US 12,480,782 B2
Machine learning-based traffic light relevancy mapping
Adi Hayat, Modiin (IL); and Jonathan Barlev, Shoeva (IL)
Assigned to Mobileye Vision Technologies Ltd., Jerusalem (IL)
Filed by Mobileye Vision Technologies Ltd., Jerusalem (IL)
Filed on Mar. 1, 2023, as Appl. No. 18/116,084.
Claims priority of provisional application 63/315,247, filed on Mar. 1, 2022.
Prior Publication US 2023/0280183 A1, Sep. 7, 2023
Int. Cl. G01C 21/00 (2006.01)
CPC G01C 21/3841 (2020.08) [G01C 21/3819 (2020.08)] 36 Claims
OG exemplary drawing
 
1. A system for generating a crowd-sourced map for use in vehicle navigation, the system comprising:
at least one processor comprising circuitry and a memory, wherein the memory includes instructions that when executed by the circuitry cause the at least one processor to:
receive, at a server via one or more networks, drive information collected from a plurality of vehicles that traversed a road segment, wherein the road segment intersects a junction associated with a plurality of traffic lights;
aggregate, by the server, the received drive information to determine a position for each of the plurality of traffic lights and to determine a spline representation for each of one or more drivable paths associated with road segment;
generate, by the server based on the determined positions for each of the plurality of traffic lights and the spline representation for each of the one or more drivable paths, a traffic light relevancy mapping including an indicator of traffic light relevancy for each of a plurality of traffic light to drivable path pairs selected from among the plurality of traffic lights and the one or more drivable paths, wherein generating the traffic light relevancy mapping includes inputting the determined positions for each of the plurality of traffic lights and the spline representation for each of the one or more drivable paths into at least one trained model;
generate, by the server, an updated traffic light relevancy mapping based on the traffic light relevancy mapping and an observed vehicle behavior represented by the received drive information, wherein generating the updated traffic light relevancy mapping includes inputting the observed vehicle behavior into the at least one trained model to modify at least one indicator of traffic light relevancy for at least one traffic light to drivable path pair of the plurality of traffic light to drivable path pairs;
store in the crowd-sourced map, based on the updated traffic light relevancy mapping, indicators of traffic light relevancy for each of the plurality of traffic light to drivable path pairs; and
transmit the crowd-sourced map from the server via the one or more networks to at least one vehicle predicted to traverse the road segment, the at least one vehicle being configured to determine at least one navigational action for navigating the road segment based on the stored indicators of traffic light relevancy for each of the plurality of traffic light to drivable path pairs.