Yield scenario encoding for autonomous systems

1 . A method comprising:

determining, based at least on perception data obtained using at least one sensor of a first vehicle in an environment, a first geometry representing a first path for the first vehicle in a scenario, a second geometry representing a second path for a second vehicle in the scenario, and at least one contention point corresponding to the first path and the second path;

matching, using the first geometry and the second geometry, the first path, the second path, and the at least one contention point to one or more traffic rules;

based at least on the matching, selecting, for the scenario, at least two categories from a plurality of categories of yield behavior, each category of yield behavior encoding a respective sequence of stages for the first vehicle to execute with respect to the second vehicle and the at least one contention point, each stage representing:

a respective yield maneuver executed during the stage, and

one or more respective conditions that, when satisfied, enable the first vehicle to transition from the respective yield maneuver to a subsequent yield maneuver in a subsequent stage of the respective sequence; and

based at least on the selecting, controlling the first vehicle through the scenario using the at least two selected categories, the controlling including, while executing the respective yield maneuver of a first category of the at least two categories, waiting to transition to the subsequent yield maneuver of the first category until a determination that:

the one or more respective conditions of the first category are satisfied, and

the one or more respective conditions of a second category of the at least two categories are satisfied or capable of being satisfied.

2 . The method of claim 1 , wherein the at least one contention point includes an entry location and an exit location for the scenario with respect to the first path, a first condition of the one or more respective conditions corresponds to the entry location, a second condition of the one or more respective conditions corresponds to the exit location, and the controlling initiates using the at least two categories based at least on the entry location and terminates using the at least two categories based at least on the exit location.

3 . The method of claim 1 , wherein the one or more respective conditions of the first category correspond to a first location of the at least one contention point and the one or more respective conditions of the second category correspond to a second location of the at least one contention point.

4 . The method of claim 1 , wherein the matching includes:

evaluating, using the first geometry, the second geometry, and the at least one contention point, propositions corresponding to the first category of yield behavior; and

determining, based at least on the evaluating, the propositions are true, wherein the selecting is based at least on the determining the propositions are true.

5 . The method of claim 1 , wherein the one or more traffic rules include at least one first traffic rule corresponding to the first category and at least one second traffic rule corresponding to a third category selected from the plurality of categories of yield behavior, and the method further includes:

detecting a conflict between the first category and the third category, wherein the matching selects the first category for the scenario over the third category based at least on determining, based at least on the conflict, that the first category has precedence over the third category of yield behavior.

6 . The method of claim 1 , wherein the one or more respective conditions of the first category includes a starting condition for the scenario, the starting condition corresponding to the first vehicle entering an intersection that represents the scenario, and the one or more respective conditions of the second category includes an ending condition for the scenario, the ending condition corresponding to the first vehicle exiting the intersection.

7 . The method of claim 1 , wherein the one or more traffic rules include:

at least one map rule corresponding to first yield behavior observed at a geographic location associated with the first vehicle and at least one base rule corresponding to second yield behavior corresponding to a geographic region associated with the first vehicle, and

wherein the matching is based at least on resolving one or more conflicts between the at least one map rule and the at least one base rule.

8 . At least one processor comprising:

one or more circuits to:

determine, based at least on sensor data obtained using at least one sensor of a first vehicle in an environment, a first geometry representing a first path for the first vehicle in a scenario, a second geometry representing a second path for a second vehicle in the scenario, and at least one contention point corresponding to the first path and the second path;

match, using the first geometry and the second geometry, the first path, the second path, and the at least one contention point to one or more traffic rules;

based at least on the matching, select, for the scenario, at least two categories from a plurality of categories of yield behavior, each category of yield behavior encoding a respective sequence of stages for the first vehicle to execute with respect to the second vehicle and the at least one contention point, each stage representing:

a respective yield maneuver executed during the stage, and

one or more respective conditions that, when satisfied, enable the first vehicle to transition from the respective yield maneuver to a subsequent yield maneuver in a subsequent stage of the respective sequence; and

based at least on the selection of the category of yield behavior, control the first vehicle through the scenario using the at least two selected categories, the controlling including, while executing the respective yield maneuver of a first category of the at least two categories, waiting to transition to the subsequent yield maneuver of the first category until a determination that:

the one or more respective conditions of the first category are satisfied, and

the one or more respective conditions of a second category of the at least two categories are satisfied or capable of being satisfied.

9 . The at least one processor of claim 8 , wherein the one or more circuits are further to detect, using perception data obtained using at least one sensor of the first vehicle in the environment, a current state of one or more traffic signals included in one or more maps, wherein the one or more respective conditions for the first category are evaluated based at least on the current state of the one or more traffic signals.

10 . The at least one processor of claim 8 , wherein the one or more circuits are further to assign to the at least one contention point a classification of a crossing point where the first path and the second path intersect or a merging point where the first path and the second path meet and join for a segment, wherein the matching includes evaluating one or more matching conditions corresponding to the one or more traffic rules using the classification.

11 . The at least one processor of claim 8 , wherein the selecting of the at least two categories occurs at a first time prior to the vehicle initiating the scenario, and the at least two categories are used for the controlling until a second time when the vehicle completes the scenario.

12 . The at least one processor of claim 8 , wherein the one or more circuits are further to classify the first geometry corresponding to the first path as a first path type and the second geometry corresponding to the second path as a second path type, wherein the matching includes evaluating, based at least on the classifying, one or more matching conditions corresponding to the one or more traffic rules using the first path type and the second path type.

13 . The at least one processor of claim 8 , wherein prior to the matching each category of the plurality of categories of yield behavior is stored in a machine-readable representation that specifies the respective sequence of stages and the one or more respective conditions.

14 . The at least one processor of claim 8 , wherein the one or more traffic rules include one or more map rules associated with localizing the first vehicle to at least one map or one or more base rules associated with a geographic region of the first vehicle, the one or more base rules associated with the first category and the one or more map rules associated with a third category of yield behavior that conflicts with the first category.

15 . A system comprising:

one or more processing units to:

determine, based at least on sensor data obtained using at least one sensor of a first vehicle in an environment, a first geometry representing a first path for the first vehicle in a scenario, a second geometry representing a second path for a second vehicle in the scenario, and at least one contention point corresponding to the first path and the second path;

match, using the first geometry and the second geometry, the first path, the second path, and the at least one contention point to one or more traffic rules;

based at least on the matching, selecting, for the scenario, at least two categories from a plurality of categories of yield behavior, each category of yield behavior encoding a respective sequence of stages for the first vehicle to execute with respect to the second vehicle and the at least one contention point, each stage representing:

a respective yield maneuver executed during the stage, and

one or more respective conditions that, when satisfied, enable the first vehicle to transition from the respective yield maneuver to a subsequent yield maneuver in a subsequent stage of the respective sequence; and

based at least on the selection of the category of yield behavior, control the first vehicle through the scenario using the at least two selected categories, the controlling including, while executing the respective yield maneuver of a first category of the at least two categories, waiting to transition to the subsequent yield maneuver of the first category until a determination that:

the one or more respective conditions of the first category are satisfied, and

the one or more respective conditions of a second category of the at least two categories are satisfied or capable of being satisfied.

16 . The system of claim 15 , wherein the one or more respective conditions of the first category include a condition for the first vehicle to stop, await further instructions, and not proceed until a state of contention regarding the second vehicle has changed.

17 . The system of claim 15 , where the one or more processing units are further to:

receive perception data that includes first geometry information associated with at least the first path;

receive map data representative of second geometry information associated with at least the first path; and

generate the first geometry based at least on fusing the first geometry information with the second geometry information.

18 . The system of claim 15 , where the one or more processing units are further to:

receive perception data that includes first signal information associated with one or more detected traffic signals in the environment;

receive map data including second signal information associated with one or more localized traffic signals from at least one map; and

generate fused signal information associated with at least one traffic signal based at least on fusing the first signal information with the second signal information, wherein the matching is based at least on the fused signal information.

19 . The system of claim 15 , where the one or more processing units are further to classify the second path as belonging to at least one class, wherein the matching is based at least on the second path belonging to the at least one class.

20 . The system of claim 15 , where the one or more processing units are comprised in at least one of:

a control system for an autonomous or semi-autonomous machine;

a perception system for an autonomous or semi-autonomous machine;

a system for performing simulation operations;

a system for performing deep learning operations;

a system implemented using an edge device;

a system implemented using a robot;

a system incorporating one or more virtual machines (VMs);

a system implemented at least partially in a data center; or

a system implemented at least partially using cloud computing resources.