US 12,124,265 B2
Autonomous decisions in traffic situations with planning control
Tommy Tram, Gothenburg (SE); and Ivo Batkovic, Gothenburg (SE)
Assigned to Zenuity AB, Gothenburg (SE)
Filed by Zenuity AB, Gothenburg (SE)
Filed on Apr. 14, 2020, as Appl. No. 16/847,785.
Claims priority of application No. 19169323 (EP), filed on Apr. 15, 2019.
Prior Publication US 2020/0326719 A1, Oct. 15, 2020
Int. Cl. G05D 1/02 (2020.01); G01C 21/34 (2006.01); G05D 1/00 (2006.01); G06N 20/00 (2019.01)
CPC G05D 1/0221 (2013.01) [G01C 21/3407 (2013.01); G05D 1/0088 (2013.01); G06N 20/00 (2019.01)] 12 Claims
OG exemplary drawing
 
1. A control device for generating maneuvering decisions for an ego-vehicle in a traffic scenario, the control device comprising:
a first circuit comprising a trained self-learning model, the first circuit being configured to:
receive data comprising information about a surrounding environment of the ego-vehicle;
determine, by means of the trained self-learning model, an action to be executed by the ego-vehicle based on the received data, the determined action corresponding to a driving maneuver to be executed by the ego-vehicle; and
a second circuit configured to:
receive the determined action from the first circuit;
receive data comprising information about the surrounding environment of the ego-vehicle during a finite time horizon;
predict an environmental state for a first time period of the finite time horizon;
determine a trajectory for the ego-vehicle based on the received action for the finite time horizon and on the predicted environmental state for the first time period; and
send a signal in order to control the ego-vehicle according to the determined trajectory during the first time period,
wherein the second circuit is further configured to:
compare the predicted environmental state with the received information about the surrounding environment of the ego-vehicle during the first time period in order to determine if the determined action is feasible based on at least one predefined criteria;
send the signal in order to control the ego-vehicle according to the determined trajectory during the first time period while the determined action is feasible based on the at least one predefined criteria; and
send a second signal to the first circuit, the second signal comprising information about the comparison between the predicted environmental state and the received information about the surrounding environment of the ego-vehicle, and wherein the first circuit is configured to:
receive the second signal transmitted by the second circuit; and
determine, by means of the trained self-learning model, the action to be executed by the ego-vehicle further based on the received second signal.