Deep neural network for detecting obstacle instances using radar sensors in autonomous machine applications

US11885907B2 · kind B2 · utility

Assignee

• NVIDIA Corporation · US

Inventors

• Alexander Popov · Kirkland, US
• Nikolai Smolyanskiy · Seattle, US
• Ryan Oldja · Redmond, US
• Shane Murray · San Jose, US
• Tilman Wekel · Santa Clara, US
• David Nister · Bellevue, US
• Joachim Pehserl · Lynnwood, US
• Ruchi Bhargava · Redmond, US
• Sangmin Oh · San Jose, US

Key dates

• Filing date: Mar 31, 2020
• Grant date: Jan 30, 2024
• Expiry date: Dec 7, 2040

Classification

• Technology area (CPC G): Physics
• CPC primary: G06V2201/12
• WIPO field: Measurement
• WIPO sector: Instruments

Abstract

In various examples, a deep neural network(s) (e.g., a convolutional neural network) may be trained to detect moving and stationary obstacles from RADAR data of a three dimensional (3D) space, in both highway and urban scenarios. RADAR detections may be accumulated, ego-motion-compensated, orthographically projected, and fed into a neural network(s). The neural network(s) may include a common trunk with a feature extractor and several heads that predict different outputs such as a class confidence head that predicts a confidence map and an instance regression head that predicts object instance data for detected objects. The outputs may be decoded, filtered, and/or clustered to form bounding shapes identifying the location, size, and/or orientation of detected object instances. The detected object instances may be provided to an autonomous vehicle drive stack to enable safe planning and control of the autonomous vehicle.