US 12,254,424 B2
Leveraging traffic patterns to understand traffic rules
Sammy Omari, Menlo Park, CA (US); and Sameer Qureshi, Sunnyvale, CA (US)
Assigned to Lyft, Inc., San Francisco, CA (US)
Filed by Lyft, Inc., San Francisco, CA (US)
Filed on Dec. 2, 2019, as Appl. No. 16/700,544.
Prior Publication US 2021/0166145 A1, Jun. 3, 2021
Int. Cl. G06N 7/01 (2023.01); G06N 20/00 (2019.01); G08G 1/01 (2006.01); G08G 1/056 (2006.01)
CPC G06N 7/01 (2023.01) [G06N 20/00 (2019.01); G08G 1/0129 (2013.01); G08G 1/056 (2013.01)] 24 Claims
OG exemplary drawing
 
1. A method comprising:
collecting, by a fleet of a plurality of vehicles managed by a transportation management system, vehicle traffic-pattern data in a region of a road network, wherein the vehicle traffic-pattern data are captured by a plurality of sensors associated with the plurality of vehicles when operating in the region;
determining, by the transportation management system, a connectivity model associated with the region, wherein the connectivity model was trained using the vehicle traffic-pattern data in the region, wherein the vehicle traffic-pattern data comprises:
a first lane identifier indicating a first lane and a second lane identifier indicating a second lane, wherein the first and second lanes are associated with a vehicle's trajectory through the region, and
a traffic-light state corresponding to signal information of a plurality of traffic lights in the region when the vehicle turned from the first lane to the second lane through the region;
determining, by the transportation management system and based on the connectivity model, for at least one egress lane in the region:
a lane relationship indicating one or more ingress lanes in the region which vehicles in the at least one egress lane can turn into; and
virtual lane dividers defining an area which vehicles turning from the at least one egress lane into the one or more ingress lanes should stay within;
identifying, by the transportation management system and based on the connectivity model, for the at least one egress lane in the region, one or more first traffic lights from the plurality of traffic lights in the region as one or more governing traffic lights that govern the at least one egress lane;
encoding, by the transportation management system, the lane relationship, the virtual lane dividers, and the one or more governing traffic lights into a map of the region; and
transmitting, by the transportation management system, the map of the region to a vehicle, wherein the map of the region is configured to be used by the vehicle to generate one or more navigation paths based on the lane relationship, the virtual lane dividers, and the one or more governing traffic lights encoded in the map of the region.
 
14. The method of claim 1, further comprising training the connectivity model by the transportation management system, wherein the training comprises:
inputting, to a probability function, the vehicle traffic-pattern data;
outputting, by the probability function, a probability distribution over one or more model-parameters associated with the connectivity model; and
determining, based on the probability distribution, one or more model parameter-values for the one or more model-parameters, respectively.
 
19. A system comprising: one or more processors and one or more computer-readable non-transitory storage media coupled to one or more of the processors, the one or more computer-readable non-transitory storage media comprising instructions operable when executed by one or more of the processors to cause the system to:
collect, by a fleet of a plurality of vehicles managed by a transportation management system, vehicle traffic-pattern data in a region of a road network, wherein the vehicle traffic-pattern data are captured by a plurality of sensors associated with the plurality of vehicles when operating in the region;
determine, by a transportation management system, a connectivity model associated with the region, wherein the connectivity model was trained using the vehicle traffic-pattern data in the region, wherein the vehicle traffic-pattern data comprises:
a first lane identifier indicating a first lane and a second lane identifier indicating a second lane, wherein the first and second lanes are associated with a vehicle's trajectory through the region, and
a traffic-light state corresponding to signal information of a plurality of traffic lights in the region when the vehicle turned from the first lane to the second lane through the region;
determine, by the transportation management system and based on the connectivity model, for at least one egress lane in the region:
a lane relationship indicating one or more ingress lanes in the region which vehicles in the at least one egress lane can turn into; and
virtual lane dividers defining an area which vehicles turning from the at least one egress lane into the one or more ingress lanes should stay within;
identify, by the transportation management system and based on the connectivity model, for the at least one egress lane in the region, one or more first traffic lights from the plurality of traffic lights in the region as one or more governing traffic lights that govern the at least one egress lane;
encode, by the transportation management system, the lane relationship, the virtual lane dividers, and the one or more governing traffic lights into a map of the region; and
transmit, by the transportation management system, the map of the region to a vehicle, wherein the map of the region is configured to be used by the vehicle to generate one or more navigation paths based on the lane relationship, the virtual lane dividers, and the one or more governing traffic lights encoded in the map of the region.