Aptiv Technologies Ag

Aptiv Technologies Limited

Xiaohui Wang

Nianxia Cao

James Nicholas Nickolaou

Kai Zhang

Narbik Manukian

Bin Jia

Shan Zhang

Kanishka Tyagi

Lingnan Hu

This document describes systems and techniques related to deep association for sensor fusion. For example, a model trained using deep machine learning techniques, may be used to generate an association score matrix that includes probabilities that tracks from different types of sensors are related to the same objects. This model may be trained using a convolutional recurrent neural network and include constraints not included in other training techniques. Focal loss can be used during training to compensate for imbalanced data samples and address difficult cases, and data expansion techniques can be used to increase the multi-sensor data space. Simple thresholding techniques can be applied to the association score matrix to generate an assignment matrix that indicates whether tracks from one sensor and tracks from another sensor match. In this manner, the track association process may be more accurate than current sensor fusion techniques, and vehicle safety may be increased.

Deep association for sensor fusion

This document describes techniques, systems, and methods for multi-scan sensor fusion for object tracking. A sensor-fusion system can obtain radar tracks and vision tracks generated for an environment of a vehicle. The sensor-fusion system maintains sets of hypotheses for associations between the vision tracks and the radar tracks based on multiple scans of radar data and vision data. The set of hypotheses include mass values for the associations. The sensor-fusion system determines a probability value for each hypothesis. Based on the probability value, matches between radar tracks and vision tracks are determined. The sensor-fusion system then outputs the matches to a semi-autonomous or autonomous driving system to control operation of the vehicle. In this way, the described techniques, systems, and methods can provide high-precision object tracking with quantifiable uncertainty.

Multi-scan sensor fusion for object tracking

This document describes a multiple hypothesis-based data fusion tracker. Each hypothesis aligns to a different pseudo measurement type. The fusion tracker automatically determines, using a predefined error covariance associated with the radar, which pseudo measurement type has a greater chance of being accurate for a current situation. The fusion tracker may rely on either one of two combined radar and vision calculations, or the fusion tracker may ignore the vision-based pseudo measurements and instead, rely on radar pseudo measurements alone. By selecting between three different bounding boxes, a vision angle based box, a vision lateral position based box, or a radar only based box, the fusion tracker can balance accuracy and speed when drawing, repositioning, or resizing bounding boxes, even under congested traffic or other high volume situations.

Multiple hypothesis-based fusion of sensor data

This document describes target curvature estimation considering vehicle dynamics constraints. Some vehicle dynamics models use several kinematic states to predict the paths of targets identified by an object-tracking system. Rather than rely on predetermined industry measurements as the kinematic inputs to a model, an example system instead estimates some of these states using near real-time data output from a tracking filter. This can include estimating position and velocity states, as well as curvature and centripetal acceleration states. The example system is particularly suited at accurately predicting the curvature and centripetal acceleration states, which greatly improves the accuracy of the object-tracking system, and further improves safety in avoiding a target collision.

Target curvature estimation considering vehicle dynamics

Methods and systems are described that enable radar detection prioritization based on downstream feedback. As a vehicle is traveling in a travel corridor, radar detections are received by a processor and feedback related to a downstream function is received from another processor that is executing the downstream function. For the radar detections that are determined to be within the travel corridor, respective priorities are assigned based on locations of the radar detections and the feedback related to the downstream function. The radar detections with the assigned priorities are then stored in a detection buffer along with indications of their respective assigned priorities. In this way, the detection buffer may be optimized for the downstream function by filtering out certain detections and prioritizing others. Consequently, computational loads on the downstream function are reduced while ensuring that important detections are received by the downstream function early.

Growing community of inventors

Oak Park, CA, United States of America

Xiaohui Wang