Method and system for context-aware decision making of an autonomous agent
A system for context-aware decision making of an autonomous agent includes a computing system having a context selector and a map. A method for context-aware decision making of an autonomous agent includes receiving a set of inputs, determining a context associated with an autonomous agent based on the set of inputs, and optionally any or all of: labeling a map; selecting a learning module (context-specific learning module) based on the context; defining an action space based on the learning module; selecting an action from the action space; planning a trajectory based on the action S 260 ; and/or any other suitable processes.
1. A method for context-aware decision making of an autonomous agent, the method comprising:
receiving a set of inputs from a sensor subsystem of the autonomous agent, wherein the set of inputs includes at least a pose of the autonomous agent;
determining a context of the vehicle based on the set of inputs and a map, wherein the map prescribes a predetermined series of contexts for a fixed route of the autonomous agent;
based on the context, selecting a context-aware learning model from a set of context-aware learning models;
with the context-aware learning model, defining an action space for the autonomous agent;
selecting an action from the action space;
transmitting the action to a motion planner of the autonomous agent, wherein the motion planner determines a trajectory for the autonomous agent based on the action;
operating the autonomous agent according to the trajectory;
receiving a second set of inputs from the sensor subsystem; and
detecting that the autonomous agent is transitioning to a second context of the predetermined series of contexts based on the second set of inputs and the map.
2. The method of claim 1 , wherein the series of contexts comprises an ordered series of contexts associated with the fixed route.
3. The method of claim 1 , further comprising determining a scenario based on the context, wherein the context-aware learning model is further determined based on the scenario.
4. The method of claim 1 , wherein the trajectory is along the fixed route.
5. The method of claim 1 , wherein each of the context-aware learning models is trained based on data collected within the context.
6. The method of claim 5 , wherein the data further includes data collected during a traversal of the fixed route.
7. The method of claim 5 , wherein each of the context-aware learning models is trained with an inverse reinforcement learning algorithm.
8. The method of claim 1 , wherein the context and the selected context-aware learning model are associated in a 1:1 mapping.
9. The method of claim 1 , wherein the action space is associated with a set of parameters, wherein a set of values of the set of parameters is determined based on the context.
10. The method of claim 9 , wherein the set of values includes a creep distance.
11. The method of claim 1 , wherein the map comprises a hand labeled, high definition map.
12. The method of claim 1 , wherein the context is further determined based on the fixed route, wherein the fixed route intersects with a second fixed route in a region associated with the context.
13. The method of claim 1 , wherein the context-aware learning model further receives as input:
information associated with a set of detected dynamic objects; and
information associated with a set of static objects.
14. A system for context-aware decision making of an autonomous agent, the system comprising:
a sensor system onboard the autonomous agent, wherein the sensor system comprises:
a localization subsystem configured to receive a set of inputs, the set of inputs comprising a location parameter of the autonomous agent;
a map prescribing a predetermined series of contexts for a fixed route of the autonomous agent;
a set of context-aware learning models;
a computing system in communication with the sensor system, wherein the computing system:
receives the set of inputs from the sensor system;
compares one or more of the set of inputs with the map;
selects a context based on the comparison;
selects a context-aware learning model of the set of context-aware learning models based on the context;
determines an action for the agent based on the context-aware learning model; and
transmits the action to a motion planner of the autonomous agent, wherein the motion planner determines a trajectory for the autonomous agent based on the action.
15. The system of claim 14 , wherein the set of context-aware learning models comprises a context-aware learning model for each of the predetermined series of contexts.
16. The system of claim 15 , wherein each of the context-aware learning models is mapped to a context of the predetermined series of contexts in a 1:1 fashion.
17. The system of claim 14 , wherein each of the context-aware learning models is trained based on data collected within the context.
18. The system of claim 17 , wherein the data further includes data collected within the fixed route.
19. The system of claim 17 , wherein each of the context-aware learning models is trained with an inverse reinforcement learning algorithm.
20. The system of claim 14 , wherein the map is a hand labeled, high definition map.