EP4320600A1 - Computerimplementiertes verfahren und system zum trainieren eines maschinellen lernverfahrens - Google Patents
Computerimplementiertes verfahren und system zum trainieren eines maschinellen lernverfahrensInfo
- Publication number
- EP4320600A1 EP4320600A1 EP22720432.8A EP22720432A EP4320600A1 EP 4320600 A1 EP4320600 A1 EP 4320600A1 EP 22720432 A EP22720432 A EP 22720432A EP 4320600 A1 EP4320600 A1 EP 4320600A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- frames
- future
- machine learning
- objects
- ego vehicle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/248—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving reference images or patches
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/242—Aligning, centring, orientation detection or correction of the image by image rotation, e.g. by 90 degrees
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/776—Validation; Performance evaluation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30241—Trajectory
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30261—Obstacle
Definitions
- the invention relates to a computer-implemented method for training a machine learning method for recognizing future trajectories of objects in relation to an ego vehicle.
- the invention also relates to a system.
- An autonomous or fully autonomous vehicle is a vehicle capable of sensing its surroundings and navigating with little or no user input. This is done through the use of sensor devices such as radar, lidar systems, cameras, ultrasound and the like.
- the vehicle analyzes the sensor data with regard to the course of the road, other road users and their trajectory. Furthermore, the vehicle must react appropriately to the recorded data and calculate control commands based on the recorded data and forward these to actuators in the vehicle.
- an autonomous vehicle In order for an autonomous vehicle to reach its destination, it not only has to perceive and interpret its surroundings, but also predict what could happen in the future. These predictions are on the order of one to three seconds, for example when a road user turns off or a pedestrian crosses the street, so that the autonomous vehicle can plan/replan its future route safely and without colliding.
- DE 10 2018222 542 A1 discloses a method for predicting the trajectory of at least one controlled object, using a physical measurement determined current position of the object is provided, at least one anticipated destination of the movement of the object is provided; using physical observations of the object, and/or the environment in which the object is moving. At least one anticipated preference is determined which occurs when the object is steered towards the at least one anticipated destination.
- the object is achieved by a computer-implemented method for training a machine learning method for recognizing future trajectories of objects in relation to an ego vehicle with the features of claim 1.
- the object is also achieved by a system with the features of claim 12.
- the object is achieved by a computer-implemented method for training a machine learning method for recognizing future trajectories of objects in relation to an ego vehicle, comprising the steps:
- Frames are, so to speak, traffic scenarios (a single traffic scenario) at a certain point in time.
- the frames can be viewed as individual images of the traffic scenarios that follow one another over time. Traffic scenarios can therefore be constructed from frames of traffic scenarios that follow one another in time.
- the ego pose is essentially at least the orientation of the ego vehicle.
- Ground truth traffic scenarios are the traffic scenarios that actually occur; i.e. the traffic scenarios that really occur after the first point in time up to the second point in time with the trajectories actually driven by the road users after the first point in time.
- a traffic scenario can consist of a number/amount of different moving objects (bicycles/cars/pedestrians) and/or stationary objects (traffic lights/road signs) in the vicinity of the ego vehicle.
- Fixed objects such as traffic signs, road markings, traffic lights, pedestrian crossings and obstacles are in a precisely defined position.
- Moving objects such as bicycles, cars, etc. exhibit dynamic behavior (trajectory) such as speed, acceleration/deceleration, distance from the center line of the road, etc.
- the term “ego vehicle” can be understood as the vehicle whose surroundings are to be monitored.
- the ego vehicle can in particular be a fully or partially autonomously driving motor vehicle for traffic on roads, which is intended to steer at least partially independently.
- sensors etc. are usually arranged on the ego vehicle, which can detect the surroundings using sensors.
- a trajectory designates a set of positions and orientations linked in time and space, i.e. a route taken by a road user along or in the frame.
- all frames are oriented based on the ego pose, so that the ego movement and ego rotation is only depicted because the ego vehicle itself is not specified.
- the last two seconds of the traffic scenarios are selected as historical frames and represent the input training data with the ground truth frames.
- the machine learning method trained using the method according to the invention makes it possible to create a prediction of the object trajectories on the basis of the complete frame as input.
- a machine learning method for example an artificial neural network, is also trained in a simplified manner by the method according to the invention.
- the machine learning method is trained using the complete knowledge about, for example, the lanes and traffic rules (static objects) as training data.
- the learning process learned in this way can now include this knowledge in the prediction.
- the entire previous knowledge about road users is used in the machine learning method trained according to the invention.
- the learned machine learning process can also include this in the later predictions.
- the past movements of road users and the category to which this road user belongs such as e.g. B. pedestrians, cars, trucks, bicycles, etc., by input of the complete frames into the learning process trained with it. Based on this, it is possible for the machine learning method trained with it to later take all road users into account without affecting the computing time.
- the social interactions can be taken into account. Based on this, it is possible for the machine learning method trained with it to later take these social interactions into account when predicting the future movement of road users.
- a machine learning method for generating anticipatory traffic scenarios can be trained using historical frames and ground truth frames.
- an improved machine learning method can be generated, which provides an improved prediction of the trajectories of moving objects in its environment.
- a machine learning method is trained using the complete frames and thus the entire map information and the entire social interactions as well as the history of the historical trajectory as input and can accordingly achieve better results after training.
- the machine learning method trained using the method is thus able to determine all trajectories in the traffic scenarios around the ego vehicle at once in advance.
- only a constant time is required for the prediction, which is independent of the number of road users, for example by including the social interaction of the respective road users in the prediction and, for example, the prior historical knowledge about the road users in the future traffic scenarios to be determined be included.
- the objects are in the form of both static objects and moving objects, with the static objects and the moving objects be identified as markers at least by their size and shape. Each object is preferably represented by its original size and length and width. Furthermore, static objects and moving objects can be identified as markers using different colors. For this purpose, an RGB color palette is used that displays all available map information such as lane centers and lane boundaries. For example, road users can be displayed in gray.
- the historical frames as well as the ground truth frames and the future frames generated by the machine learning method have a time stamp.
- each moving gray object represents a point in time when the frame was created. This allows the time steps in connection with the objects to be displayed together in one frame. The decoding of the objects in relation to their history and the associated time step is therefore given in the data structure itself.
- the frames are designed as an image section from a respective traffic scenario, with each image section being formed by a specified radius around the coordinates of the ego vehicle, so that the ego vehicle is centered in the middle of the image section.
- This enables better tracking of the moving objects from the perspective of the ego vehicle and faster processing. Since all frames contain the tracking of all objects and their poses, only those objects that can be perceived from the perspective of the ego vehicle are required to determine relevant trajectories.
- the individually generated frames reduced by the image detail thus only contain objects that are visible to this specific ego vehicle in its field of vision.
- the radius can be chosen freely. In particular, one can be selected as 50 m.
- All frames are based on the Ego coordinates, ie the coordinates of the ego vehicle and the direction centered and oriented so that the ego movement and ego rotation is only represented by this and the ego vehicle itself is not indicated.
- the ego vehicle is always in the middle of a frame as the coordinate origin.
- the historical trajectory of moving objects i.e. road users
- the expected future trajectories generated by the machine learning method are determined on the basis of the future frames.
- the future trajectories are extracted from the future frames generated by the machine learning process and assigned to the associated object (road users).
- the future frames are preferably rotated according to the ego pose in order to obtain the same orientation as the ego vehicle or the historical frames; i.e. the historical and future frames are aligned to each other.
- Ego pose means the position and orientation of the ego vehicle.
- the contours and thus the objects (road users) and their trajectories are preferably further recognized and their pose, i.e. orientation and coordinates, determined and compared with the pose of the individual road users in the historical frames. This allows an assignment to be made. If the assignment has been obtained as a result, the future trajectories can be assigned to the known road users.
- the future trajectories of moving objects are determined on the basis of the ground truth frames.
- the machine learning method can then be trained using the historical trajectories from the historical frames and the future trajectories determined by the machine learning method.
- a machine learning method can be taught in a targeted manner, for example by means of iterative gradient methods.
- a quality of the machine learning method is determined by determining the difference between the ground truth trajectories and the expected future trajectories generated by the machine learning method as the mean absolute error MAE:
- the traffic scenarios can be simulated in virtual space in a bird's-eye view. This makes it easy to generate the historical frames as well as the ground truth frames.
- the machine learning method is a deep learning method, which is trained using a gradient method.
- This learning process can be designed as a deep neural network, for example. Based on the trajectories or the frames, the network can be learned iteratively using gradient descent.
- a decoder-encoder structure can be used as the architecture of the artificial neural network.
- the artificial neural network can be a convolutional network, in particular a deep convolutional neural network.
- the encoder is responsible for compressing the input signal using convolution and transforming the input into a low-dimensional vector.
- the decoder is responsible for the restoration. The decoder then transforms the low-dimensional vector into the desired output.
- the object is also achieved by a system for training a machine learning method for recognizing future trajectories of objects in relation to an ego vehicle, comprising:
- - a storage unit for providing chronologically consecutive global traffic scenarios as chronologically consecutive frames in a global coordinate system, the global traffic scenarios having objects and all objects in the global traffic scenarios being marked with different markers,
- a processor for determining the ego pose of the ego vehicle in the chronologically consecutive frames and for transforming the frames with the marked objects based on the determined ego pose in a local coordinate system as a local traffic scenario, so that the respective frame has the same orientation how the ego vehicle has, the transformed frames being used as historical frames up to a first point in time and the transformed frames being used as ground truth frames from the first point in time to a second point in time,
- the processor for training the machine learning method using the historical frames, for determining future local traffic scenarios up to a second point in time as future frames and comparing the future frames generated by the machine learning method with the corresponding ground truth frames.
- FIG 1 different historical frames
- FIG 2 immovable objects in a frame
- FIG 3 the ground truth frames
- FIG 4 the historical frames and ground truth frames as a table
- FIG 5 stacked frames as a single frame
- FIG 6 encoder and decoder of the neural network
- FIG 7 a calculated future trajectory.
- sensors with which the autonomous vehicle is equipped are used, which record the environment. The recorded sensor data must be processed and interpreted.
- a machine learning method can, for example, be a neural one network are used. However, this must be reliably trained in order to correctly interpret the sensor data received.
- the computer-implemented method for training the machine learning method for recognizing future trajectories of objects in relation to an ego vehicle can be used.
- the current and previous positions of a road user in Cartesian coordinates can be used.
- the trajectories or trajectory data are therefore inherent time series data.
- the traffic scenarios are preferably represented by objects.
- the objects can essentially be divided into static and moving objects (road users).
- Static objects are, for example, roadways and roadway borders, traffic lights, traffic signs, etc.
- Movable objects here are primarily road users such as cars, pedestrians and cyclists. These generate a so-called trajectory. Trajectory denotes a set of temporally and spatially linked positions and orientations, i.e. the travel route of the moving object.
- traffic scenarios are preferably created/simulated using a data set based on simulation data. Furthermore, the traffic scenarios are preferably simulated in relation to different cities in order to ensure that the simulation data is of sufficient quality. In this way, large amounts of different traffic scenarios can be generated, which can be used to train the machine learning process.
- Each traffic scenario is represented as a frame.
- the historical frames indicate the history, i.e. in the case of moving objects, the trajectory that has already been driven.
- Each object is preferably represented by its original size and length and width.
- static objects and moving objects can be identified by different colors as markers (RGB color palette) in the simulation.
- RGB color palette is used to display all available map information such as lane centers and lane boundaries.
- Road users or their historical trajectories 3 can thus be displayed in gray in each of the simulated historical frames 1a, . . . , 1e and ground truth frames 2a, . . . , 2e.
- 1 shows various historical frames 1a, .., 1e, which contain the trajectories 3 of all objects.
- the frames and thus the objects are rotated around the ego pose so that they correspond to the view from the ego vehicle.
- the trajectory 3, here of an individual object, is recognized as it were by the fact that the historical frames 1a, 1e can be displayed/perceived as an image sequence.
- the frames 1a, . . . are displayed which can be perceived from the perspective of the ego vehicle, i.e. which would be perceived from the "ego perspective".
- the ego vehicle or its coordinates are thus centered in the middle of the image section (origin of coordinates). This enables better tracking of the objects from the perspective of the ego vehicle and faster processing. Since all frames include the tracking of all objects and their poses, the ego vehicle itself does not need to be shown in the frames.
- the individually generated frames with a reduced image section then only contain objects that are visible to this specific ego vehicle in its field of vision.
- the radius can be chosen freely. In particular, one can be selected as 50 m. This ensures that all moving and immobile objects are detected, which are necessary to autonomously control the ego vehicle for the next few seconds / minutes.
- the objects are centered in the direction of the ego vehicle, so that the ego vehicle with the ego coordinates is in the center, i.e. the origin of the coordinates is here, so that the ego movement and ego rotation is only mapped through and that Ego vehicle itself is not specified.
- the ego vehicle is always in the middle of the respective frame 1a, 1e, and is not shown.
- immovable objects can be displayed, which are also displayed rotated around the pose of the ego vehicle.
- the various lanes 5 are shown in green (dashed here) as an example of immovable objects.
- FIG 4 shows the representation of the historical frames 1a, .1, e and ground truth frames 2a, 2e as a table.
- FIG. 5 shows such an image, in which individual frames have been placed one on top of the other as an image sequence, so to speak, for recognizing different objects and object trajectories, shown here as an example of an object trajectory 6 .
- a machine learning method is then trained, preferably using the historical frames 1a, . . . , 1e and the ground truth frames 2a, . . . , 2e.
- Such a learning method is preferably designed as an artificial deep neural network, which is described in more detail in FIG.
- This is preferably designed as an encoder and decoder, which are trained iteratively using a gradient method.
- the artificial neural network can use the trajectories 3, 4 from the historical frames 1a, . . . 1e and the ground truth frames 2a, . . . 2e and/or the frames 1a, . . . 1e, 2a, ... 2e itself can be learned iteratively by means of the gradient descent.
- the neural network can be a convolutional network, in particular a deep convolutional neural network.
- the encoder is responsible for compressing the input signal using convolution.
- the decoder is responsible for recovering inputs.
- the encoder transforms the input into a low-dimensional vector.
- the decoder then transforms the low-dimensional vector into the desired output.
- GNA network Geneative Adversarial Networks
- the neural network uses the historical frames 1a, . . . 1e to calculate the future frames.
- the trajectories can then be extracted from the future frames calculated by the neural network and assigned to the associated object (road users).
- the future frames are preferably first rotated according to the ego pose in order to obtain the same orientation as the ego vehicle or the historical frames; i.e. the historical frames 1a, .. , 1e and future frames are aligned in the same way. Then, in the rotated future frames, the contours and thus the objects (road users) are preferably still recognized and their pose, i.e. orientation and coordinates, are determined and compared with the pose of the individual known objects at time tO. If an assignment has been obtained as a result, the future trajectories can be assigned to the known road users or objects.
- FIG. 7 shows a calculated future trajectory, with the last six steps in FIG. 7 being summarized as “prediction (right) trajectory” and a ground truth trajectory (left).
- the machine learning method can be evaluated using the soft-dice-loss method (similarity measure function). This indicates the degree of overlap between the future bird's-eye view frames and the ground truth bird's-eye view frames in relation to the original object size.
- a quality of the machine learning method can be determined by determining the difference between the ground truth trajectories 4 and the expected future trajectories generated by the machine learning method as the mean absolute error MAE: (ground truth trajectories) ; — (future trajectories) ;
- the neural network can be trained using the method according to the invention in such a way that it uses map information and driving context when predicting the future trajectories of road users, as well as prior knowledge about the road users when predicting the future trajectories of road users, as well as social interactions when prediction of future trajectories between road users.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102021203492.6A DE102021203492B3 (de) | 2021-04-08 | 2021-04-08 | Computerimplementiertes Verfahren und System zum Trainieren eines maschinellen Lernverfahrens |
| PCT/EP2022/058835 WO2022214416A1 (de) | 2021-04-08 | 2022-04-04 | Computerimplementiertes verfahren und system zum trainieren eines maschinellen lernverfahrens |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4320600A1 true EP4320600A1 (de) | 2024-02-14 |
Family
ID=81256440
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP22720432.8A Withdrawn EP4320600A1 (de) | 2021-04-08 | 2022-04-04 | Computerimplementiertes verfahren und system zum trainieren eines maschinellen lernverfahrens |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US20240185437A1 (de) |
| EP (1) | EP4320600A1 (de) |
| CN (1) | CN117121060A (de) |
| DE (1) | DE102021203492B3 (de) |
| WO (1) | WO2022214416A1 (de) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2024019849A (ja) * | 2022-08-01 | 2024-02-14 | 株式会社Subaru | 車両の走行制御装置 |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102017204404B3 (de) | 2017-03-16 | 2018-06-28 | Audi Ag | Verfahren und Vorhersagevorrichtung zum Vorhersagen eines Verhaltens eines Objekts in einer Umgebung eines Kraftfahrzeugs und Kraftfahrzeug |
| US10788830B2 (en) * | 2017-07-28 | 2020-09-29 | Qualcomm Incorporated | Systems and methods for determining a vehicle position |
| US10782693B2 (en) * | 2017-09-07 | 2020-09-22 | Tusimple, Inc. | Prediction-based system and method for trajectory planning of autonomous vehicles |
| US11282389B2 (en) * | 2018-02-20 | 2022-03-22 | Nortek Security & Control Llc | Pedestrian detection for vehicle driving assistance |
| US11966838B2 (en) * | 2018-06-19 | 2024-04-23 | Nvidia Corporation | Behavior-guided path planning in autonomous machine applications |
| DE102018222542A1 (de) | 2018-12-20 | 2020-06-25 | Robert Bosch Gmbh | Bewegungsvorhersage für gesteuerte Objekte |
| DE102020100685A1 (de) * | 2019-03-15 | 2020-09-17 | Nvidia Corporation | Vorhersage zeitlicher informationen in autonomenmaschinenanwendungen |
| US11579629B2 (en) * | 2019-03-15 | 2023-02-14 | Nvidia Corporation | Temporal information prediction in autonomous machine applications |
| US11131993B2 (en) * | 2019-05-29 | 2021-09-28 | Argo AI, LLC | Methods and systems for trajectory forecasting with recurrent neural networks using inertial behavioral rollout |
| US11520037B2 (en) * | 2019-09-30 | 2022-12-06 | Zoox, Inc. | Perception system |
| CN111002980B (zh) * | 2019-12-10 | 2021-04-30 | 苏州智加科技有限公司 | 基于深度学习的道路障碍物轨迹预测方法和系统 |
| CN111626097A (zh) * | 2020-04-09 | 2020-09-04 | 吉利汽车研究院(宁波)有限公司 | 一种障碍物未来轨迹的预测方法、装置、电子设备及存储介质 |
| US11919545B2 (en) * | 2020-05-15 | 2024-03-05 | Perceptive Automata, Inc. | Scenario identification for validation and training of machine learning based models for autonomous vehicles |
| US12485889B2 (en) * | 2020-11-19 | 2025-12-02 | Nvidia Corporation | Object detection and collision avoidance using a neural network |
-
2021
- 2021-04-08 DE DE102021203492.6A patent/DE102021203492B3/de not_active Expired - Fee Related
-
2022
- 2022-04-04 CN CN202280026728.4A patent/CN117121060A/zh active Pending
- 2022-04-04 EP EP22720432.8A patent/EP4320600A1/de not_active Withdrawn
- 2022-04-04 US US18/554,288 patent/US20240185437A1/en not_active Abandoned
- 2022-04-04 WO PCT/EP2022/058835 patent/WO2022214416A1/de not_active Ceased
Also Published As
| Publication number | Publication date |
|---|---|
| CN117121060A (zh) | 2023-11-24 |
| WO2022214416A1 (de) | 2022-10-13 |
| DE102021203492B3 (de) | 2022-05-12 |
| US20240185437A1 (en) | 2024-06-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3380811B1 (de) | Verfahren und system zum erstellen einer digitalen karte | |
| DE102015203016B4 (de) | Verfahren und Vorrichtung zur optischen Selbstlokalisation eines Kraftfahrzeugs in einem Umfeld | |
| DE102020127043A1 (de) | Sicherheitssystem für ein fahrzeug | |
| DE102018203583B4 (de) | Verfahren, Fahrerassistenzsystem sowie Kraftfahrzeug zur Prädiktion einer Position oder einer Trajektorie mittels eines graphbasierten Umgebungsmodells | |
| DE102019101938A1 (de) | Erstellung kognitiver Karten für Fahrzeuge | |
| DE102020113848A1 (de) | Ekzentrizitätsbildfusion | |
| DE102019122826A1 (de) | Adaptives fahrzeuglernen | |
| DE102016211182A1 (de) | Verfahren, Vorrichtung und System zum Durchführen einer automatisierten Fahrt eines Fahrzeugs entlang einer aus einer Karte bereitgestellten Trajektorie | |
| DE102019108644A1 (de) | Verfahren und vorrichtung zum automatischen lernen von regeln für autonomes fahren | |
| DE102017201665A1 (de) | Verfahren zur Einbindung eines dynamischen Objektes in eine digitale Karte eines hochautomatisierten Fahrzeugs (HAF) | |
| DE102021119871B4 (de) | Verfahren und Prozessorschaltung zum Betreiben einer automatisierten Fahrfunktion mit Objektklassifikator in einem Kraftfahrzeug, sowie Kraftfahrzeug | |
| DE102018213378B4 (de) | Fahrassistenzsystem für ein Fahrzeug, Fahrzeug mit demselben und Fahrassistenzverfahren für ein Fahrzeug | |
| DE102021114724A1 (de) | Verbesserter fahrzeugbetrieb | |
| DE102015206546A1 (de) | Fahrbahnmarkierungserkennungsvorrichtung | |
| DE102019122536A1 (de) | Exzentrizitätskarten | |
| DE102020200876B4 (de) | Verfahren zum Verarbeiten von Sensordaten einer Sensorik eines Fahrzeugs | |
| WO2022043203A1 (de) | Training eines generators zur erzeugung realistischer bilder mit einem semantisch segmentierenden diskriminator | |
| DE102023114645A1 (de) | Systeme und Verfahren zur Bildklassifizierung unter Verwendung eines neuronalen Netzwerks in Kombinaiton mit einer Korrelationsstruktur | |
| DE102020122610A1 (de) | Neuronales fahrzeugnetzwerk | |
| WO2022106336A1 (de) | Verfahren zum steuern eines kraftfahrzeugs und steuergerät | |
| DE102024201831A1 (de) | Verfahren zum Erstellen von hochaufgelösten Umgebungskarten für ein Fahrzeug mit einer autonomen Fahrfunktion | |
| EP4320600A1 (de) | Computerimplementiertes verfahren und system zum trainieren eines maschinellen lernverfahrens | |
| DE102017215844A1 (de) | Verfahren zum Betreiben eines Fahrzeugs | |
| DE102022210504B4 (de) | System und Verfahren zur Prädiktion einer zukünftigen Position eines Verkehrsteilnehmers | |
| DE102020208421A1 (de) | Verfahren sowie Steuergerät zum Steuern eines Kraftfahrzeugs |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20231031 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN |
|
| 18D | Application deemed to be withdrawn |
Effective date: 20240516 |