Autonomous vehicles featuring machine-learned yield model
The present disclosure provides autonomous vehicle systems and methods that include or otherwise leverage a machine-learned yield model. In particular, the machine-learned yield model can be trained or otherwise configured to receive and process feature data descriptive of objects perceived by the autonomous vehicle and/or the surrounding environment and, in response to receipt of the feature data, provide yield decisions for the autonomous vehicle relative to the objects. For example, a yield decision for a first object can describe a yield behavior for the autonomous vehicle relative to the first object (e.g., yield to the first object or do not yield to the first object). Example objects include traffic signals, additional vehicles, or other objects. The motion of the autonomous vehicle can be controlled in accordance with the yield decisions provided by the machine-learned yield model.
1 . A computing system for an autonomous vehicle, the computing system comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:
obtaining data descriptive of respective lengths of a plurality of gaps at one or more times, the data associated with a plurality of vehicles perceived by the autonomous vehicle, wherein the plurality of gaps comprises:
a current gap between two or more of the plurality of vehicles; and
a predicted gap that is predicted to be created by at least two of the plurality of vehicles;
processing the data descriptive of the respective lengths of the plurality of gaps with a machine-learned yield model to generate an output indicative of a plurality of gap classifications respectively for the plurality of gaps;
based at least in part on the plurality of gap classifications, selecting a particular gap of the plurality of gaps through which the autonomous vehicle should proceed; and
causing the autonomous vehicle to navigate through the particular gap of the plurality of gaps.
2 . The computing system of claim 1 , wherein each of the plurality of gap classifications is indicative of an ability of the autonomous vehicle to enter the corresponding gap.
3 . The computing system of claim 1 , wherein each of the plurality of gap classifications comprises a corresponding traversal metric of a plurality of traversal metrics; and
wherein selecting the particular gap of the plurality of gaps through which the autonomous vehicle should proceed comprises:
selecting the particular gap of the plurality of gaps based at least in part on the traversal metric for the particular gap, wherein the traversal metric for the particular gap comprises a value greater than a corresponding value of any other traversal metric of the plurality of traversal metrics.
4 . The computing system of claim 1 , wherein each of the plurality of gap classifications is indicative of a binary yield decision.
5 . The computing system of claim 1 , wherein the machine-learned yield model is trained to process feature data associated with the plurality of vehicles perceived by the autonomous vehicle, wherein the output is based at least in part on the feature data.
6 . The computing system of claim 5 , wherein the feature data is indicative of at least one of: (i) a location of at least one vehicle relative to a travel way or (ii) a location of at least one vehicle relative to the autonomous vehicle.
7 . The computing system of claim 5 , wherein the feature data is indicative of at least one of: (i) an acceleration of the autonomous vehicle relative to at least one vehicle of the plurality of vehicles or (ii) a deceleration for the autonomous vehicle to yield relative to at least one vehicle of the plurality of vehicles.
8 . The computing system of claim 5 , wherein the feature data is indicative of at least one of a location or a movement of at least one vehicle in the plurality of vehicles relative to a location or a movement of another vehicle in the plurality of vehicles.
9 . The computing system of claim 1 , wherein each of the plurality of gap classifications is indicative of a jerk limit indicative of an acceleration threshold associated with passenger discomfort.
10 . The computing system of claim 1 , wherein the operations further comprise:
obtaining predicted trajectories for one or more of the plurality of vehicles, wherein the plurality of gaps are based at least in part on the predicted trajectories.
11 . The computing system of claim 1 , wherein the autonomous vehicle is an autonomous truck.
12 . A computer-implemented method comprising:
(a) obtaining data descriptive of respective lengths of a plurality of gaps at one or more times, the data associated with a plurality of vehicles perceived by an autonomous vehicle, wherein the plurality of gaps comprises:
a current gap between two or more of the plurality of vehicles; and
a predicted gap that is predicted to be created by at least two of the plurality of vehicles;
(b) processing the data descriptive of the respective lengths of the plurality of gaps with a machine-learned yield model to generate an output indicative of a plurality of gap classifications respectively for the plurality of gaps;
(c) based at least in part on the plurality of gap classifications, selecting a particular gap of the plurality of gaps through which the autonomous vehicle should proceed; and
(d) causing the autonomous vehicle to navigate through the particular gap of the plurality of gaps.
13 . The computer-implemented method of claim 12 , wherein causing the autonomous vehicle to navigate through the particular gap comprises:
generating a motion plan that comprises a trajectory for the autonomous vehicle to enter the particular gap of the plurality of gaps; and
controlling motion of the autonomous vehicle based at least in part on the trajectory.
14 . The computer-implemented method of claim 12 , wherein the machine-learned yield model is trained to generate the output based at least in part on feature data associated with the plurality of vehicles perceived by the autonomous vehicle.
15 . The computer-implemented method of claim 14 , wherein the feature data is indicative of at least one of: (i) a location of at least one vehicle or (ii) a location of at least one vehicle relative to the autonomous vehicle.
16 . The computer-implemented method of claim 14 , wherein the feature data is indicative of at least one of: (i) an acceleration for the autonomous vehicle relative to at least one vehicle of the plurality of vehicles or (ii) a deceleration for the autonomous vehicle to yield relative to at least one vehicle of the plurality of vehicles.
17 . The computer-implemented method of claim 12 , wherein each of the plurality of gap classifications is indicative of an ability of the autonomous vehicle to enter the corresponding gap.
18 . The computer-implemented method of claim 12 , wherein each of the plurality of gap classifications comprises a traversal metric; and
wherein selecting the particular gap of the plurality of gaps through which the autonomous vehicle should proceed comprises:
selecting the particular gap of the plurality of gaps based at least in part on the traversal metric for the particular gap, wherein the traversal metric for the particular gap comprises a value greater than a corresponding value of any other traversal metric of the plurality of traversal metrics.
19 . An autonomous vehicle comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the autonomous vehicle to perform operations, the operations comprising:
(a) obtaining data descriptive of respective lengths of a plurality of gaps at one or more times, the data associated with a plurality of objects perceived by the autonomous vehicle, wherein the plurality of gaps comprises:
a current gap between two or more of the plurality of objects; and
a predicted gap that is predicted to be created by at least two of the plurality of objects;
(b) processing the data descriptive of the respective lengths of the plurality of gaps with a machine-learned yield model to generate an output indicative of a plurality of gap classifications respectively for the plurality of gaps;
(c) based at least in part on the plurality of gap classifications, selecting a particular gap of the plurality of gaps through which the autonomous vehicle should proceed;
(d) generating a motion plan that comprises a trajectory for the autonomous vehicle to travel through the particular gap of the plurality of gaps; and
(e) controlling motion of the autonomous vehicle based at least in part on the trajectory.
20 . The autonomous vehicle of claim 19 , wherein the trajectory allows the autonomous vehicle to travel through the particular gap to be located between two objects of the plurality of objects.