A vehicle controller includes a processor configured to detect an object region including another vehicle near a vehicle from each of time series images obtained by a camera mounted on the vehicle; detect a predetermined action taken by the other vehicle, based on a trajectory of the other vehicle estimated from the object region of each image; identify the state of a signal light of the other vehicle, based on characteristics obtained from pixel values of the object region of each image; extract information indicating characteristics of an action of the other vehicle or the state of a signal light at the predetermined action taken by the other vehicle, based on the predetermined action detected in a tracking period and the state of the signal light related to the predetermined action; and predict behavior of the other vehicle, using the extracted information.

In various examples, a multi-sensor fusion machine learning model—such as a deep neural network (DNN)—may be deployed to fuse data from a plurality of individual machine learning models. As such, the multi-sensor fusion network may use outputs from a plurality of machine learning models as input to generate a fused output that represents data from fields of view or sensory fields of each of the sensors supplying the machine learning models, while accounting for learned associations between boundary or overlap regions of the various fields of view of the source sensors. In this way, the fused output may be less likely to include duplicate, inaccurate, or noisy data with respect to objects or features in the environment, as the fusion network may be trained to account for multiple instances of a same object appearing in different input representations.

The present technology pertains to predicting trajectories of objects near an autonomous vehicle. The predictions may be obtained as output from a trajectory prediction machine learning model. The inputs to the trajectory prediction machine learning model may be based on a first map of an area surrounding an autonomous vehicle, and a second map of an area around an object within the first area. The second map may have a smaller area and a higher resolution relative to the first map.

An autonomous driving control apparatus for determining a driving path of an autonomous vehicle and a method thereof are provided. A path calculation device calculates paths swept by a part or all of a body of an autonomous vehicle with respect to two or more driving path candidates of the autonomous vehicle. A controller determines risks for the two or more driving path candidates based on the swept paths, and determines a driving path of the autonomous vehicle based on the risks determined for the two or more driving path candidates. The autonomous driving control apparatus prevents a risk of collision between a trailer part and an object to ensure stability of autonomous driving.

Disclosed are systems and techniques for managing an autonomous vehicle (AV). In some aspects, an AV may predict a first predicted position of an object perceived by one or more sensors of the autonomous vehicle, wherein the first predicted position of the object is associated with an uncertainty metric. The AV may determine that a first error between a first actual position of the object and the first predicted position of the object is greater than the uncertainty metric. The AV may increase the uncertainty metric corresponding to a second predicted position of the object based on the first error to result in a revised uncertainty metric. The AV may provide the revised uncertainty metric to a planning stack for maneuvering the AV.

System, methods, and computer-readable media for training an object path prediction model to reduce an uncertainty of a predicted path when the predicted path of an object adjacent to another object. The training penalizes an uncertainty area prediction associated with a predicted future location of a nearby object to an autonomous vehicle (AV) when the uncertainty area prediction overlaps with another object to which the first detected object would be adjacent at the predicted future location. The training also penalizes a set of predicted future locations that implies improbable vehicle kinematics, whereby the object path prediction model becomes trained to avoid predicting similar sets of predicted future locations with improbable vehicle kinematics.

A computer-implemented method and system for classifying traffic situations of a virtual test. The method comprises concatenating a plurality of determined data segments of the lateral and longitudinal behavior of the ego vehicle to identify vehicle actions and classifying traffic situations by linking a subset of the determined data segments of the lateral and longitudinal behavior of the ego vehicle with the identified vehicle actions. The invention further comprises a computer-implemented method for providing a trained machine learning algorithm for classifying traffic situations of a virtual test.

Systems and methods for communicating uncertainty around stationary objects are provided. For example, a method for communicating uncertainty around stationary objects includes receiving sensor data corresponding to a stationary object at a location along a road segment. The sensor data is captured via one or more sensors of a first vehicle. The method also includes based on the sensor data, determining a level of uncertainty corresponding to the stationary object. The method also includes based on the determined level of uncertainty, providing an instruction for one or more sensors of a second vehicle to capture additional sensor data corresponding to the stationary object at the location along the road segment.

During the autonomous driving, the movement trails or moving history of obstacles, as well as, an autonomous driving vehicle (ADV) may be maintained in a corresponding buffer. For the obstacles and the ADV, the vehicle states at different points in time are maintained and stored in one or more buffers. The vehicle states representing the moving trails or moving history of the obstacles and the ADV may be utilized to reconstruct a history trajectory of the obstacles and the ADV, which may be used for a variety of purposes. For example, the moving trails or history of obstacles may be utilized to determine lane configuration of one or more lanes of a road, particularly, in a rural area where the lane markings are unclear. The moving history of the obstacles may also be utilized predict the future movement of the obstacles, tailgate an obstacle, and infer a lane line.

Disclosed herein are methods and systems for determining an attribute of an object at a pre-determined time point. Data representing a respective property of the object and a plurality of further objects at a plurality of time points different from the pre-determined time point are determined, and the data is arranged in an image-like data structure. The image-like data structure has a plurality of columns and a plurality of rows. The data is arranged in the image-like data structure such that each of one of the rows or the columns corresponds to respective properties of the object or of one of the plurality of further objects and each of the other of the rows or the columns corresponds to respective properties at one of the plurality of time points. The attribute is then determined using a pre-determined rule based on the image-like data structure.