Intelligent engine activation planner

1. A system, comprising:

a battery;

an engine; and

a processor configured to:

plan, according to a model, an activation action of the engine of a vehicle for a next road segment subsequent to a current road segment,

wherein the model includes a state space comprising a navigation map including the current road segment of the vehicle, a current charge level of the battery, and whether the engine is currently on or off, and

wherein the activation action is selected from a set comprising a first action to turn on the engine to charge the battery and a second action to turn off the engine; and

activate, for the next road segment, the engine according to the activation action.

2. The system of claim 1 , wherein to plan, according to the model, the activation action of the engine comprises to:

solve the model using an optimization technique to obtain a policy.

3. The system of claim 2 , wherein the optimization technique is a multi- objective Markov decision (MOMDP) process.

4. The system of claim 3 , wherein the MOMDP uses a cost function that includes at least one cost, wherein the at least one cost is a total energy, a wasted energy, a battery regenerated power, a battery consumed power, a fuel consumed amount, whether the engine toggled, whether the engine auto-toggled from off to on or from on to off, a time, or an extra noise level.

5. The system of claim 4 , wherein the MOMDP is solved using constrained optimization on the at least one cost.

6. The system of claim 4 , wherein the MOMDP is solved using a scalarizing function of the at least one cost.

7. The system of claim 2 , wherein the optimization technique is one of a partially observable Markov decision process (POMDP) or a classical planning (CP) process.

8. The system of claim 1 , wherein to plan, according to the model, the activation action comprises to:

plan the activation action according to an objective, wherein the objective is at least one of minimizing energy consumption of the vehicle, minimizing noise related to activation of the engine, or minimizing a number of times that the engine is toggled.

9. The system of claim 1 , wherein the state space further comprises whether a power-consuming component of the vehicle is on or off.

10. The system of claim 1 , wherein the model includes a transition function comprising at least one of a movement in edges of the navigation map, a battery change function, and a change in engine status based on the navigation map and a performed activation action.

11. A method for charging a battery of a vehicle by an engine comprising:

obtaining an engine activation policy by modeling engine activation as a Markov decision process (MDP), wherein the MDP comprises a state space, an action space, and a state transition function;

obtaining, using the engine activation policy, an activation action from the action space, wherein the action space comprises a first activation action of turning the engine on and a second activation action of turning the engine off; and

activating the engine according to the activation action, wherein activating the engine using the first activation action causes the engine to turn on to charge a batter of the vehicle.

12. The method of claim 11 , wherein the state transition function comprises a first component indicating how a location of the vehicle changes stochastically following a topological graph of the edges formed by various road segments connecting start nodes to end nodes of a navigation map.

13. The method of claim 12 , wherein the navigation map comprises nodes and edges, wherein each edge is connected by a start node and an end node of the nodes.

14. The method of claim 12 , wherein the state transition function further comprises a second component indicating a predefined engine activation based on a range of a state of charge of the battery.

15. The method of claim 14 , wherein the state transition function further comprises a third component indicating at least one of:

how a battery level of the battery changes given energy consumption or regeneration information of an edge of the navigation map;

an average traversal time of the edge;

an engine activation action;

a power at which the battery is charged from the battery;

a number of times of stopping while traversing the edge;

a number of times of traversing the edge; or

an average battery power regeneration on a stop.

16. The method of claim 15 , wherein the battery level is discretized.

17. A non-transitory computer-readable storage medium, comprising executable instructions that, when executed by a processor, facilitate performance of operations for charging a battery, the operations comprising:

determining, by optimizing an objective, an activation action of an engine of a vehicle,

wherein the activation action is selected from a set comprising a first activation action of turning the engine on and a second activation action of turning the engine off, and

wherein the battery of the vehicle is charged by the engine when the engine is turned on; and

activating the engine according to the activation action.

18. The non-transitory computer-readable storage medium of claim 17 , wherein the objective is based on at least two of an energy expenditure of the engine, a wasted energy, and an extra noise generated by the engine.

19. The non-transitory computer-readable storage medium of claim 18 , wherein the objective is to minimize, over time, a total energy consumption of the engine.

20. The non-transitory computer-readable storage medium of claim 17 , wherein costs associated with a next activation action of the engine are used in optimizing the objective, the costs comprise at least one of:

an expected energy expenditure cost;

an expected wasted energy cost;

an expected cost of the engine switching; or

an extra noise generated if the engine is activated.