US 11,781,874 B2
Transportation route error detection and adjustment
Shashi Kant Sharma, Redwood City, CA (US); Xianzhe Liang, Belmont, CA (US); Ricky Chachra, New York, NY (US); Xabier Azagirre Lekuona, New York, NY (US); Gerardo dela Rosa Michicol, Dublin, CA (US); Piyush Garg, San Francisco, CA (US); Adriel Frederick, Berkeley, CA (US); Chirag Chhagan Chheda, San Francisco, CA (US); Jack Chun Zhou, San Francisco, CA (US); and Amarnath Pundalika Pai, San Francisco, CA (US)
Assigned to Lyft, Inc., San Francisco, CA (US)
Filed by Lyft, Inc., San Francisco, CA (US)
Filed on Sep. 12, 2019, as Appl. No. 16/569,648.
Prior Publication US 2021/0080269 A1, Mar. 18, 2021
Int. Cl. G01C 21/34 (2006.01); G06Q 10/047 (2023.01); G06Q 50/30 (2012.01); G06Q 10/02 (2012.01); G06Q 30/0283 (2023.01); G06Q 10/063 (2023.01)
CPC G01C 21/3438 (2013.01) [G01C 21/3453 (2013.01); G06Q 10/02 (2013.01); G06Q 10/047 (2013.01); G06Q 10/063 (2013.01); G06Q 30/0284 (2013.01); G06Q 50/30 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving, by a transportation matching system, a transportation request from a mobile device, wherein the transportation request comprises at least two locations;
generating, in response to the transportation request, an initial route prediction, wherein the initial route prediction is generated utilizing a first predictive model based on a first set of previously-completed routes associated with the at least two locations;
comparing, utilizing a second predictive model, the initial route prediction generated utilizing the first predictive model with route information associated with a second set of previously-completed routes identified at least in part based on the initial route prediction;
generating, based on the comparison, a corrected route prediction for the transportation request; and
training the second predictive model by:
receiving, after fulfillment of the transportation request, actual route information corresponding to the fulfillment of the transportation request;
determining one or more errors between the corrected route prediction and the actual route information; and
updating the second predictive model based on the one or more errors.