US 11,892,857 B2
Distributed data sampling
John Hayes, Mountain View, CA (US); Volkmar Uhlig, Cupertino, CA (US); and Akash J. Sagar, Redwood City, CA (US)
Assigned to GHOST AUTONOMY INC., Mountain View, CA (US)
Filed by GHOST AUTONOMY INC., Mountain View, CA (US)
Filed on Jun. 5, 2020, as Appl. No. 16/893,852.
Prior Publication US 2021/0382499 A1, Dec. 9, 2021
Int. Cl. G05D 1/02 (2020.01); G06F 16/245 (2019.01); G06N 20/00 (2019.01); G05D 1/00 (2006.01); G01C 21/34 (2006.01); G05B 13/02 (2006.01)
CPC G05D 1/0287 (2013.01) [G01C 21/3453 (2013.01); G05B 13/029 (2013.01); G05D 1/0011 (2013.01); G05D 1/0088 (2013.01); G06F 16/245 (2019.01); G06N 20/00 (2019.01)] 12 Claims
OG exemplary drawing
 
1. A method for distributed data sampling, comprising:
receiving, by a vehicle, a sampling target, wherein the sampling target comprises a first multidimensional value comprising a first plurality of values associated with sensor data and a second plurality values indicating targeted driving actions, wherein the second plurality of values include a position of a steering wheel and at least one of: an amount of brake applied or an amount of acceleration applied;
generating, by the vehicle, a plurality of portions of sampled data, wherein generating the plurality of portions of sampled data comprises:
generating, for each portion of the sampled data, using one or more sensors of the vehicle, a portion of sensor data; and
generating, for each portion of the sampled data, data indicating driving actions input to the vehicle from a driver when a corresponding portion of the sensor data was generated, wherein the data indicating the driving actions corresponds to the second plurality of values of the sampling target;
wherein each portion of the sampled data is expressed as a second multidimensional value;
determining, of each portion of the sampled data, a degree of similarity between a respective portion of the sampled data and the sampling target, wherein the degree of similarity is based on both the portion of the sensor data and the data indicating the driving actions for the respective portion of the sampled data, wherein the degree of similarity comprises a distance in multidimensional space between the first multidimensional value of the sampling target and the second multidimensional value of the respective portion of the sampled data;
generating a selected set of sampled data comprising a first subset of the plurality of portions of sampled data having degrees of similarity exceeding a threshold;
providing, from the vehicle via a network, the selected set to a remotely disposed computing device; and
deleting, from storage of the vehicle, a remainder of the plurality of portions of sampled data not included in the selected set without providing the remainder to the remotely disposed computing device.