A vehicle headlight illumination intensity control system, comprising: an object feature network 720 that locates and identifies objects in an environment from data provided by a perception apparatus 706, (900, Fig. 9); an image feature network 710 that identifies environmental (ambient) illumination information from an image 704; and a feature fusion network 730 that receives the object features and the image features and outputs a classification of whether the vehicle should use a high illumination intensity (high beams) or a low illumination intensity (low beams) 710. The feature fusion network may comprise a multilayer perceptron that operates on concatenated 707 object features and the image features. The object feature network may be trained using manual driving data (816, Fig.8) and outputs generated by the perception system (814A, Fig.8). The image feature network may be trained using manual driving data and image sensor data (812A, Fig.8). The object feature network, image feature network, and feature fusion network may be retrained once deployed using low occurrence data conflicting data to satisfy a predetermined safety confidence level (832, Fig.8).
Deep learning based beam control for autonomous vehicles
2023-05-31
Patent
Electronic Resource
English
IPC: | G06V / B60Q ARRANGEMENT OF SIGNALLING OR LIGHTING DEVICES, THE MOUNTING OR SUPPORTING THEREOF OR CIRCUITS THEREFOR, FOR VEHICLES IN GENERAL , Anordnung von Signal- oder Beleuchtungsvorrichtungen, deren Einbau oder Halterung oder deren Schaltkreise bei Fahrzeugen allgemein / F21S Ortsfeste Leuchten , NON-PORTABLE LIGHTING DEVICES |
DEEP LEARNING BASED BEAM CONTROL FOR AUTONOMOUS VEHICLES
European Patent Office | 2023
|Deep learning based beam control for autonomous vehicles
European Patent Office | 2023
|Deep Learning Based Beam Control for Autonomous Vehicles
European Patent Office | 2023
|Deep Learning & Autonomous Vehicles
TIBKAT | 2018
|