Qualcomm Incorporated

Avdhut Joshi

Jayakrishnan Unnikrishnan

Yoga Y Nadaraajan

Murali Chari

Volodimir Slobodyanyuk

Arvind Ramanandan

Amir Salimi

Gautam Sachdeva

Sundar Subramanian

Shivam Agarwal

Sree Sesha Aravind Vadrevu

Makesh Pravin John Wilson

Xinzhou Wu

Ahmed Kamel Sadek

Urs Niesen

Christopher Brunner

Amin Ansari

Ravi Teja Sukhavasi

Duck Hoon Kim

Radhika Dilip Gowaikar

Young-Ki Baik

Hyun-Mook Cho

Shantanu Chaisson Sanyal

Behnaz Rezaei

Yiming Chen

John Steven Lima

Shubham Sanjeev Garg

In some aspects, a device may receive sensor data associated with a vehicle and a set of frames. The device may aggregate, using a first pose, the sensor data associated with the set of frames to generate an aggregated frame, wherein the aggregated frame is associated with a set of cells. The device may obtain an indication of a respective occupancy label for each cell from the set of cells, wherein the respective occupancy label includes a first occupancy label or a second occupancy label, and wherein a subset of cells from the set of cells are associated with the first occupancy label. The device may train, using data associated with the aggregated frame, a machine learning model to generate an occupancy grid, based on a loss function that only calculates a loss for respective cells from the subset of cells. Numerous other aspects are described.

Machine learning based occupancy grid generation

A processor-implemented method for radar-based tracking of an object includes transmitting radio frequency (RF) signals. In response to the transmitted RF signals, one or more return RF signals are received. Features of the one or more return RF signals are extracted. A graph comprising multiple nodes is generated. Each node of the graph corresponds to the one or more return RF signals and indicates a potential target object detection. An existence of a plurality of edges is determined. Each edge connects a pair of nodes in the graph based on features of the return RF signals. The existence of each edge indicates that the pair of nodes connected correspond to a same target object.

Radar clustering using machine learning

In some aspects, a device may receive, from a radar scanner or a LIDAR scanner of a vehicle, point data that identifies a first point and a second point. The device may receive grid information that identifies cells of a grid that is associated with mapping a physical environment of the vehicle. The device may designate, based on determining that a distance between the first point and the second point satisfies a distance threshold, a subset of the cells as an occupied cluster that is associated with the first point and the second point. The device may perform an action associated with the vehicle based on location information associated with the occupied cluster. Numerous other aspects are described.

Occupancy clustering according to radar data

Systems and techniques are described for performing object detection and tracking. For example, a tracking object can obtain an image comprising a target object at least partially in contact with a surface. The tracking object can obtain a plurality of two-dimensional (2D) keypoints based on one or more features associated with one or more portions of the target object in contact with the surface in the image. The tracking object can obtain information associated with a contour of the surface. Based on the plurality of 2D keypoints and the information associated with the contour of the surface, the tracking object can determine a three-dimensional (3D) representation associated with the plurality of 2D keypoints.

Three-dimensional target estimation using keypoints

In some aspects, a device may receive point data associated with a cell of an occupancy grid for controlling a vehicle. The device may determine, based on the point data, a characteristic of the cell that is associated with an occupancy probability of the cell, wherein the occupancy probability is determined according to a first technique based on the point data. The device may configure, based on the characteristic, the occupancy probability for the cell, within the occupancy grid, according to a second technique. Numerous other aspects are described.

Occupancy mapping for autonomous control of a vehicle

Techniques and systems are provided for determining static occupancy. For example, an apparatus can be configured to determine one or more pixels associated with one or more static objects depicted in one or more images of a three-dimensional space. The apparatus can be configured to obtain a point map including a plurality of map points, the plurality of map points corresponding to a portion of the three-dimensional space. The apparatus can be configured to determine, based on the point map and the one or more pixels associated with the one or more static objects, a probability of occupancy by the one or more static objects in the portion of the three-dimensional space. The apparatus can be configured to combine information across multiple images of the three-dimensional space, and can determine probabilities of occupancy for all cells in a static occupancy grid that is associated with the three-dimensional space.

Static occupancy tracking

Methods, system, non-transitory media, and devices for supporting safety compliant computing in a heterogeneous computing system, such as a vehicle heterogeneous computing system are disclosed. Various aspects include methods enabling a vehicle, such as an autonomous vehicle, a semi-autonomous vehicle, etc., to achieve algorithm safety for various algorithms on a heterogeneous compute platform with various safety levels.

System to achieve algorithm safety in heterogeneous compute platform

Disclosed are techniques for determining a motion state of a target object. In an aspect, an on-board computer of an ego vehicle detects the target object in one or more images, determines one or more first attributes of the target object based on measurements of the one or more images, determines one or more second attributes of the target object based on measurements of a map of a roadway on which the target object is travelling, and determines the motion state of the target object based on the one or more first attributes and the one or more second attributes of the target object.

Determining a motion state of a target object

Techniques and systems are provided for determining one or more sizes of one or more objects. For example, a bounding region identifying a first object detected in an image can be obtained. A map including map points can also be obtained. The map points correspond to one or more reference locations in a three-dimensional space. The bounding region identifying the first object can be associated with at least one map point of the map points included in the map. Using the bounding region and the at least one map point, an estimated three-dimensional position and an estimated size of the first object detected in the image can be determined. In some examples, other information can be used to estimate the estimated three-dimensional position and an estimated size of the first object, such as radar information and/or other information.

Object size estimation using camera map and/or radar information

An electronic device is described. The electronic device includes a memory and a processor in communication with the memory. The memory is configured to store precalibration data for a camera mounted on a vehicle, the precalibration data including a camera height determined relative to a road plane the vehicle is configured to contact during operation. The processor is configured to receive a plurality of images. The processor is also configured to classify one or more features in the plurality of images as road features based on the precalibration data.

Systems and methods for classifying road features

A method performed by an electronic device is described. The method includes determining a predicted velocity relative to Earth corresponding to a first epoch using a camera and an inertial measurement unit (IMU). The method also includes determining, using a Global Positioning System (GPS) receiver, a GPS velocity relative to Earth. The method further includes determining a difference vector between the predicted velocity and the GPS velocity. The method additionally includes refining a bias estimate and a scale factor estimate of IMU measurements proportional to the difference vector. The method also includes refining a misalignment estimate between the camera and the IMU based on the difference vector. The method further includes providing pose information based on the refined bias estimate, the refined scale factor, and the refined misalignment estimate.

Systems and methods for using a global positioning system velocity in visual-inertial odometry

A method for visual inertial odometry (VIO)-aided global positioning is described. The method includes updating an extended Kalman filter (EKF) state including a current pose and a sliding window of multiple prior poses. The sliding window includes poses at a number of most recent global positioning system (GPS) time epochs. Updating the EKF includes updating an EKF covariance matrix for the prior poses and the current pose in the EKF state. The method also includes determining, at a GPS epoch, a relative displacement between each of the updated prior poses and the current pose. The method further includes determining an error covariance of each of the relative displacements based on cross-covariances between each of the updated prior poses and the current pose in the EKF covariance matrix. The method additionally includes using the relative displacements and the error covariances to fuse pseudorange measurements taken over multiple epochs.

Growing community of inventors

Carlsbad, CA, United States of America

Avdhut Joshi