavior of the driver of its own vehicle V1, as a trained model M to the vehicle motion control device 2, based on the nominal model. In this case, a different trained model M will be generated for each vehicle.
[0071] By placing the external learning processing unit 6, which performs computationally intensive learning, outside the in-vehicle system 1, the performance required for the ECU implementing the in-vehicle system 1 can be reduced. This allows for a lower cost for the in-vehicle system 1. Furthermore, by performing additional learning based on the nominal model to match the driving characteristics of the driver of the vehicle V1, the automated driving control can be brought closer to matching the driving characteristics of the driver of the vehicle V1 each time data such as the driving behavior of the driver of the vehicle V1 is learned.
[0072] According to the third embodiment described above, the following effects can be obtained. (12) The data acquisition unit 22a communicates with an external learning processing unit 6 that outputs a trained model generated by learning using information acquired during manual driving. Therefore, by placing the external learning processing unit 6, which performs learning with a heavy processing load, outside the in-vehicle system 1, additional learning can be performed while reducing the performance required for the ECU that implements the in-vehicle system 1.
[0073] —Fourth Embodiment— A fourth embodiment of the vehicle motion control device, in-vehicle system, and vehicle control method will be described with reference to Figure 11. In the following description, the same reference numerals are used for components that are the same as in the first embodiment, and the differences will be mainly explained. Points that are not specifically described are the same as in the first embodiment. This embodiment differs from the first embodiment mainly in that the user can select to update the learned model.
[0074] Figure 11 is a diagram showing the configuration of the HMI unit 15 in the fourth embodiment. The HMI unit 15 comprises a data acquisition unit 15a, a user selection reception unit 15b, and a data output unit 15c. The data acquisition unit 15a acquires a new learned model M and model update data, which will be described later, from an external source. The user selection reception unit 15b presents the model update data to the occupant of the vehicle V1 and accepts the occupant's selection. The data output unit 15c outputs the learned model M to the trajectory planning unit 22 based on the occupant's selection.
[0075] Model update data refers to data relating to updates based on the learning results of a trained model M. Ideally, the model update data should be understandable to the occupants, showing changes before and after the update. Examples of model update data include the trained model M's accuracy in matching the driver's characteristics to the vehicle V1, travel time to the destination, changes in acceleration, and driving position.
[0076] The user selection reception unit is, for example, a display and a pointing device, which can display data and accept selections from the crew. As shown in the lower part of Figure 11, the user selection reception unit displays, for example, two trained models, one before and one after the update, and displays them in a way that allows for easy comparison of the individual fit rate and travel time to the destination for each model. The crew uses the pointing device to select their preferred model from the two trained models M. The model selected by the crew is then output to the trajectory planning unit 22 and used in subsequent processing.
[0077] In this way, the trajectory planning unit 22 generates a trajectory based on trajectory generation parameters predicted by a learned model M selected by the occupant of the vehicle V1 at the user selection reception unit 15b of the HMI unit 15, thereby realizing vehicle control that is highly comfortable and safe, tailored to the driver's driving characteristics.
[0078] According to the fourth embodiment described above, the following effects can be obtained: (13) The system includes a user selection reception unit 15b that displays information regarding the update content based on the learning results of the learned model M, and allows the occupant of the vehicle V1 to select whether or not to allow the update of the learned model M. As a result, a route that is more in line with the driver's preferences can be generated.
[0079] In the embodiments and modifications described above, the configuration of the functional blocks is merely an example. Several functional configurations shown as separate functional blocks may be integrated, or a configuration represented in one functional block diagram may be divided into two or more functions. Furthermore, some of the functions of each functional block may be provided by other functional blocks.
[0080] In each of the embodiments and modifications described above, the vehicle motion control device 2 is equipped with an input / output interface (not shown), and a program may be read from another device via the input / output interface and a medium available to the vehicle motion control device 2 when necessary. Here, the medium refers to, for example, a storage medium that can be attached to or detached from the input / output interface, or a communication medium, i.e., a wired, wireless, optical network, or a carrier wave or digital signal propagating through such network. Furthermore, some or all of the functions realized by the program may be realized by hardware circuits or FPGAs.
[0081] The embodiments and modifications described above may be combined in any way. Although various embodiments and modifications have been described above, the present invention is not limited to these. Other embodiments that can be conceivable within the scope of the technical idea of the present invention are also included within the scope of the present invention.
[0082] 1: In-vehicle system 2: Vehicle motion control device 6: External learning processing device 11: External communication device 14: Sensor 15b: User selection reception unit 21: Operation management unit 21b: Vehicle control unit 22: Driving trajectory planning unit 22a: Data acquisition unit 22b: Area determination unit 22c: Trajectory generation parameter prediction unit 22d: Trajectory generation unit 22e: Data output unit 23: Driving control unit
Claims
1. A vehicle motion control device comprising: a data acquisition unit that acquires data relating to the vehicle, which is vehicle data, and external data relating to the surrounding environment of the vehicle, which includes vehicle data relating to the vehicle and a drivable area, which is an area in which the vehicle can travel; an area determination unit that determines a first drivable area, which is an area for generating a trajectory on which the vehicle travels, with the direction of travel of the vehicle being forward, and at least one of a second drivable area adjacent to the rear of the first drivable area or a third drivable area adjacent to the front of the first drivable area; a trajectory generation parameter prediction unit that inputs the vehicle data, the first drivable area, and at least one of the second drivable area or the third drivable area into a trained model and outputs trajectory generation parameters used to generate a trajectory on which the vehicle travels; and a trajectory generation unit that inputs the trajectory generation parameters and generates a trajectory on which the vehicle travels.
2. A vehicle motion control device according to claim 1, wherein the trajectory generation parameter prediction unit inputs the first drivable region, the second drivable region, and the third drivable region to the learned model, and outputs the trajectory generation parameters including the position of the vehicle at the boundary between the first drivable region and the second drivable region, and the position of the vehicle at the boundary between the first drivable region and the third drivable region.
3. A vehicle motion control device according to claim 1, wherein the trajectory generation parameter prediction unit inputs the first drivable region and the third drivable region to the learned model, and outputs the trajectory generation parameters under the condition that the position of the vehicle at the boundary between the first drivable region and the second drivable region is assumed to be the current position of the vehicle.
4. A vehicle motion control device according to any one of claims 1 to 3, wherein the area determination unit determines the third drivable area based on the range recognizable by the external sensor mounted on the vehicle and the first drivable area.
5. A vehicle motion control device according to any one of claims 1 to 3, wherein the region determination unit determines the second drivable region based on the second drivable region and the first drivable region prior to a predetermined time, and the first drivable region.
6. A vehicle motion control device according to any one of claims 1 to 3, wherein the trajectory generation parameter prediction unit outputs the trajectory generation parameter including a margin width that specifies the distance from the left and right edges of the drivable area to the trajectory on which the vehicle travels.
7. A vehicle motion control device according to any one of claims 1 to 3, wherein the trajectory generation parameter prediction unit outputs the trajectory generation parameter, which includes at least one of the vehicle's speed, acceleration, and jerk.
8. A vehicle motion control device according to any one of claims 1 to 3, wherein the learned model is generated by learning using data acquired by the data acquisition unit during manual driving of the vehicle.
9. A vehicle motion control device according to any one of claims 1 to 3, wherein the trajectory generation unit comprises a model based on at least one of the following methods: rule-based, machine learning, optimization, Bayesian estimation, and maximum likelihood estimation.
10. A vehicle motion control device according to any one of claims 1 to 3, comprising a vehicle control unit that performs at least one of the following: automatic driving control of the vehicle based on a trajectory generated by the trajectory generation unit; switching control between automatic driving and manual driving of the vehicle; and support control for manual driving by the driver of the vehicle.
11. A vehicle motion control device according to any one of claims 1 to 3, wherein the data acquisition unit communicates with an external learning processing device that outputs the learned model generated by learning using information acquired during manual driving.
12. A vehicle motion control device according to any one of claims 1 to 3, comprising an information selection unit that displays information regarding update content based on the learning results of the learned model, and allows the occupant of the vehicle to select whether or not to allow the update of the learned model.
13. An on-board system comprising: a vehicle motion control device according to any one of claims 1 to 3; a powertrain system for controlling the driving force of the vehicle; a brake system for controlling the braking force of the vehicle; and a steering system for controlling the steering angle of the wheels of the vehicle, wherein the vehicle motion control device controls the powertrain system, the brake system, and the steering system so that the vehicle travels along a trajectory generated by the trajectory generation unit.
14. A vehicle control method executed by a computer, comprising: an information acquisition step of acquiring data relating to the vehicle, which is data relating to the vehicle itself, and external data relating to the surrounding environment of the vehicle, which is data relating to the vehicle itself, including vehicle data relating to the vehicle itself and a drivable area, which is an area in which the vehicle itself can travel; an area determination step of determining a first drivable area, which is an area for generating a trajectory on which the vehicle itself travels, with the direction of travel of the vehicle itself being forward, and at least one of a second drivable area adjacent to the rear of the first drivable area or a third drivable area adjacent to the front of the first drivable area; a trajectory generation parameter prediction step of inputting the vehicle data, the first drivable area, and at least one of the second drivable area or the third drivable area into a trained model and outputting trajectory generation parameters used to generate a trajectory on which the vehicle itself travels; and a trajectory generation step of generating a trajectory on which the vehicle itself travels using the trajectory generation parameters.
