Methods for adaptive speed control

Abstract

An adaptive speed control method for a vehicle may include performing a multitude of measurements of the vehicle's surrounding environment using a vehicle perception sensor. The vehicle's surrounding environment includes a multitude of distant vehicles. The method may further include determining one or more speed decision algorithm inputs, at least partially based on the multitude of measurements. The method may further include determining an optimal vehicle speed, at least partially based on the inputs of the one or more speed decision algorithm. The method may further include controlling the vehicle to maintain the optimal vehicle speed.

Description

[0001] The present description refers to control units and procedures for the speed control of vehicles.

[0002] To enhance occupant alertness and comfort, vehicles can be equipped with Advanced Driver Assistance Systems (ADAS). ADAS systems can use various sensors, such as cameras, radar, and LiDAR, to detect and identify objects in the vehicle's surroundings, including other vehicles, pedestrians, road configurations, traffic signs, and road markings. Based on the environmental conditions around the vehicle, ADAS systems can take action, such as applying the brakes and / or warning a vehicle occupant. ADAS systems can also provide assistive functions, such as adaptive cruise control. Adaptive cruise control is a type of ADAS function that automatically adjusts the vehicle's speed to maintain an ideal distance from vehicles ahead.Adaptive cruise control can use sensor information from various sensors, such as cameras, radar, and LiDAR, to measure the distance between the vehicle and a lead vehicle and adjust the vehicle's acceleration and / or braking to maintain the tracking distance. However, current ADAS systems and adaptive cruise control may not take into account the behavior and trajectories of other distant vehicles close to the lead vehicle.

[0003] DE 10 2019 122 757 A1 describes a vehicle speed control system in which a first vehicle speed target is determined on the basis of a speed of a second vehicle and at least one factor of a road curvature, a road inclination, a road cross slope angle, visibility or a coefficient of friction on a road surface.

[0004] DE 10 2016 225 855 A1 describes a method for operating at least one motor vehicle, wherein the motor vehicle's speed is predetermined depending on at least one other vehicle located in its path. It describes how the driving behavior of several other vehicles in the vicinity of the motor vehicle is recorded in order to detect a traffic jam, and how, upon detection of a traffic jam, the motor vehicle's speed is limited to a maximum speed depending on the recorded driving behavior of the other vehicles.

[0005] DE 10 2015 120 996 A1 describes a vehicle system comprising a sensor that detects the speed of at least one nearby vehicle and outputs a speed signal representing the speed of that vehicle. The vehicle system further includes a processing unit programmed to determine a target speed based on the speed signal output by the sensor. The processing unit generates a command signal to control a host vehicle in accordance with the target speed.

[0006] While ADAS and adaptive cruise control systems and procedures fulfill their purpose, there is a need for improvement. It can be considered a task to specify a new and improved adaptive cruise control procedure for a vehicle.

[0007] The problem is solved by a method according to claim 1. Furthermore, a system is described with which the invention can be implemented in an application case.

[0008] A method according to the invention for adaptive speed control of a vehicle is described. The method comprises performing a plurality of measurements of the vehicle's environment using a vehicle perception sensor. The environment surrounding the vehicle includes a plurality of distant vehicles. The method further comprises determining one or more speed decision algorithm inputs, at least partially based on the plurality of measurements. The method further comprises determining an optimal vehicle speed for the vehicle, at least partially based on the inputs of the one or more speed decision algorithm. The method further comprises controlling the vehicle so that the vehicle speed corresponds to the optimal vehicle speed.Performing the multitude of measurements further includes measuring the position of each of the multitude of remote vehicles using the vehicle perception sensor. Performing the multitude of measurements further includes storing the position of each of the multitude of remote vehicles and the time at which the position of each of the multitude of remote vehicles was measured in non-volatile memory. Performing the multitude of measurements further includes repeating the measurement step and the storage step at a predetermined interval in order to accumulate a multitude of position measurements over time for each of the multitude of remote vehicles. Performing the multitude of measurements further includes determining a trajectory for each of the multitude of remote vehicles, at least partially based on the multiple position measurements over time.Determining the optimal vehicle speed further involves training a multitude of pre-trained Gaussian mixture models (GMMs) using a multitude of training data. Each of the multitude of pre-trained GMMs is associated with one of the multitude of trained optimal speeds. Determining the optimal vehicle speed further involves training a GMM for the current observation, based at least partially on the trajectory of each of the multitude of distant vehicles. Determining the optimal vehicle speed further involves comparing the current observation GMM with the multitude of pre-trained GMMs. Determining the optimal vehicle speed further involves selecting a suitable pre-trained GMM from the multitude of pre-trained GMMs that best matches the current observation GMM.Determining the optimal vehicle speed also includes determining the optimal vehicle speed at least partially based on one of the many trained optimal speeds associated with the appropriate pre-trained GMM.

[0009] In one embodiment, determining the optimal vehicle speed further includes receiving a desired speed range from a vehicle occupant. Determining the optimal vehicle speed further includes determining the optimal vehicle speed at least partially based on one of the several trained optimal speeds associated with the appropriate pre-trained GMM and the desired speed range.

[0010] In one embodiment, determining the optimal vehicle speed further comprises training a Gaussian mixture model (GMM) for the current observation, which is based at least partially on the trajectory of each of the plurality of distant vehicles. Determining the optimal vehicle speed further comprises receiving a desired speed range from a vehicle occupant. Determining the optimal vehicle speed further comprises executing a machine learning algorithm to determine the optimal speed. The machine learning algorithm for determining the optimal speed is configured to receive the current observation GMM, the desired speed range, and the inputs of one or more speed decision algorithm inputs as inputs and to provide the optimal vehicle speed as output.

[0011] In one embodiment, determining the one or more speed decision algorithm inputs further comprises determining the speed of a lead vehicle from among the multitude of distant vehicles, at least partially based on the multitude of measurements. The lead vehicle is located directly in front of the vehicle and in the same lane as the vehicle. Determining the one or more speed decision algorithm inputs further comprises determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the speed of the lead vehicle.

[0012] In one embodiment, determining the one or more speed decision algorithm inputs further comprises determining the speed of one or more neighboring vehicles from the plurality of distant vehicles, at least partially based on the plurality of measurements. The one or more neighboring vehicles are located within a predetermined radius around the vehicle. The one or more neighboring vehicles are located in a different lane than the vehicle. Determining the one or more speed decision algorithm inputs further comprises determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the speed of the one or more neighboring vehicles.

[0013] In one embodiment, determining the one or more speed decision algorithm inputs further includes determining a following distance of a trailing vehicle from the multitude of distant vehicles, at least partially based on the multitude of measurements. The trailing vehicle is located directly behind the vehicle and in the same lane as the vehicle. The following distance is the distance between the trailing vehicle and the vehicle. Determining the one or more speed decision algorithm inputs further includes determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs comprise at least the following distance.

[0014] In one embodiment, determining the one or more speed decision algorithm inputs further includes determining the current lane of the vehicle, at least partially, based on the multitude of measurements. Determining the one or more speed decision algorithm inputs further includes determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs comprise at least the current lane of the vehicle.

[0015] In one embodiment, determining the one or more speed decision algorithm inputs further comprises receiving one or more occupant inputs from a vehicle occupant. Determining the one or more speed decision algorithm inputs further comprises determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the one or more occupant inputs.

[0016] In one application, an adaptive speed control system for a vehicle is described, with which the method according to the invention can be carried out. The system can comprise a vehicle perception sensor, an automated driving system, and a control unit electrically connected to the vehicle perception sensor and the automated driving system. The control unit is programmed to perform a plurality of measurements in the vehicle's environment using the vehicle perception sensor. The environment surrounding the vehicle includes a plurality of distant vehicles. The plurality of measurements includes a trajectory of each of the plurality of distant vehicles. The control unit is further programmed to determine one or more speed decision algorithm inputs, at least partially, based on the plurality of measurements.The control unit is further programmed to determine an optimal vehicle speed, at least partially, based on one or more speed decision algorithm inputs. The control unit is also programmed to control the vehicle using the automated driving system to maintain the optimal vehicle speed.

[0017] In this use case, the control unit for determining the optimal vehicle speed is further programmed to train a current observation GMM, which is based at least partially on the trajectory of each of the multiple remote vehicles and the inputs of one or more speed decision algorithm inputs. To determine the optimal vehicle speed, the control unit is further programmed to compare the current observation GMM with a multiple of pretrained GMMs. Each of the multiple pretrained GMMs is associated with one of multiple trained optimal speeds. To determine the optimal vehicle speed, the control unit is further programmed to select a suitable pretrained GMM from the multiple of pretrained GMMs that best matches the current observation GMM.In order to determine the optimal vehicle speed, the control unit is further programmed to determine the optimal vehicle speed at least partially on the basis of one of the many trained optimal speeds that is associated with the appropriate pre-trained GMM.

[0018] In this use case, the control unit for determining the optimal vehicle speed is further programmed to receive a desired speed range from a vehicle occupant. To determine the optimal vehicle speed, the control unit is also programmed to determine the optimal vehicle speed, at least partially, based on one of several trained optimal speeds associated with the appropriate pre-trained GMM and the desired speed range.

[0019] In this use case, the control unit for determining the optimal vehicle speed is further programmed to train a GMM for the current observation, which is based at least partially on the trajectory of each of the multitude of distant vehicles. To determine the optimal vehicle speed, the control unit is further programmed to receive a desired speed range from a vehicle occupant. To determine the optimal vehicle speed, the control unit is further programmed to execute a machine learning algorithm to determine the optimal speed.The machine learning algorithm for determining the optimal speed is configured to receive the current observation GMM, the desired speed range, and the inputs of one or more speed decision algorithm inputs as inputs and to provide the optimal vehicle speed as output.

[0020] In this application, the control unit for determining the one or more speed decision algorithm inputs is further programmed to determine the speed of a lead vehicle from among a multitude of distant vehicles, at least partially, based on the multitude of measurements. The lead vehicle is located directly in front of the vehicle and in the same lane as the vehicle. To determine the one or more speed decision algorithm inputs, the control unit is further programmed to determine the speed of one or more neighboring vehicles from among the multitude of distant vehicles, at least partially, based on the multitude of measurements. The one or more neighboring vehicles are located within a predetermined radius around the vehicle. The one or more neighboring vehicles are located in a different lane than the vehicle.To determine one or more speed decision algorithm inputs, the control unit is further programmed to determine, at least partially, the following distance of a trailing vehicle from the multitude of distant vehicles based on the multitude of measurements. The trailing vehicle is located directly behind the vehicle and in the same lane. The following distance is the distance between the trailing vehicle and the vehicle. To determine one or more speed decision algorithm inputs, the control unit is further programmed to determine, at least partially, the current lane of the vehicle based on the multitude of measurements.To determine the one or more speed decision algorithm inputs, the control unit is further programmed to determine the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the speed of the lead vehicle, the speed of the one or more adjacent vehicles, the following distance and the current lane.

[0021] In this application, the control unit for determining the one or more speed decision algorithm inputs is further programmed to receive one or more occupant inputs from a vehicle occupant. The one or more occupant inputs comprise at least one desired speed range. To determine the one or more speed decision algorithm inputs, the control unit is further programmed to determine the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the one or more occupant inputs.

[0022] In this application, the control unit for controlling the vehicle using the automated driving system is further programmed to continuously monitor the one or more occupant inputs. To control the vehicle using the automated driving system, the control unit is also programmed to instruct the automated driving system to control the acceleration and deceleration of the vehicle, at least partially, based on the optimal vehicle speed and the one or more occupant inputs. Fig. Figure 1 is a schematic representation of an adaptive speed control system for a vehicle; Fig. Figure 2 is a schematic diagram of an environment with a roadway and a large number of distant vehicles; Fig. Figure 3 is a flowchart of a procedure for adaptive speed control for a vehicle; Fig. Figure 4 is a flowchart of a first procedure for determining an optimal vehicle speed; and Fig. Figure 5 is a flowchart of a second procedure for determining the optimal vehicle speed.

[0023] In aspects of this description, adaptive speed control systems and methods are used to enable semi-autonomous driving of a vehicle, for example, when traveling on a highway. This description presents a new and improved adaptive speed control system and method for a vehicle, including the determination of an optimal vehicle speed based on several factors, including the trajectories and behavior of several distant vehicles in the vicinity.

[0024] In Fig. Figure 1 is an adaptive cruise control system for a vehicle, generally designated by reference number 10. System 10 is illustrated with an example vehicle 12. Although a passenger car is depicted, vehicle 12 can be any type of vehicle without this altering the scope of this description. System 10 generally comprises a control unit 14, a vehicle perception sensor 16, and an automatic driving system 18.

[0025] The control unit 14 is used to implement a method 100 for adaptive speed control of a vehicle, as described below. The control unit 14 comprises at least one processor 20 and a non-volatile, computer-readable device or medium 22. The processor 20 may be a custom or commercially available processor, a central processing unit (CPU), a graphics processing unit (GPU), an auxiliary processor among several processors associated with the control unit 14, a semiconductor-based microprocessor (in the form of a microchip or chipset), a macroprocessor, a combination thereof, or generally an instruction-executing device.

[0026] The computer-readable devices or media 22 can include volatile and non-volatile memory, for example, read-only memory (ROM), random-access memory (RAM), and keep-alive memory (KAM). KAM is a persistent or non-volatile memory that can be used to store various operating variables while the processor 20 is turned off. The computer-readable memory device or media 22 can be implemented using a variety of memory devices such as PROMs (programmable read-only memory), EPROMs (electrical PROMs), EEPROMs (electrically erasable PROMs), flash memory, or other electrical, magnetic, optical, or combined memory devices capable of storing data, some of which are executable instructions used by the control unit 14 to control various systems of the vehicle 12.

[0027] The control unit 14 can also consist of several control units that are electrically interconnected. The control unit 14 can be associated with additional systems and / or control units of the vehicle 12, so that the control unit 14 can access data such as speed, acceleration, braking and steering angle of the vehicle 12.

[0028] The control unit 14 is electrically connected to the vehicle perception sensor 16. In an exemplary embodiment, the electrical communication is established, for example, via a CAN network, a FLEXRAY network, a local area network (e.g., WiFi, Ethernet, and the like), a serial peripheral interface network (SPI), or the like. It is understood that various additional wired and wireless techniques and communication protocols for communicating with the control unit 14 fall within the scope of this description. Within the scope of this description, electrical communication also includes the transfer of power and / or energy between electrical devices (e.g., using conductive wires and / or wireless power transfer techniques).

[0029] The vehicle perception sensor 16 is used to detect objects in an environment 26 around the vehicle 12. Furthermore, the vehicle perception sensor 16 is used to measure distances between multiple objects in the environment 26 and distances between the vehicle 12 and objects in the environment 26. In an exemplary embodiment, the vehicle perception sensor 16 comprises at least one of the following components: a camera 28, a LiDAR (Light Detection and Ranging) sensor 30, and an ultrasonic sensor 32.

[0030] The camera 28 is used to capture images and / or videos of the vehicle 12's surroundings 26. In one exemplary embodiment, the camera 28 comprises one or more cameras that provide a view of the vehicle 12's surroundings 26. In a non-limiting example, the camera 28 is mounted inside the vehicle 12, for example, on the vehicle's headliner or windshield, with a view through the windshield. In another non-limiting example, the camera 28 is mounted on the outside of the vehicle 12, for example, on the vehicle's roof, and provides a view of the surroundings 26.

[0031] In another exemplary embodiment, the camera 28 comprises a surround-view camera system with a plurality of cameras (also known as satellite cameras) arranged to provide a view of the surroundings 26 on all sides of the vehicle 12. In a non-limiting example, the camera system includes a forward-facing camera (for example, mounted in a radiator grille of the vehicle 12), a rear-facing camera (for example, mounted on a tailgate of the vehicle 12), and two side-facing cameras (for example, mounted below each of the two side mirrors of the vehicle 12). In a further non-limiting example, the camera system also includes an additional reversing camera mounted near a high-mounted center brake light of the vehicle 12.

[0032] It is understood that camera systems with additional cameras and / or additional mounting locations fall within the scope of this description. It should further be understood that the camera 28 may comprise a stereoscopic camera with distance measurement capabilities, an infrared camera, a thermal imaging camera, and / or any other type of camera or image sensor device, without this deviating from the scope of this description. As described above, the camera 28 is electrically connected to the control unit 14.

[0033] The LiDAR sensor 30 is used for remote sensing and environmental mapping. The LiDAR sensor 30 operates by emitting laser pulses and measuring the time it takes for the laser pulses to return to the LiDAR sensor 30 after striking objects. In an exemplary embodiment, the LiDAR sensor 30 comprises a LiDAR laser source, a LiDAR scanner or mirror, a LiDAR photodetector, and a LiDAR time-of-flight measurement system. In a non-restrictive example, the LiDAR laser source emits laser pulses that travel into the target area, and the LiDAR scanner deflects these pulses in various directions. The emitted laser pulses interact with objects in the environment, and their reflections are detected by the LiDAR photodetector.The LiDAR time-of-flight measurement system calculates the distance to the objects based on the time between the emission of the laser pulses by the LiDAR laser source and the reception of the reflected laser pulses by the LiDAR photodetector. The LiDAR sensor 30 is electrically connected to the control unit 14, as described above.

[0034] The ultrasonic sensor 32 is used to measure distances in the vicinity 26 of the vehicle 12. In one exemplary embodiment, the ultrasonic sensor 32 comprises an ultrasonic transducer and an ultrasonic receiver. The ultrasonic transducer emits ultrasonic pulses, and the ultrasonic receiver captures the ultrasonic pulses after they have been reflected by an object. In a non-limiting example, the ultrasonic transducer emits ultrasonic waves that are reflected by the object and return to the ultrasonic receiver. The ultrasonic sensor 32 measures the time it takes for the ultrasonic pulses to travel to and from the ultrasonic transducer and calculates a distance based on the speed of sound in a given medium (for example, air). The ultrasonic sensor 32 is electrically connected to the control unit 14, as described above.

[0035] It should be understood that the foregoing discussion of the camera 28, the LiDAR sensor 30 and the ultrasonic sensor 32 is merely exemplary and that the vehicle perception sensor 16 may include any number of additional or alternative sensors, including, for example, radar sensors and / or the like, without this deviating from the scope of the present description.

[0036] The automated driving system 18 is used to assist a vehicle occupant, to enhance the occupant's awareness, and / or to control the behavior of the vehicle 12. For the purposes of this description, the automated driving system 18 includes systems that provide any level of assistance to the occupant (for example, blind spot warning, lane keeping assist, and / or similar), as well as systems capable of autonomously driving the vehicle 12 under some or all conditions (for example, automatic lane keeping, adaptive cruise control, fully autonomous driving, and / or similar). It is understood that all levels of driving automation defined, for example, by SAE J3016 (i.e., SAE Level 0, SAE Level 1, SAE Level 2, SAE Level 3, SAE Level 4, and SAE Level 5) fall within the scope of this description.

[0037] In one exemplary embodiment, the automated driving system 18 is configured to detect and / or receive information about the environment 26 surrounding the vehicle 12 and to process this information to assist the occupant. In some embodiments, the automated driving system 18 is a software module running in the control unit 14. In other embodiments, the automated driving system 18 comprises a separate control unit for the automated driving system, similar to the control unit 14, which is capable of processing the information about the environment 26 surrounding the vehicle 12. In one exemplary embodiment, the automated driving system 18 can operate in a manual operating mode, a semi-automated operating mode, and a fully automated operating mode.

[0038] Within the context of this description, manual operating mode means that the automated driving system 18 warns or notifies the occupant but does not directly intervene in or control the vehicle 12. In a non-restrictive example, the automated driving system 18 receives information from the vehicle perception sensor 16. Using techniques such as computer vision, the automated driving system 18 understands the vehicle 12's surroundings 26 and provides assistance to the occupant. For example, if the automated driving system 18 detects, based on data from the vehicle perception sensor 16, that the vehicle 12 is likely to collide with a distant vehicle, the automated driving system 18 can use a display to warn the occupant.

[0039] Within the scope of this description, the semi-automated operating mode means that the automated driving system 18 can warn or notify the occupants and, in certain situations, directly intervene in or control the vehicle 12. In a non-restrictive example, the automated driving system 18 is also in electrical communication with components of the vehicle 12, such as a braking system, a drive system, and / or a steering system, so that the automated driving system 18 can control the behavior of the vehicle 12. In a non-restrictive example, the automated driving system 18 can control the behavior of the vehicle 12 by applying the vehicle's brakes to avoid an imminent collision. In another non-restrictive example, the automated driving system 18 can control the vehicle's steering system to provide an automatic lane-keeping function.In another, non-restrictive example, the automated driving system 18 can control the braking system, the drive system, and the steering system of the vehicle 12 to temporarily drive the vehicle 12 to a predetermined destination. However, occupant intervention may be required at any time. In an exemplary embodiment, the automated driving system 18 can include additional components, such as an eye-monitoring device, configured to monitor the occupant's level of attention and ensure that the occupant is ready to take control of the vehicle 12.

[0040] Within the context of this description, the fully automatic operating mode means that the automated driving system 18 uses the data from the vehicle perception sensor 16 to understand the environment 26 and to control the vehicle 12 to drive to a predetermined destination without requiring any control or intervention by the occupant.

[0041] The automated driving system 18 operates with a path planning algorithm configured to generate a safe and efficient trajectory for the vehicle 12 to navigate its surrounding environment. In one exemplary embodiment, the path planning algorithm is a machine learning algorithm trained to output control signals for the vehicle 12 based on input data acquired by the vehicle perception sensor 16. In another exemplary embodiment, the path planning algorithm is a deterministic algorithm programmed to output control signals for the vehicle 12 based on data acquired by the vehicle perception sensor 16.

[0042] In a non-restrictive example, the path planning algorithm generates a sequence of waypoints or a continuous path that the vehicle 12 should follow to reach a destination while complying with rules, regulations, and safety restrictions. The sequence of waypoints or the continuous path is created, at least in part, based on a detailed map and the current state of the vehicle 12 (that is, its position, speed, and orientation). The detailed map contains, for example, information about lane boundaries, road geometry, speed limits, traffic signs, and / or other relevant features. In an exemplary embodiment, the detailed map is stored in the media 22 of the control unit 14 and / or in a remote database or server.In another exemplary embodiment, the path planning algorithm performs perception and mapping tasks to interpret the data collected by the vehicle perception sensor 16 and to create, update and / or extend the detailed map.

[0043] It is understood that the automated driving system 18 can contain any software and / or hardware module configured to operate in manual mode, semi-automated mode, or fully automated mode as described above. The automated driving system 18 is electrically connected to the control unit 14 as described above.

[0044] Fig. Figure 2 shows a schematic diagram of the environment 26 with a roadway 40 and a plurality of remote vehicles 42. In an exemplary embodiment, the roadway 40 comprises a first lane 44a, a second lane 44b, a third lane 44c, and a fourth lane 44d. In a non-restrictive example, vehicle 12 is located in the third lane 44c. The plurality of remote vehicles 42 includes a lead vehicle 42a, which is located directly in front of vehicle 12 in the same lane as vehicle 12 (i.e., in the third lane 44c). The plurality of remote vehicles 42 further includes one or more adjacent vehicles 42b, which are located within a predetermined radius 46 around vehicle 12 and in a different lane than vehicle 12 (i.e., the first lane 44a, the second lane 44b, or the fourth lane 44d).The multitude of distant vehicles 42 also includes a following vehicle 42c, which is located directly behind vehicle 12 in the same lane as vehicle 12 (that is, the third lane 44c). A following distance 48 is defined as the distance between the following vehicle 42c and vehicle 12.

[0045] Fig. Figure 3 shows a flowchart of the method 100 for adaptive speed control of a vehicle. The method 100 begins at block 102 and proceeds to block 104. In block 104, the control unit 14 uses the vehicle perception sensor 16 to measure the position of each of the plurality of distant vehicles 42. In a non-restrictive example, the control unit 14 uses the LiDAR sensor 30 to measure the position of each of the plurality of distant vehicles 42 relative to the vehicle 12. In another non-restrictive example, the control unit 14 uses the camera 28 to capture multiple images of the plurality of distant vehicles 42 and processes the multiple images using a computer vision algorithm to determine the position of each of the plurality of distant vehicles 42 relative to the vehicle 12.It is understood that the control unit 14 can continue to use the vehicle perception sensor 16 to perceive, measure, and / or understand additional relevant objects in the environment 26, including, for example, road signs, road edges, lane markings, and / or the like. After block 104, procedure 100 proceeds to block 106.

[0046] In block 106, the control unit 14 stores the position of each of the plurality of remote vehicles 42, measured in block 104, in the media 22 of the control unit 14. The control unit 14 also stores the time at which the position of each of the plurality of remote vehicles 42 was measured in the media 22. In an exemplary embodiment, the control unit 14 repeats the measurement step in block 104 and the storage step in block 106 at a predetermined interval (for example, every second) to accumulate a plurality of position measurements over time for each of the plurality of remotely controlled vehicles 42 in the media 22.

[0047] In an exemplary embodiment, if the environment 26 is sparsely populated with vehicles, historical data for lane 40, acquired using crowdsourcing and / or a process similar to blocks 104 and 106, can be used to accumulate a multitude of historical position measurements over time for average vehicles on lane 40. This multitude of historical position measurements over time can be used in subsequent steps of procedure 100 instead of the multitude of position measurements over time for each of the multitude of distant vehicles 42. After block 106, procedure 100 proceeds to block 108.

[0048] In block 108, the control unit 14 determines a trajectory for each of the plurality of remote vehicles 42. In an exemplary embodiment, the trajectory of each of the plurality of remote vehicles 42 comprises at least a position, a velocity (i.e., a first derivative of the position), an acceleration (i.e., a second derivative of the position), and a jerk (i.e., a third derivative of the position) for each of the plurality of remote vehicles 42. It is understood that the trajectory may also include additional higher-order derivatives of the position without this deviating from the scope of the present description. In an exemplary embodiment, to determine the trajectory in block 106, the control unit 14 analyzes the multiple position measurements over time for each of the plurality of remote vehicles 42 that are stored in the medium 22.In a non-restrictive example, the trajectory of each of the plurality of remote vehicles 42 is stored in the media 22 and continuously updated in block 104 upon receipt of updated position measurements. Following block 108, the procedure 100 proceeds through blocks 110, 112, 114, 116, and 118 to determine one or more speed decision algorithm inputs. For the purposes of this description, the one or more speed decision algorithm inputs are data, information, or observations used to determine an optimal speed for the vehicle 12, as further explained below.

[0049] Referring again to Fig. 2 with continued reference to Fig. 3. In block 110, the control unit 14 determines a speed of the lead vehicle 42a, at least partially, based on the trajectory of each of the plurality of remote vehicles 42 (and thus the trajectory of the lead vehicle 42a). The speed of the lead vehicle 42a is one of the speed decision algorithm inputs relevant for determining the optimal vehicle speed. For example, the optimal vehicle speed should be determined to be less than or equal to the speed of the lead vehicle 42a in order to maintain a comfortable tracking distance between the lead vehicle 42a and vehicle 12. After block 110, the procedure 100 proceeds to block 120, as will be explained in more detail below.

[0050] In block 112, the control unit 14 determines the speed of one or more neighboring vehicles 42b, at least partially, based on the trajectory of each of the plurality of distant vehicles 42 (and thus the trajectory of the one or more neighboring vehicles 42b). The speed of the one or more neighboring vehicles 42b is one of the one or more speed decision algorithm inputs relevant for determining the optimal vehicle speed. For example, the optimal vehicle speed should be determined such that the time the vehicle 12 remains within a blind spot of the one or more neighboring vehicles 42b is minimized.In another example, the optimal vehicle speed should be determined so that it is relatively similar to the speed of one or more adjacent vehicles 42b, such that the optimal vehicle speed corresponds to the traffic flow on lane 40. Following Block 112, the procedure 100 proceeds to Block 120, as explained in more detail below.

[0051] In block 114, the control unit 14 determines the following distance 48 of the following vehicle 42c, at least partially, based on the trajectory of each of the plurality of distant vehicles 42 (and thus the trajectory of the following vehicle 42c, for example, the position of the following vehicle 42c relative to vehicle 12). The speed of the following vehicle 42c is one of the speed decision algorithm inputs relevant for determining the optimal vehicle speed. For example, if the following distance 48 is below a certain threshold, it can be assumed that the following vehicle 42c will collide with vehicle 12, so the optimal vehicle speed can be increased over time to reduce the likelihood of a collision.It is understood that the following distance threshold can be a fixed threshold (for example, twenty meters) or a variable threshold that depends on the vehicle speed (for example, the vehicle speed in kilometers per hour divided by 1.2 to obtain a following distance threshold of three seconds in meters). Following Block 114, Procedure 100 proceeds to Block 120, as explained in more detail below.

[0052] In block 116, the control unit 14 determines the current lane of the vehicle 12 (for example, the third lane 44c, as in Fig. 2 shown), which is based at least partially on measurements taken with the vehicle perception sensor 16 and / or additional vehicle sensors. In one exemplary embodiment, the control unit 14 uses data on lane lines on the roadway 40, measured by the vehicle perception sensor 16, to determine the current lane. In another exemplary embodiment, the control unit 14 uses the trajectory of each of the plurality of distant vehicles 42 (for example, the position of vehicle 12 relative to the plurality of distant vehicles 42) to determine the current lane. In yet another exemplary embodiment, the control unit 14 uses additional vehicle sensor measurements (for example, measurements from the global navigation satellite system (GNSS)) to determine the current lane.

[0053] The current lane of vehicle 12 is one of the speed decision algorithm inputs relevant for determining the optimal vehicle speed. For example, the optimal vehicle speed is determined, at least in part, based on the regulated or generally accepted use of the current lane. If the current lane is an overtaking lane (that is, a lane intended for overtaking slower vehicles), the optimal vehicle speed should be relatively higher. If the current lane is a slow lane (that is, a lane intended for slower vehicles), the optimal vehicle speed should be relatively lower. After Block 116, Procedure 100 transitions to Block 120, as explained in more detail below.

[0054] In block 118, the control unit 14 receives one or more occupant inputs from a vehicle occupant. In an exemplary embodiment, the one or more occupant inputs include at least one desired speed range. For the purposes of this description, the desired speed range is a driving speed range for the vehicle 12 requested by the occupant. In some examples, the desired speed range is defined between fixed limits (for example, between ninety and one hundred kilometers per hour). In some examples, the desired speed range is defined in relation to the speeds of the surrounding traffic (that is, the multitude of distant vehicles 42). In some examples, the desired speed range is defined in relation to a posted speed limit on the roadway 40.

[0055] In a non-restrictive example, the desired speed range can be represented as slower than the surrounding traffic (for example, 10 km / h slower than the average speed of the surrounding traffic), appropriate to the surrounding traffic (for example, equal to the average speed of the surrounding traffic), or faster than the surrounding traffic (for example, 10 km / h faster than the average speed of the surrounding traffic). The desired speed range can also include a minimum and / or maximum speed acceptable to the occupant. In some examples, the minimum and / or maximum speed is defined as fixed values ​​relative to the desired speed range (for example, plus or minus 10 km / h relative to the desired speed range).In some examples, the minimum and / or maximum speed is defined as percentages in relation to the desired speed range (for example, plus or minus ten percent of the desired speed range).

[0056] In one exemplary embodiment, the one or more occupant inputs are received via an accelerator and / or brake pedal of the vehicle 12. In a non-limiting example, actuation of the brake pedal indicates a decrease in the desired speed range or a deactivation of the system 10, while actuation of the accelerator pedal indicates an increase in the desired speed range. In another exemplary embodiment, the one or more occupant inputs are received via physical and / or software buttons or other devices located in an occupant compartment of the vehicle 12 (for example, steering wheel buttons, turn signal lever, foot pedals, gearshift lever, and / or the like). In a non-limiting example, the one or more occupant inputs are received via electrical and / or electromechanical buttons on the steering wheel of the vehicle 12.In another, non-restrictive example, one or more occupant inputs are received via speech recognition (i.e., receiving and interpreting voice commands given by a vehicle occupant), gesture recognition (i.e., capturing and interpreting physical gestures performed by a vehicle occupant), and / or similar methods.

[0057] In another exemplary embodiment, the one or more occupant inputs are received via the occupant's interaction with a display in the vehicle 12 (for example, an instrument display, an infotainment display, and / or the like). In a non-limiting example, the display provides a user interface that prompts the occupant to select the desired speed range and the minimum and / or maximum speed. It is understood that additional systems and methods for interacting with the occupant to provide the control unit 14 with the desired speed range and the minimum and / or maximum speed are also within the scope of this description. The one or more occupant inputs are one of the one or more speed decision algorithm inputs relevant for determining the optimal vehicle speed.After block 118, procedure 100 transitions to block 120.

[0058] In block 120, the control unit 14 trains a Gaussian mixture model (GMM) for the current observation, which is based at least partially on the trajectory of each of the plurality of remote vehicles 42 determined in block 108 and the inputs of one or more speed decision algorithm inputs determined in blocks 110, 112, 114, 116, and 118. Within the scope of this description, the GMM for the current observation is a probabilistic model for grouping and representing the plurality of remote vehicles 42 based on the trajectory of each of the plurality of remote vehicles 42 determined in block 108 and the one or more speed decision algorithm inputs. In an exemplary embodiment, the GMM for the current observation contains a mixture of a plurality of Gaussian distributions.

[0059] Each of the multiple Gaussian distributions is defined by a mean vector, a covariance matrix, and a set of mixing coefficients that determine the weight of each of the multiple Gaussian distributions in the GMM of the current observation. The GMM for the current observation is trained, for example, with an expectation-maximization (EM) algorithm that iteratively adjusts the mean vector, covariance matrix, and mixing coefficients of each of the multiple Gaussian distributions until the GMM for the current observation optimally represents the environment 26 based on the trajectory of each of the plurality of distant vehicles 42 and the one or more speed decision algorithm inputs.

[0060] In a non-restrictive example, subgroups within the multitude of distant vehicles 42 are identified that exhibit a similarity in the inputs of the trajectory decision algorithm and / or velocity decision algorithm. Each subgroup is represented by one of several Gaussian distributions, and the mean vector, covariance matrix, and mixture coefficients of each of the several Gaussian distributions are iteratively adjusted until the current observational GMM optimally represents the environment 26. After block 120, the procedure 100 proceeds to block 122.

[0061] In block 122, the control unit 14 determines the optimal vehicle speed for the vehicle 12. Methods for determining the optimal vehicle speed for the vehicle 12 are discussed in more detail below. After block 122, the method 100 transitions to block 124. In block 124, the control unit 14 controls the vehicle so that the vehicle speed corresponds to the optimal vehicle speed determined in block 122. In an exemplary embodiment, the control unit 14 uses the automated driving system 18 to control the acceleration and deceleration of the vehicle 12 in order to maintain the optimal vehicle speed.In another exemplary embodiment, the control unit 14 continuously monitors the one or more occupant inputs and uses the automated driving system 18 to control the acceleration and deceleration of the vehicle 12 in order to maintain the optimal vehicle speed, at least partially, based on the one or more occupant inputs. If the occupant presses the brake pedal of the vehicle 12, the control unit 14 commands the automated driving system 18 to reduce the vehicle speed, or it deactivates the automated driving system 18 so that the occupant controls the vehicle speed manually.

[0062] In one exemplary embodiment, the control unit 14 proposes increased levels of automation based at least partially on the inputs of the one or more speed decision algorithm inputs, the one or more occupant inputs, and / or the trajectory of each of the plurality of remote vehicles 42. In a non-restrictive example, the control unit 14 proposes to activate the semi-automated operating mode of the automated driving system 18 when the current lane is a slow lane in order to autonomously perform a lane change maneuver. In another non-restrictive example, the control unit 14 proposes to activate the semi-automated operating mode of the automated driving system 18 when the occupant activates a turn signal of the vehicle 12 in order to autonomously perform a lane change maneuver.In a non-restrictive example, the control unit 14 uses a display or other human interface device (HID) to prompt and / or inform the occupant before entering a higher level of automation. Following block 124, the procedure 100 in block 126 enters a standby state.

[0063] In one exemplary embodiment, the control unit 14 repeatedly exits the standby state 126 and restarts the procedure 100 in block 102. In a non-restrictive example, the control unit 14 exits the standby state 126 and restarts the procedure 100 according to a timer, for example, every three hundred milliseconds.

[0064] In Fig. Figure 4 is a flowchart of a first exemplary embodiment 122a of Block 122 (that is, a first method for determining the optimal vehicle speed). The first exemplary embodiment 122a of Block 122 begins in Block 402. In Block 402, a plurality of pretrained Gaussian Mixing Models (GMMs) are trained. In an exemplary embodiment, the plurality of pretrained GMMs are trained using a plurality of training data. In a non-restrictive example, the plurality of training data comprises a plurality of sets of training vehicle trajectories and training speed decision algorithm inputs. Each of the plurality of training vehicle trajectories and training speed decision algorithm inputs is labeled with one of a plurality of trained optimal speeds.

[0065] In an exemplary embodiment, the multitude of training data is collected by a training vehicle (not shown) equipped with sensors similar to those of vehicle 12 (i.e., the vehicle perception sensor 16) and occupied by a human passenger. As the training vehicle travels, it collects the multitude of sets of training vehicle trajectories and training speed decision algorithm inputs and records a trained optimal speed driven by the passenger, corresponding to each set of training vehicle trajectories and training speed decision algorithm inputs. It should be understood that additional methods for obtaining the multitude of training data, including, for example, crowdsourcing, computer simulation, and / or the like, are within the scope of this description.

[0066] In one exemplary embodiment, each of the multiple pre-trained GMMs is trained using sets of training vehicle trajectories and training speed decision algorithm inputs corresponding to one of the multiple trained optimal speeds. In a non-restrictive example, the plurality of pre-trained GMMs includes a first pre-trained GMM trained for a first optimal speed (for example, 100 kilometers per hour), a second pre-trained GMM trained for a second optimal speed (for example, 110 kilometers per hour), and a third pre-trained GMM trained for a third optimal speed (for example, 120 kilometers per hour). Thus, each of the multiple pre-trained GMMs is associated with one of the multiple trained optimal speeds.

[0067] In an exemplary embodiment, each of the multiple pre-trained GMMs contains a mixture of several Gaussian distributions. Each of the multiple Gaussian distributions is defined by a mean vector, a covariance matrix, and a set of mixing coefficients that determine the weight of each of the multiple Gaussian distributions in each of the multiple pre-trained GMMs. Each of the multiple pre-trained GMMs is trained, for example, with an expectation-maximization (EM) algorithm that iteratively adjusts the mean vector, covariance matrix, and mixing coefficients of each of the multiple Gaussian distributions until each of the multiple pre-trained GMMs optimally represents each set of training vehicle trajectories and training speed decision algorithm inputs, each labeled with one of the multiple trained optimal speeds.

[0068] In a non-restrictive example, subgroups within each set of training vehicle trajectories and training speed decision algorithm inputs are identified that exhibit similarities in the trajectories and / or speed decision algorithm inputs. Each subgroup is represented by one of several Gaussian distributions, and the mean vector, covariance matrix, and mixture coefficients of each of the several pretrained GMMs represent precisely each set of training vehicle trajectories and training speed decision algorithm inputs that are labeled with one of the several trained optimal speeds.

[0069] In one exemplary embodiment, the plurality of pre-trained GMMs is trained using the control unit 14 of system 10. In another exemplary embodiment, the plurality of pre-trained GMMs is trained using an external system, such as a cloud-based server or other device. In a non-restrictive example, the plurality of pre-trained GMMs is transferred to the control unit 14 and stored in the media 22 of the control unit 14. Following Block 402, the first exemplary embodiment 122a transitions from Block 122 to Block 404.

[0070] In block 404, the control unit 14 selects a suitable pre-trained GMM from the multitude of GMMs pre-trained in block 402. In an exemplary embodiment, the control unit 14 compares the current observation GMM trained in block 120 with the multitude of pre-trained GMMs trained in block 402 to select the suitable pre-trained GMM. In an exemplary embodiment, the control unit 14 uses probabilistic comparison methods to compare the current observation GMM with the multitude of pre-trained GMMs, such as the Kullback-Leibler divergence (KL), the Bhattacharyya distance, the Hellinger distance, the Earth Mover's Distance (EMD) / Wasserstein distance, the likelihood ratio test, the overlap integral, component fitting, entropy-based measures, the log likelihood comparison, the log likelihood ratio test, and / or similar methods.The matching pre-trained GMM is selected from the multitude of pre-trained GMMs that most closely matches the current observation GMM. For the purposes of this description, the highest match means that the matching pre-trained GMM exhibits the highest statistical correlation with the current observation GMM compared to any of the multiple pre-trained GMMs. Following Block 404, the first exemplary embodiment 122a transitions from Block 122 to Block 406.

[0071] In block 406, the control unit 14 determines the optimal vehicle speed, at least partially, based on the matching pre-trained GMM. In one exemplary embodiment, the optimal vehicle speed is determined as the one from the multitude of trained optimal speeds associated with the matching pre-trained GMM. In another exemplary embodiment, the optimal vehicle speed is determined, at least partially, based on the desired speed range determined in block 118. In a non-restrictive example, the occupant is notified if one of the multiple trained optimal speeds associated with the matching pre-trained GMM deviates significantly from the desired speed range (for example, by a difference greater than or equal to 20 percent).In another, non-restrictive example, system 10 is deactivated if one of the several trained optimal speeds associated with the matching pre-trained GMM deviates significantly from the desired speed range. Following block 406, the first exemplary embodiment 122a of block 122 is completed, and procedure 100 continues as described above.

[0072] In Fig.Figure 5 shows a flowchart of a second exemplary embodiment 122b of Block 122 (that is, a second method for determining the optimal vehicle speed). The second exemplary embodiment 122b of Block 122 begins in Block 502. In Block 502, the control unit 14 executes a machine learning algorithm for determining the optimal speed to determine the optimal vehicle speed. In a non-restrictive example, the machine learning algorithm for determining the optimal speed comprises several layers, including an input layer and an output layer, as well as one or more hidden layers. The input layer receives the GMM trained in Block 120 for the current observation, the desired speed range determined in Block 118, and the inputs determined in Blocks 110-118 for one or more speed decision algorithm inputs as inputs.The inputs are then passed to the hidden layers. Each hidden layer applies a transformation (for example, a nonlinear transformation) to the data and passes the result to the next hidden layer, until the last hidden layer. The output layer provides the optimal vehicle speed.

[0073] To train the machine learning algorithm for determining the optimal speed, a dataset containing inputs and their corresponding optimal vehicle speeds is used. The algorithm is trained by adjusting the internal weights between nodes in each hidden layer to minimize the prediction error. During training, an optimization procedure (for example, gradient descent) is used to further adjust the internal weights and reduce the prediction error. The training process is repeated with the entire dataset until the prediction error is minimized, and the trained model is then used to process new input data.

[0074] After sufficient training of the machine learning algorithm for determining the optimal speed, the algorithm is able to determine the optimal vehicle speed based on the GMM trained in Block 120 for the current observation, the desired speed range determined in Block 118, and the speed decision algorithm inputs determined in Blocks 110-118. By adjusting the weights between the nodes in each hidden layer during training, the algorithm "learns" to recognize patterns in the input data that indicate an optimal vehicle speed. After Block 502, the second exemplary embodiment 122b of Block 122 is completed, and Procedure 100 continues as described above.

[0075] System 10 and Method 100 of the present description offer several advantages. By considering the speed of the lead vehicle 42a, the speed of one or more adjacent vehicles 42b, the following distance of the trailing vehicle 42c, the current lane, and one or more occupant inputs when determining the optimal vehicle speed for vehicle 12, System 10 and Method 100 ensure comprehensive consideration of the conditions surrounding the vehicle. By using Gaussian mixing models to characterize the trajectories of the multitude of distant vehicles 42, large amounts of data are efficiently aggregated and presented statistically for interpretation or further processing by the machine learning algorithm for determining the optimal speed.Overall, the system 10 and the method 100 of the present description enable the determination of an optimal vehicle speed for the vehicle 12 on the basis of several factors, including the behavior of the multitude of distant vehicles 42 in the vicinity of the vehicle 12.

Claims

[1] Method (100) for adaptive speed control for a vehicle (12), wherein the method (100) comprises: Performing a plurality of measurements of an environment (26) surrounding the vehicle (12) using a vehicle perception sensor (16), wherein the environment (26) surrounding the vehicle (12) includes a plurality of distant vehicles (42); Determining one or more velocity decision algorithm inputs, at least partially, based on the multitude of measurements; Determining an optimal vehicle speed for the vehicle (12) at least partially based on the one or more speed decision algorithm inputs; and Steering the vehicle (12) so that the vehicle speed corresponds to the optimal vehicle speed; the performance of the multitude of measurements further comprising: Measuring the position of each of the multitude of remote vehicles (42) using the vehicle perception sensor (16); Storing the position of each of the multiple remote vehicles (42) and a time at which the position of each of the multiple remote vehicles (42) was measured in a non-volatile memory (22); Repeating the measurement step and the storage step at a predetermined interval to accumulate a multitude of position measurements over time for each of the multitude of remote vehicles (42); and Determining a trajectory of each of the plurality of remote vehicles (42) at least partially on the basis of the multiple position measurements over time; and further comprising determining the optimal vehicle speed: Training a variety of pre-trained Gaussian mixture models (GMMs) using a variety of training data, wherein each of the variety of pre-trained GMMs is associated with one of a variety of trained optimal velocities; Training a GMM for the current observation at least partially on the basis of the trajectory of each of the multitude of distant vehicles (42); Comparing the GMM of the current observation with the multitude of pre-trained GMMs; Selecting a suitable pre-trained GMM from the multitude of pre-trained GMMs that best matches the current observation GMM; and Determining the optimal vehicle speed, at least partially, based on one of the many trained optimal speeds associated with the appropriate pre-trained GMM. [2] Method (100) according to claim 1, wherein determining the optimal vehicle speed further comprises: Receiving a desired speed range from a vehicle occupant; and Determining the optimal vehicle speed, at least partially, based on one of several trained optimal speeds associated with the appropriate pre-trained GMM and the desired speed range. [3] Method (100) according to claim 1, wherein determining the optimal vehicle speed further comprises: Training a Gaussian mixture model, GMM, for the current observation at least partially on the basis of the trajectory of each of the multitude of distant vehicles (42); Receiving a desired speed range from a vehicle occupant; and Executing a machine learning algorithm to determine the optimal speed, wherein the machine learning algorithm for determining the optimal speed is configured to receive the current observation GMM, the desired speed range and the one or more speed decision algorithm inputs as inputs and to provide the optimal vehicle speed as output. [4] Method (100) according to claim 1, wherein determining the one or more speed decision algorithm inputs further comprises: Determining the speed of a lead vehicle (42a) from the plurality of distant vehicles (42) at least partially on the basis of the plurality of measurements, wherein the lead vehicle (42a) is directly in front of the vehicle (12) and on the same lane (44a-d) as the vehicle (12); and Determining one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the speed of the lead vehicle (42a). [5] Method (100) according to claim 1, wherein determining the one or more speed decision algorithm inputs further comprises: Determining the speed of one or more adjacent vehicles (42b) from the plurality of distant vehicles (42) at least partially on the basis of the plurality of measurements, wherein the one or more adjacent vehicles (42b) are located within a predetermined radius (46) around the vehicle (12) and wherein the one or more adjacent vehicles (42b) are located in a different lane (44a-d) than the vehicle (12); and Determining the one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the speed of the one or more adjacent vehicles (42b). [6] Method (100) according to claim 1, wherein determining one or more speed decision algorithm inputs further comprises: Determining a following distance (48) of a following vehicle (42c) from the plurality of distant vehicles (42) at least partially on the basis of the plurality of measurements, wherein the following vehicle (42c) is directly behind the vehicle (12) and in the same lane (44a-d) as the vehicle (12) and wherein the following distance (48) is a distance between the following vehicle (42c) and the vehicle (12); and Determining the one or more velocity decision algorithm inputs, wherein the one or more velocity decision algorithm inputs include at least the tracking distance (48). [7] Method (100) according to claim 1, wherein determining the one or more speed decision algorithm inputs further comprises: Determining a current lane (44a-d) of the vehicle (12) at least partially on the basis of the multitude of measurements; and Determining one or more speed decision algorithm inputs, wherein the one or more speed decision algorithm inputs include at least the current lane (44a-d) of the vehicle (12). [8] Method (100) according to claim 1, wherein determining the one or more speed decision algorithm inputs further comprises: Receiving one or more occupant inputs from a vehicle occupant; and Determining one or more velocity decision algorithm inputs, wherein the one or more velocity decision algorithm inputs include at least one or more occupant inputs.

Images