US 12,462,676 B2
Movement prediction for road users
Faris Janjos, Stuttgart (DE); and Maxim Dolgov, Renningen (DE)
Assigned to Robert Bosch GmbH, Stuttgart (DE)
Appl. No. 18/570,376
Filed by Robert Bosch GmbH, Stuttgart (DE)
PCT Filed Jun. 1, 2022, PCT No. PCT/EP2022/064936
§ 371(c)(1), (2) Date Feb. 2, 2024,
PCT Pub. No. WO2022/263175, PCT Pub. Date Dec. 22, 2022.
Claims priority of application No. 10 2021 206 014.5 (DE), filed on Jun. 14, 2021.
Prior Publication US 2024/0282190 A1, Aug. 22, 2024
Int. Cl. G06V 10/82 (2022.01); G06V 20/58 (2022.01); G08G 1/01 (2006.01)
CPC G08G 1/0125 (2013.01) [G06V 10/82 (2022.01); G06V 20/58 (2022.01); G08G 1/0112 (2013.01); G08G 1/0145 (2013.01)] 18 Claims
OG exemplary drawing
 
1. A method for predicting a movement of at least one traffic-relevant object based on observations of a surroundings of the object, the method comprising:
mapping an observation of the surroundings at a time by a trained encoder network to a first representation with reduced dimensionality;
based on a first action performed by the object at the time and the first representation, using at least one prediction network to:
determine a first representation prediction of a second representation to which a first future observation at a first future time is expected to be mapped by the trained encoder network, and/or
determine a first action prediction of a second action that the object is likely to perform at the first future time; and
determine a first dynamic state prediction for a dynamic state of the object at the first future time based on the first representation prediction and/or the first action prediction.