US 11,807,272 B2
Systems and methods for multiple algorithm selection
Blake Warren Wulfe, San Francisco, CA (US); Guy Rosman, Newton, MA (US); Noah J. Epstein, Cambridge, MA (US); and Luke D. Fletcher, Cambridge, MA (US)
Assigned to TOYOTA RESEARCH INSTITUTE, INC., Los Altos, CA (US)
Filed by TOYOTA RESEARCH INSTITUTE, INC., Los Altos, CA (US)
Filed on Jul. 28, 2020, as Appl. No. 16/941,434.
Prior Publication US 2022/0032957 A1, Feb. 3, 2022
Int. Cl. B60W 60/00 (2020.01); G06N 5/02 (2023.01); G07C 5/08 (2006.01); B60W 30/095 (2012.01); G08G 1/00 (2006.01); G01C 21/34 (2006.01)
CPC B60W 60/0016 (2020.02) [B60W 30/0956 (2013.01); B60W 60/0011 (2020.02); B60W 60/0027 (2020.02); G01C 21/3407 (2013.01); G06N 5/02 (2013.01); G07C 5/085 (2013.01); G08G 1/22 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A vehicle comprising:
an autonomous control system adapted to provide one or more commands to autonomously control one or more systems of the vehicle;
an obstacle recognition unit adapted to:
apply a plurality of prediction algorithms to a state of an object generating a plurality of predictions regarding object trajectory;
determine which of the plurality of prediction algorithms generates a most accurate prediction of the plurality of predictions regarding trajectory of the object by collecting log data from a plurality of vehicles and applying each of the plurality of prediction algorithms to the collected log data;
collect a dataset from the application of each of the plurality of prediction algorithms, the dataset comprising a first subset of data to be used as input to a classification algorithm, the first subset of data comprising one or more characteristics of a driving experience represented by the collected log data, and a second subset of data to be specified as a desired output of the classification algorithm, the desired output comprising one of the plurality of prediction algorithms generating the most accurate prediction; and
a navigation plan generation unit adapted to generate a navigation plan based on the most accurate prediction of the plurality of predictions regarding the trajectory of the object and according to which, the vehicle is autonomously controlled, the prediction algorithm of the plurality of prediction algorithms generating the most accurate prediction being subsequently applied to another object, generating a subsequent prediction, and generating another navigation plan or portion of the navigation plan based on the subsequent prediction.