Apparatus, system and method for planning intersection turn

Abstract

The invention relates to a device, a system and a method for planning intersection turning. Devices, systems, and methods disclosed relate to trajectory planning. An example apparatus includes one or more processors, the one or more processors are operable to receive instructions to turn within an intersection, plan a trajectory for an autonomous vehicle to turn within the intersection based on map data including one or more optimal turn paths associated with the intersection, and direct the autonomous vehicle to follow the trajectory to pass the intersection. The one or more optimal turn paths are generated based on turn paths of the one or more human-driven vehicles.

Description

Technical Field

[0001] The present disclosure relates to devices, systems, and methods for vehicle control, and more particularly, to devices, systems, and methods for vehicle control by planning turns at intersections. Background Art

[0002] Intersections pose challenges for vehicles as they compete for space within the same area. An intersection involves multiple lanes, oncoming traffic, and traffic signs, which require determination of the right of way. Relying on real-time decision-making at intersections can lead to uncertainty. Therefore, in order to reduce potential traffic risks and enhance the efficiency of navigating intersections, it is necessary to establish planned turns at intersections for vehicles. Summary of the Invention

[0003] In one embodiment, a device for trajectory planning includes one or more processors operable to receive an instruction to turn within an intersection, plan a trajectory for an autonomous vehicle to turn within the intersection based on map data including one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on the turning paths of one or more human-driven vehicles, and guide the autonomous vehicle to follow the trajectory to pass through the intersection.

[0004] In another embodiment, a method for trajectory planning includes receiving an instruction to turn within an intersection, planning a trajectory for an autonomous vehicle to turn within the intersection based on map data including one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on the turning paths of one or more human-driven vehicles, and guide the autonomous vehicle to follow the trajectory to pass through the intersection.

[0005] These and additional features provided by the embodiments of the present disclosure will be more thoroughly understood in conjunction with the following detailed description and the accompanying drawings. Brief Description of the Drawings

[0006] The embodiments illustrated in the drawings are substantially illustrative and exemplary and are not intended to limit the disclosure. The following detailed description of the illustrative embodiments can be understood in conjunction with the following drawings, in which like structures are indicated by like reference numerals, where:

[0007] Figure 1 Schematically depicts an example system for trajectory planning for intersection turning of the present disclosure according to one or more embodiments shown and described herein;

[0008] Figure 2 Schematically depicts example components of a device and system for trajectory planning for intersection turning of the present disclosure according to one or more embodiments shown and described herein;

[0009] Figure 3 An illustrative block diagram depicting the generation of a planned trajectory at an intersection of the present disclosure in accordance with one or more embodiments shown and described herein; and

[0010] Figure 4 A flowchart depicting illustrative steps for generating a planned intersection turn of the present disclosure in accordance with one or more embodiments shown and described herein. DETAILED DESCRIPTION

[0011] The disclosed embodiments include apparatuses, systems, and methods for trajectory planning at an intersection based on map data including one or more optimal turn paths. The optimal turn paths may be generated based on the turn paths of one or more human-driven vehicles. The disclosed embodiments include apparatuses, systems, and methods that assist an autonomous vehicle in planning a trajectory to pass through an intersection. An autonomous vehicle relying on real-time decision-making at an intersection may have difficulty navigating through the intersection, resulting in an undesired wide turn that causes the autonomous vehicle to extend beyond the intersection. Such wide turns may cause traffic flow disruptions and increase the risk of collisions with other vehicles, pedestrians, or cyclists within the intersection.

[0012] By encoding the optimal turn paths of human-driven vehicles into the intersections within the map data, an autonomous vehicle can turn more naturally and desirably within the intersection. The disclosed embodiments also include apparatuses, systems, and methods having artificial intelligence capabilities to continuously improve the optimal turn paths based on the generated planned trajectories. Thus, the disclosed apparatuses, systems, and methods help alleviate problems associated with navigating a vehicle through an intersection (such as wide turns that may exceed the intersection), thereby improving overall traffic flow and vehicle interactions. Additionally, the artificial intelligence capabilities enhance the system's ability to ensure that the trajectory planning system evolves, adapts to changes in traffic dynamics, and contributes to the efficient and desired safe navigation of intersections by autonomous vehicles, and to ensure that the trajectory planning remains robust and responsive to evolving road conditions, traffic patterns, and other dynamic factors by continuously improving the optimal turn paths based on the generated planned trajectories. The trajectory planning system not only prioritizes safe and natural maneuvers but also demonstrates efficiency in its use of map data. Integrating the optimal turn paths into the map data facilitates precise decision-making for autonomous vehicles, leveraging spatial information to effectively navigate intersections.

[0013] Various embodiments of methods and systems for trajectory planning are described in more detail herein. Whenever possible, the same reference numerals will always be used in the drawings to refer to the same or similar parts. As used herein, the singular forms "a", "an", and "the" include plural referents unless the context clearly dictates otherwise. Thus, a reference to "a" component includes aspects having two or more such components unless the context indicates otherwise.

[0014] Referring to the accompanying drawings, Figure 1 and Figure 2 an example trajectory planning system 100 is schematically depicted. The trajectory planning system 100 may include one or more controllers 201, and the one or more controllers 201 may also include one or more modules, such as a trajectory generation module 222 and an optimal turning path module 232.

[0015] The trajectory generation module 222 may include one or more first machine learning algorithms, such as a first neural network 322. The trajectory generation module 222 may generate a planned trajectory 105. The planned trajectory 105 may include an upcoming turning path and the velocity, time, and kinematic characteristics of the vehicle 101 associated with the upcoming turning path.

[0016] The optimal turning path module 232 may include one or more second machine learning algorithms, such as a second neural network 432. The second machine learning algorithm may generate one or more optimal turning paths 315 associated with an intersection 135 (as Figure 3 illustrated). Information about the optimal turning paths 315 and the intersection 135 may be stored in the map data 217. The map data 217 may include crosswalks, traffic lights, traffic signs, obstacles, road lanes, road edges, shoulders, medians, road markings, poles, or combinations thereof. In an embodiment, the map data 217 may be high-definition (HD) map data or standard-definition (SD) map data. An SD map containing SD map data is based on the curves, elevations, and coordinates of the road. The SD map may have a resolution of meters. An HD map containing HD map data may include more details than an SD map containing SD map data. The HD map may have a resolution ranging from sub-meter to centimeter scale. The HD map data may include all the information included in the SD map data and more information, such as road shape, road markings, traffic signs, poles, guardrails, walls, and obstacles.

[0017] The trajectory planning system 100 may include one or more vehicles 101, which may be autonomous vehicles. In some embodiments, one or more controllers 201 are included in the vehicle 101. In some embodiments, some of the vehicles 101 may include a communication device (such as vehicle network interface hardware) that operably communicates wirelessly with the controller 201. In some embodiments, the controller 201 may be included in one or more servers that include a server communication device (such as network interface hardware 206) that operably communicates with the vehicle 101.

[0018] Each of the vehicles 101 may be an automobile or any other passenger or non-passenger vehicle, such as, for example, land, waterborne, and / or airborne vehicles. Each of the vehicles 101 may be an autonomous vehicle that navigates its environment with limited human input or no human input. Each of the vehicles 101 may drive on a road and perform vision-based lane centering, for example, using one or more sensors. Each of the vehicles 101 may include brakes for driving the vehicle, such as motors, engines, or any other power system. The vehicle 101 may move on various surfaces, such as, but not limited to, roads, highways, streets, freeways, bridges, tunnels, parking lots, garages, off-road tracks, railways, or any surface on which a vehicle may operate.

[0019] In an embodiment, the vehicle 101 may move on a road 137 that includes one or more intersections 135. The intersection 135 may include one or more lanes. The intersection 135 may include traffic signs, signal lights, roundabouts, and other structures that control traffic flow. The intersection 135 may be four-way, cross, three-way (such as T-intersections and Y-intersections), or five-way or more-way. For example, as Figure 1 illustrated, the road 137 may include a vertical road and a horizontal road. The horizontal road may include two-way lanes, namely, a southbound lane 121 and a northbound lane 122. The horizontal road may include two-way lanes, a westbound lane 131 and an eastbound lane 132. The vertical road and the horizontal road may intersect at the intersection 135. The intersection 135 may include one or more curbs 125 at the junction of the vertical road and the horizontal road.

[0020] Reference Figure 2 illustrates an example component of the controller 201. Although Figure 2FIG. illustrates a controller 201, but in some embodiments, the trajectory planning system 100 may include two or more controllers 201. The controller 201 or the vehicle 101 may include one or more vision sensors 208 and vehicle sensors 212. The vision sensors 208 may be used to capture images or videos of the environment around the vehicle 101. In some embodiments, one or more vision sensors 208 include one or more imaging sensors configured to operate in the visible and / or infrared spectra to sense visible and / or infrared light. Additionally, although the specific embodiments described herein are described in terms of hardware for sensing light in the visible and / or infrared spectra, it is understood that other types of sensors are contemplated. For example, the systems described herein may include one or more LIDAR sensors, radar sensors, sonar sensors, or other types of sensors for collecting data that may be incorporated into or supplement the data collection described herein. A range sensor such as radar may be used to obtain approximate depth and rate information of the view of the vehicle 101. One or more vision sensors 208 may include a forward camera mounted in the vehicle 101. One or more vision sensors 208 may be an array of sensing devices capable of detecting radiation in the ultraviolet, visible, or infrared bands. One or more vision sensors 208 may have any resolution. In some embodiments, one or more optical components, such as mirrors, fisheye lenses, or any other type of lens, may be optically coupled to one or more vision sensors 208. In the embodiments described herein, one or more vision sensors 208 may provide image data to one or more processors 204 or another component communicatively coupled to the communication path 203. In some embodiments, one or more vision sensors 208 may also provide navigation support. That is, the data captured by one or more vision sensors 208 may be used to automatically or semi-automatically navigate the vehicle.

[0021] The controller 201 or the vehicle 101 may include one or more vehicle sensors 212. Each of the one or more sensors 212 is coupled to the communication path 203 and communicatively coupled to one or more processors 204. One or more vehicle sensors 212 may include one or more rate sensors or motion sensors for detecting and measuring the motion and changes in motion of the vehicle (e.g., vehicle 101). The motion sensors may include an inertial measurement unit. Each of the one or more motion sensors may include one or more accelerometers and one or more gyroscopes. Each of the one or more motion sensors converts the sensed physical movement of the vehicle into a signal indicative of the direction, rotation, speed, or acceleration of the vehicle. The data obtained from the vehicle sensors 212 may be used to determine the vehicle kinematic characteristics of the vehicle 101.

[0022] The vision sensor 208 and the vehicle sensor 212 can be used to collect vehicle control data, road condition data, and vehicle kinematic data. The vehicle control data, road condition data, and vehicle dynamics data can be used to monitor the actual trajectory of the vehicle 101 passing through the intersection 135. The vehicle control data can include the throttle position, brake state, steering angle, and gear selection of the vehicle 101. The road condition data can include the road type, friction coefficient, and surface irregularities (e.g., bumps). The vehicle kinematic data can include the speed, acceleration, position, and direction of the vehicle 101.

[0023] The controller 201 can include one or more processors 204. Each of the one or more processors 204 can be any device capable of executing machine-readable and executable instructions. The instructions can be in the form of a set of machine-readable instructions stored in the data storage component 207 and / or the memory component 202. Thus, each of the one or more processors 204 can be a controller, integrated circuit, microchip, computer, or any other computing device. The one or more processors 204 are coupled to a communication path 203 that provides signal interconnectivity between the various modules of the system. Thus, the communication path 203 can communicatively couple any number of processors 204 to each other and allow the modules coupled to the communication path 203 to operate in a distributed computing environment. Specifically, each of the modules can operate as a node that can send and / or receive data. As used herein, the term "communicatively coupled" means that the coupled components are capable of exchanging data signals with each other, such as, for example, electrical signals via a conductive medium, electromagnetic signals via air, optical signals via an optical waveguide, and so on.

[0024] Thus, the communication path 203 can be formed from any medium capable of transmitting signals, such as, for example, wires, conductive traces, optical waveguides, etc. In some embodiments, the communication path 203 can facilitate the transmission of wireless signals, such as WiFi, Bluetooth®, near field communication (NFC), and so on. Additionally, the communication path 203 can be formed from a combination of media capable of transmitting signals. In one embodiment, the communication path 203 includes a combination of conductive traces, wires, connectors, and buses that cooperate to allow the transmission of electrical data signals to components such as processors, memories, sensors, input devices, output devices, and communication devices. Thus, the communication path 203 can include a vehicle bus, such as, for example, a LIN bus, a CAN bus, a VAN bus, and so on. Additionally, it should be noted that the term "signal" means a waveform (e.g., electrical, optical, magnetic, mechanical, or electromagnetic) capable of propagating through a medium, such as DC, AC, sine wave, triangular wave, square wave, vibration, and so on.

[0025] The controller 201 may include one or more memory components 202 coupled to a communication path 203. The one or more memory components 202 may include RAM, ROM, flash memory, a hard disk drive, or any device capable of storing machine-readable and executable instructions such that the machine-readable and executable instructions can be accessed by one or more processors 204. The machine-readable and executable instructions may include logic or (one or more) algorithms written in any programming language in any generation (e.g., 1GL, 2GL, 3GL, 4GL, or 5GL), such as machine language that can be directly executed by a processor, or assembly language, object-oriented programming (OOP), scripting language, microcode, etc. that can be compiled or assembled into machine-readable and executable instructions and stored in the one or more memory components 202. Alternatively, the machine-readable and executable instructions may be written in a hardware description language (HDL), such as logic implemented via field-programmable gate array (FPGA) configuration or application-specific integrated circuit (ASIC) or their equivalents. Thus, the methods described herein may be implemented as pre-programmed hardware elements or as a combination of hardware and software components in any conventional computer programming language.

[0026] The one or more memory components 202 may include one or more modules, including a trajectory generation module 222 and an optimal turning path module 232. Each of the one or more modules may include, but is not limited to, routines, subroutines, programs, objects, components, data structures, etc. for performing specific tasks or operating on specific data types as will be described below. The data storage component 207 may store map data including intersections 135, roads 137, and optimal turning paths 315. The data storage component 207 may also store training data 227 for training the trajectory generation module 222 and the optimal turning path module 232. The training data 227 may include standard ground truth data related to trajectory planning at one or more intersections for human-driven vehicles and autonomous vehicles. The data storage component 207 may store historical data 237, such as historical vehicle kinematic data, historical vehicle control data, historical road condition data, historical trajectory data, and other historical data related to the operation of the vehicle 101, and (e.g., as Figure 3 illustrated) historical sensor data of the human-driven vehicle 301. The trajectory generation module 222 and the optimal turning path 232 may also be stored in the data storage component 207 during or after operation.

[0027] One or more modules including the trajectory generation module 222 and the optimal turning path 232 may include one or more machine learning algorithms, such as neural networks. Machine learning capabilities may be provided via and to a neural network training module as described herein. By way of example and not limitation, a neural network may utilize one or more artificial neural networks (ANNs). In an ANN, the connections between nodes may form a directed acyclic graph (DAG). An ANN may include node inputs, one or more hidden activation layers, and node outputs, and may be utilized with activation functions in one or more of the hidden activation layers, such as linear functions, step functions, logistic (Sigmoid) functions, tanh functions, rectified linear (ReLu) functions, or combinations thereof. The ANN is trained by applying such activation functions to a training dataset to determine an optimized solution from adjustable weights and biases applied to nodes within the hidden activation layers, thereby generating one or more outputs as an optimized solution with minimized error. In machine learning applications, new inputs (such as one or more generated outputs) may be provided to the ANN model as training data to continue to improve accuracy and minimize the error of the ANN model. One or more ANN models may utilize one-to-one, one-to-many, many-to-one, and / or many-to-many (e.g., sequence-to-sequence) sequence modeling. One or more ANN models may employ a combination of artificial intelligence techniques, such as but not limited to, deep learning, random forest classifiers, feature extraction from audio, images, clustering algorithms, or combinations thereof. In some embodiments, a convolutional neural network (CNN) may be utilized. For example, a convolutional neural network (CNN) may be used as an ANN in the field of machine learning, such as a type of deep, feedforward ANN applied to audio analysis of recorded sound. A CNN may be shift or space invariant and may utilize a shared weight architecture and translation. Additionally, each of the respective modules may include one or more generative artificial intelligence algorithms. Generative artificial intelligence algorithms may include generative adversarial networks (GANs) having two networks, a generator model and a discriminator model. Generative artificial intelligence algorithms may also be variational autoencoder (VAE)-based or transformer-based models.

[0028] The controller 201 may include input / output hardware 205 coupled to a communication path 203. The input / output hardware 205 may include a monitor, keyboard, mouse, printer, camera, microphone, speaker, and / or other devices for receiving, sending, and / or displaying data. The controller 201 may include network interface hardware 206 for communicatively coupling the controller 201 to external resources (e.g., vehicle 101 or smart device), the Internet of Things (IoT), and / or a server. The network interface hardware 206 may be communicatively coupled to the communication path 203 and may be any device capable of transmitting and / or receiving data via a network. Thus, the network interface hardware 206 may include a communication transceiver for sending and / or receiving any wired or wireless communication. For example, the network interface hardware 206 may include an antenna, modem, LAN port, WiFi card, WiMAX card, mobile communication hardware, near field communication hardware, satellite communication hardware, and / or any wired or wireless hardware for communicating with other networks and / or devices. In one embodiment, the network interface hardware 206 includes hardware configured to operate according to the Bluetooth® wireless communication protocol. For example, the network interface hardware 206 of the trajectory planning system 100 may receive and / or transmit map data 217, a planned trajectory 105, sensed data of the vehicle 101 (such as speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time of day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii) with a server or the vehicle 101.

[0029] Reference Figure 1 and Figure 3 , depicts an example of the trajectory planning system 100 using the trajectory generation module 222 to generate a planned trajectory 105 for the vehicle 101 (such as an autonomous vehicle) to approach an upcoming intersection 135 to pass through the intersection 135 based on one or more optimal turning paths 315 included in the map data 217. Figure 1 and Figure 3 also depicts an example of the trajectory planning system 100 using the optimal turning path module 232 to generate one or more optimal turning paths 315 associated with the intersection 135 in the map data 217.

[0030] In an embodiment, vehicle 101 may detect an upcoming intersection 135 on road 137 and send an instruction to controller 201 to assist in turning within intersection 135. After receiving the instruction to turn, controller 201 may generate a planned trajectory 105 for vehicle 101 (such as an autonomous vehicle) based on one or more optimal turning paths 315 stored in map data 217, and guide vehicle 101 to follow the trajectory to pass through intersection 135. The planned trajectory 105 may include the upcoming turning path and the speed, time, and kinematic characteristics of vehicle 101 associated with the upcoming turning path. Each upcoming turning path may be an entry point 151 at the boundary of intersection 135, an existing point 153 at the boundary of intersection 135, and a geometric path 152 between entry point 151 and existing point 153. In some embodiments, the upcoming turning path may be one of the one or more optimal turning paths 315.

[0031] In some embodiments, vehicle 101 may use vision sensor 208 and vehicle sensors 212 to generate and transmit sensed data, such as vehicle kinematic data, vehicle control data, and road condition data, to controller 201. In an embodiment, the vehicle kinematic data may be the vehicle position, vehicle speed, and vehicle acceleration of vehicle 101. The vehicle control data may be the vehicle steering, vehicle throttle, and brake inputs of vehicle 101. The road condition data may include the road surface condition, road geometry, and traffic condition of road 137. By receiving the sensed data and other input data (such as local weather conditions), controller 201 may use trajectory generation module 222 to generate a planned trajectory 105 for vehicle 101.

[0032] In some embodiments, the planned trajectory 105 may also be generated based on the parameters of vehicle 101 relative to the parameters of a human-driven vehicle 301 associated with one or more optimal turning paths 3l5. The parameters of vehicle 101 and the parameters of human-driven vehicle 301 may include vehicle length, minimum turning radius, steering system, acceleration and deceleration performance, or other parameters related to the turning performance at intersection 135. Trajectory planning system 100 may compare the similarity of the parameters between vehicle 101 and human-driven vehicle 301, and may also adjust the planned trajectory 105 inversely proportional to the similarity. In some embodiments, trajectory planning system 100 may determine whether the similarity is equal to or greater than a threshold similarity. Trajectory planning system 100 may select the optimal turning path 3l5 associated with a similarity equal to or greater than the threshold similarity when generating the planned trajectory 105, and ignore the optimal turning path 3l5 associated with a similarity less than the threshold similarity.

[0033] In some embodiments, the trajectory planning system 100 may collect sensed data from the vision sensor 208 and the vehicle sensors 212, and may also operate the vehicle 101 to follow the planned trajectory 105 by controlling or adjusting the steering, throttle, and brake inputs of the vehicle 101. The trajectory planning system 100 may monitor the vehicle trace 154 of the vehicle 101 as the vehicle 101 passes through an intersection, and determine whether the vehicle trace 154 deviates from the planned trajectory 105. In response to determining that the vehicle trace 154 deviates from the planned trajectory 105 (e.g., as illustrated in Figure 1 wherein the vehicle trace 154 is different from the geometric path 152), the trajectory planning system 100 may generate an updated trajectory for the vehicle 101 to pass through the intersection 135 based on the historical turning paths 303 and 305 of the human-driven vehicle 301.

[0034] As Figure 3 illustrated, the trajectory planning system 100 may use the optimal turning path module 232 to generate one or more optimal turning paths 315 associated with the intersection 135 based on the turning paths 303 and 305 of one or more human-driven vehicles 301. In some embodiments, the optimal turning path 315 may be the average turning path of the input turning paths 303 and 305 of the human-driven vehicle 301. In some embodiments, the optimal turning path 315 may be the shortest turning path of the input turning paths 303 and 305 of the human-driven vehicle 301. In some embodiments, the optimal turning path 315 may be the most energy-efficient turning path of the input turning paths 303 and 305 of the human-driven vehicle 301. The one or more optimal turning paths 315 may also be generated based on the historical sensor data of the human-driven vehicle 301 over time. The sensor data may include speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time of day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or a combination thereof.

[0035] Referring to Figure 2 and Figure 3 , in an embodiment, the trajectory generation module 222 and the optimal turning path module 232 may include one or more neural networks 322 and 432. Each of the neural networks 322 and 432 may include an encoder, one or more layers of hidden layers, and a decoder. The neural networks 322 and 432 may feed the training data 227 into the encoder during a pre-training process to generate a lower-dimensional representation of the target input-output pairs. For example, the lower-dimensional representation may include the input of the sensed data collected using the vision sensor 208 and the vehicle sensors 212 paired with the planned trajectory and the trajectory detected by the vehicle 101. The first neural network 322 may output the planned trajectory 105. The second neural network 432 may output the optimal turning path 315 to be integrated into the map data 217.

[0036] In some embodiments, the trajectory generation module 222 and the optimal turning path module 232 may include one or more neural networks 322 and 432 that have been trained with training data 227 and historical data 237. The neural networks 322 and 432 may include an encoder or / and a decoder that incorporates layer normalization operations or / and activation function operations. The encoded input data may be normalized and weighted by an activation function before being fed to the hidden layer. The hidden layer may generate a representation of the input data at the bottleneck layer. After passing the data processed by the neural network to the final layer of the neural network, global layer normalization may be performed to normalize the planned trajectory and the optimal turning path. For training and validation purposes, the output may be normalized and may be transformed using an activation function, as described in further detail below. The activation function may be linear or non-linear. The activation function may be, but is not limited to, the Sigmoid function, the Softmax function, the hyperbolic tangent function (Tanh), or the rectified linear unit (ReLu). The neural networks 322 and 432 may be fed the encoder with historical data 237 (such as historical vehicle kinematic data, historical vehicle control data, historical road condition data, historical planned trajectories, historical detected trajectories, historical turning paths of human-driven vehicles, and other historical vehicle operation data) for continuous training.

[0037] In an embodiment, one or more vehicle modules may be pre-trained using training data 227 that includes standard ground truth examples and scenarios in which multiple entities (such as vehicle 101 and human-driven vehicle 301) are traveling on a road 137 that includes multiple lanes 121, 122, 131, and 132, an intersection 135, and a curb 125. The pre-training may include labeling the entities and the desired planned trajectories and optimal turning paths based on the entity and intersection 135 labels in the examples and scenarios, and learning to predict desired and undesired trajectories and optimal turning paths using one or more neural networks 322 and 432 based on the training data 227. The pre-training may also include fine-tuning, evaluation, and testing steps. The module may be continuously trained using real-world collected data as historical data 237 to adapt to changing conditions and factors over time and improve performance. The neural network may be trained based on the activation functions mentioned above. The encoder may generate encoded input data h=(Wx + b) transformed from the input data of one or more input channels. The encoded input data of one of the input channels may be represented from the original input data as which is then used to reconstruct the output . The neural network may reconstruct the output (such as the planned trajectory 105 and the optimal turning path 315) as x’=(W Th + b’), where W is the weight, b is the bias, and W T and b’ are the transposed values of W and b and are learned through backpropagation. In this operation, the neural network can calculate the distance between the input data x and the reconstructed input data x’ for each input data to generate a distance vector |x - x’|. The neural networks 322 and 432 can minimize a loss function, which is a utility function that is the sum of all the distance vectors. The training process can enable the neural networks 322 and 432 to learn the linear or non-linear representation of the input data. The accuracy of the predicted output can be evaluated by meeting a preset value, such as a preset accuracy and the area under the curve (AUC) value calculated using the output scores from an activation function (e.g., Softmax function or Sigmoid function). For example, the trajectory planning system 100 can assign a preset value of AUC with values from 0.7 to 0.8 as acceptable simulations, 0.8 to 0.9 as excellent simulations, or greater than 0.9 as remarkable simulations. After the training meets the preset value, the updated neural networks 322 and 432 can be stored in the trajectory generation module 222 and the optimal steering path module 232, respectively, and the neural networks 322 and 432 are used to generate future planned trajectories and optimal turning paths.

[0038] Reference Figure 4 , depicts a flowchart of a method 400 for trajectory planning. At block 401, the method 400 includes receiving an instruction to turn within the intersection 135. At block 403, the method 400 includes planning a trajectory 105 for the vehicle 101 to turn within the intersection 135 based on one or more optimal turning paths 315 associated with the intersection 135 in the map data 217. The one or more optimal turning paths 315 are generated based on the turning paths 303 and 305 of one or more human-driven vehicles 301.

[0039] In some embodiments, the map data 217 can include crosswalks, traffic lights, traffic signs, obstacles, road lanes, road edges, shoulders, medians, road painted markings, poles, or combinations thereof. The map data 217 can be HD map data or SD map data.

[0040] In some embodiments, the one or more optimal turning paths 315 can also be generated based on the historical sensors of the human-driven vehicle 301 over time. The sensor data can include speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time of day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or combinations thereof. The planned trajectory 105 can include the upcoming turning path and the speed, time, and kinematic characteristics of the vehicle 101 associated with the upcoming turning path.

[0041] In some embodiments, the planned trajectory 105 may be generated using a trained first machine learning algorithm (such as the first neural network 322 in Figure 2 and Figure 3 ) based on one or more optimal turning paths 315 and the parameters of vehicle 101 relative to the parameters of a human-driven vehicle 301 associated with the one or more optimal turning paths 315. The parameters of vehicle 101 and the parameters of human-driven vehicle 301 may include vehicle length, minimum turning radius, steering system, acceleration and deceleration performance, or a combination thereof. Figure 2 and Figure 3 In some embodiments, one or more optimal turning paths 315 may be generated by a second trained machine learning algorithm (such as the second neural network 432 in Figure 2 and Figure 3 ) configured to reduce the path length of the optimal turning paths.

[0042] At block 405, the method 400 includes guiding vehicle 101 to follow the planned trajectory 105 to pass through intersection 135. The method 400 may also include operating vehicle 101 to follow the planned trajectory 105 by controlling or adjusting the steering, throttle, brake inputs of vehicle 101, or a combination thereof. The method 400 may also include monitoring the vehicle trace 154 of vehicle 101 when passing through intersection 135, determining whether the vehicle trace 154 deviates from the planned trajectory 105, and in response to determining that the vehicle trace 154 deviates from the planned trajectory 105, generating an updated trajectory for vehicle 101 to pass through intersection 135 based on the historical turning paths 303 and 305 of human-driven vehicle 301. Figure 3

[0043] Note that the terms "substantially" and "about" may be utilized herein to represent the inherent degree of uncertainty that may be attributed to any quantity of comparison, numerical, measurement, or other representation. These terms are also utilized herein to represent the degree to which a quantity representation may differ from a stated reference without causing a fundamental change in the basic function of the subject matter under discussion.

[0044] Although specific embodiments have been illustrated and described herein, it should be understood that various other changes and modifications may be made without departing from the spirit and scope of the subject matter claimed in the claims. Additionally, although various aspects of the subject matter claimed in the claims have been described herein, such aspects need not be utilized in combination. Accordingly, the appended claims cover all such changes and modifications within the scope of the subject matter claimed in the claims.

[0045] ​​​​

Claims

1. An apparatus for trajectory planning, comprising one or more processors operable to: Receive an instruction to turn within an intersection; Plan a trajectory for an autonomous vehicle to turn within the intersection based on map data including one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on the turning paths of one or more human-driven vehicles; and Guide the autonomous vehicle to pass through the intersection following the trajectory.

2. The apparatus according to claim 1, wherein the one or more optimal turning paths are further generated based on historical sensor data of human-driven vehicles over time.

3. The apparatus according to claim 2, wherein the trajectory is generated using a trained first machine learning algorithm based on the one or more optimal turning paths and parameters of the autonomous vehicle relative to parameters of the human-driven vehicles associated with the one or more optimal turning paths.

4. The apparatus according to claim 3, wherein the parameters of the autonomous vehicle and the parameters of the human-driven vehicles include vehicle length, minimum turning radius, steering system, acceleration and deceleration performance, or a combination thereof.

5. The apparatus according to claim 2, wherein the sensor data includes speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time of day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or a combination thereof.

6. The apparatus according to claim 1, wherein the one or more optimal turning paths are generated by a second trained machine learning algorithm configured to reduce the path length of the optimal turning paths.

7. The apparatus according to claim 1, wherein the trajectory includes an upcoming turning path and the speed, time, and kinematic characteristics of the autonomous vehicle associated with the upcoming turning path.

8. The apparatus according to claim 7, wherein the upcoming turning path is one of the one or more optimal turning paths.

9. The apparatus according to claim 1, wherein the one or more processors are further operable to operate the autonomous vehicle to follow the trajectory by controlling or adjusting the steering, throttle, brake inputs of the autonomous vehicle, or a combination thereof.

10. The apparatus according to claim 1, wherein the one or more processors are further operable to: Monitor the vehicle track of the autonomous vehicle while passing through the intersection; Determine whether the vehicle track deviates from the trajectory; and In response to determining that the vehicle track deviates from the trajectory, generate an updated trajectory for the autonomous vehicle to pass through the intersection based on the historical turning paths of the human-driven vehicles.

11. The apparatus according to claim 1, wherein the map data includes crosswalks, traffic lights, traffic signs, obstacles, road lanes, road edges, shoulders, medians, road markings, poles, or a combination thereof.

12. The apparatus according to claim 1, wherein the map data is high-definition map data or standard-definition map data.

13. A method for trajectory planning, comprising: Receiving an instruction to turn within an intersection; Planning a trajectory for an autonomous vehicle to turn within the intersection based on map data including one or more optimal turning paths associated with the intersection, wherein the one or more optimal turning paths are generated based on the turning paths of one or more human-driven vehicles; And Guiding the autonomous vehicle to pass through the intersection following the trajectory.

14. The method according to claim 13, wherein: The one or more optimal turning paths are further generated based on historical sensor data of human-driven vehicles over time; The sensor data includes speed data, acceleration data, steering data, yaw data, wheel slip data, lane departure data, time of day, weather conditions, vehicle type, vehicle size, minimum and maximum vehicle turning radii, or a combination thereof.

15. The method according to claim 13, wherein: The trajectory is generated using a trained first machine learning algorithm based on the one or more optimal turning paths and parameters of the autonomous vehicle relative to parameters of the human-driven vehicle associated with the one or more optimal turning paths; The parameters of the autonomous vehicle and the parameters of the human-driven vehicle include vehicle length, minimum turning radius, steering system, acceleration and deceleration performance, or a combination thereof.

16. The method according to claim 13, wherein the one or more optimal turning paths are generated by a second trained machine learning algorithm configured to reduce the path length of the optimal turning paths.

17. The method according to claim 13, wherein the trajectory includes an upcoming turning path, and the speed, time, and kinematic characteristics of the autonomous vehicle associated with the upcoming turning path.

18. The method according to claim 13, wherein the method further comprises operating the autonomous vehicle to follow the trajectory by controlling or adjusting the steering, throttle, brake inputs, or a combination thereof of the autonomous vehicle.

19. The method according to claim 13, wherein the method further comprises: Monitoring the track of the autonomous vehicle while passing through the intersection; Determining whether the track deviates from the trajectory; And In response to determining that the track deviates from the trajectory, generating an updated trajectory for the autonomous vehicle to pass through the intersection based on the historical turning paths of human-driven vehicles.

20. The method according to claim 13, wherein the map data is high-definition map data or standard-definition map data.

Images