A plurality of microphones, mounted to an autonomous vehicle, receive audio streams 502, from which audio features are used to detect a direction to 510, and identify a class for 508 the sound source. If the source is classified as a vehicle 512, obstacle avoidance (i.e. steering, acceleration, braking) with respect to the direction of the sound source is invoked 518. The classification may be carried out using a machine learning model (i.e. deep neural network), which outputs a confidence value relating to the determined classification 512. The system may also comprise cameras for identifying potential obstacles, wherein the sound source may be hidden from the field of view of any camera.
Collision avoidance using auditory data
2017-06-21
Patent
Electronic Resource
English
Training algorithm for collision avoidance using auditory data
European Patent Office | 2018
|Collision Avoidance Using Auditory Data Augmented With Map Data
European Patent Office | 2018
|