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 decision making of an autonomous agent, the method comprising:
determining a characterization of an environment of the autonomous agent;
mapping the characterization to a learned model from a set of multiple learned models, wherein each of the set of multiple learned models is trained with an inverse reinforcement learning algorithm;
with the learned model, producing a first output;
determining a trajectory for the autonomous agent based on the first output; and
operating the autonomous agent based on the trajectory.
2. The method of claim 1 , wherein the trajectory is determined with a second learned model.
3. The method of claim 2 , wherein the second learned model is selected from a second set of multiple learned models.
4. The method of claim 1 , wherein the characterization is selected from a set of multiple characterizations.
5. The method of claim 1 , wherein the characterization is a context.
6. The method of claim 5 , wherein determining the characterization comprises selecting the context from a predetermined set of contexts.
7. The method of claim 6 , wherein the predetermined set of contexts is labeled on a map, wherein the context is selected based on the map and a pose of the autonomous agent.
8. The method of claim 1 , wherein mapping the characterization to the learned model is performed with a mapping which prescribes at most one learned model from the set of multiple learned models to the characterization, wherein the at most one learned model is the learned model.
9. The method of claim 1 , wherein the first output comprises an action associated with the autonomous agent.
10. The method of claim 9 , further comprising selecting the action from an action space defined by the learned model.
11. A system for decision making of an autonomous agent, the system comprising:
a set of multiple learned models; and
a computing system, wherein the computing system:
determines a characterization of an environment of the autonomous agent;
maps the characterization to a learned model from the set of multiple learned models, wherein each of the set of multiple learned models is trained with an inverse reinforcement learning algorithm;
produces a first output with the learned model;
determines a trajectory for the autonomous agent based on the first output; and
operates the autonomous agent based on the trajectory.
12. The system of claim 11 , wherein the computing system maps the characterization to the learned model with a predetermined mapping.
13. The system of claim 12 , wherein the predetermined mapping prescribes at most one learned model from the set of multiple learned models to the characterization, wherein the at most one learned model is the learned model.
14. The system of claim 11 , wherein the characterization is selected from a set of predetermined characterizations, wherein each of the set of predetermined characterizations is associated with a learned model of the set of multiple learned models in a 1:1 fashion.
15. The system of claim 11 , wherein the characterization comprises a context.
16. The system of claim 15 , further comprising a map, wherein the map prescribes a predetermined set of contexts, and wherein the context is selected from the predetermined set of contexts based on the map.
17. The system of claim 11 , wherein the first output comprises an action associated with the autonomous agent.
18. The system of claim 11 , wherein the trajectory is determined with a second learned model, wherein the second learned model is selected from a second set of multiple learned models.