PROBABILISTIC AUTONOMOUS VEHICLE ROUTING AND NAVIGATION

• US 20150345967A1
• Filed: 06/03/2014
• Published: 12/03/2015
• Est. Priority Date: 06/03/2014
• Status: Abandoned Application

• Current Assignee: Nissan North America Incorporated (Nissan Motor Co., Ltd.)
• Original Assignee: Nissan North America Incorporated (Nissan Motor Co., Ltd.)
• Inventors: MEULEAU, NICOLAS
• Application Number:
  US14/294,462
• Publication Number:
  US 20150345967A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G01C 21/3453 Special cost functions, i.e. other than distance or default speed limit of road segments

Abstract

A method and apparatus for probabilistic autonomous vehicle routing and navigation are disclosed. Probabilistic autonomous vehicle routing and navigation may include an autonomous vehicle identifying transportation network information, identifying an origin, identifying a destination, generating a plurality of candidate routes from the origin to the destination based on the transportation network information, wherein each route from the plurality of routes indicates a distinct combination of road segments and lanes, generating an action cost probability distribution for each action in each candidate route, generating a route cost probability distribution based at least in part on the action cost probability distribution, identify an optimal route from the plurality of candidate routes based at least in part on the route cost probability distribution, and operate the autonomous vehicle to travel from the origin to the destination using the optimal route.

Claims

1. An autonomous vehicle comprising:

a processor configured to execute instructions stored on a non-transitory computer readable medium to;

identify transportation network information, the transportation network information including road segment information representing a plurality of road segments, the road segment information including lane information representing at least one lane for each respective road segment, the lane information including waypoint information representing at least one waypoint for each respective lane,

identify an origin,

identify a destination,

generate a plurality of candidate routes from the origin to the destination based on the transportation network information, wherein each route from the plurality of routes indicates a distinct combination of road segments and lanes,

for at least one candidate route from the plurality of candidate routes;

identify a first routing state, the first routing state including an indication of a first road segment, an indication of a first lane associated with the first road segment, and an indication of a first waypoint associated with the first lane;

identify a second routing state, the second routing state including an indication of a second road segment, an indication of a second lane associated with the second road segment, and an indication of a second waypoint associated with the second lane, such that the second waypoint is immediately adjacent to the first waypoint;

generate an action cost probability distribution including a plurality of action cost probabilities, each action cost probability from the plurality of action cost probabilities representing a probable cost of transitioning from the first routing state to the second routing state; and

generate a route cost probability distribution based at least in part on the action cost probability distribution, the route cost probability distribution including a plurality of route cost probabilities, each route cost probability from the plurality of route cost probabilities representing a probable cost of traveling from the origin to the destination using the candidate route,

identify an optimal route from the plurality of candidate routes based at least in part on the route cost probability distribution, the optimal route having a minimal probable route cost; and

a trajectory controller configured to operate the autonomous vehicle to travel from the origin to the destination using the optimal route.

2. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

identify the optimal route using a Markov decision process or a Hybrid Markov decision process.

3. The autonomous vehicle of claim 1, wherein the road segment information for at least one road segment from the plurality of road segments includes lane information representing a plurality of adjacent lanes, and wherein at least one candidate route from the plurality of candidate routes includes an action that represents a transition from a first adjacent lane from the plurality of adjacent lanes to a second adjacent lane from the plurality of adjacent lanes.

4. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution using a normal distribution and an action cost uncertainty variance modifier.

5. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution by generating each route cost probability from the plurality of route cost probabilities as a combination of a discrete cost and a discrete probability associated with the discrete cost.

6. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution using a linear model of resources and costs.

7. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution using piece-wise constant functions representing action transition probability distributions.

8. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution using piece-wise linear functions representing action transition costs.

9. The autonomous vehicle of claim 1, wherein each action cost probability from the plurality of action cost probabilities is based on at least one of a plurality of cost metrics.

10. The autonomous vehicle of claim 9, wherein the plurality of cost metrics includes at least one of a distance cost metric, a duration cost metric, a fuel cost metric, or an acceptability cost metric.

11. The autonomous vehicle of claim 1, further comprising:

a geographic location unit, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to identify the origin by;

controlling the geographic location unit identify a currently geographic location of the autonomous vehicle; and

identifying a waypoint from the transportation network information, the waypoint proximal to the current geographic location.

12. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to output or store the optimal route.

13. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

receive, from an off-vehicle sensor, current transportation network state information indicating a state of at least a portion of at least one road segment from the plurality of road segments; and

generate the action cost probability distribution based on the current transportation network state information.

14. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate an updated action cost probability distribution including at least one updated action cost probability representing an updated probable cost of transitioning from the first routing state to the second routing state; and

identify an updated optimal route based on the updated action cost probability distribution, wherein the trajectory controller is configured to operate the autonomous vehicle to travel from the origin to the destination using at least a portion of the optimal route and at least a portion of the updated optimal route.

15. The autonomous vehicle of claim 14, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the updated action cost probability distribution in response to receiving, from an off-vehicle sensor, current transportation network state information indicating a state of at least a portion of at least one road segment from the plurality of road segments.

16. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

update the plurality of candidate routes in response to successfully transitioning from the first routing state to the second routing state, wherein updating the plurality of candidate routes includes;

generating an updated plurality of candidate routes from the second routing state to the destination based on the transportation network information, wherein each candidate route from the updated plurality of candidate routes indicates a distinct combination of road segments and lanes,

for at least one candidate route from the updated plurality of candidate routes;

identify a third routing state, the third routing state including an indication of a third road segment, an indication of a third lane associated with the third road segment, and an indication of a third waypoint associated with the third lane;

identify a fourth routing state, the fourth routing state including an indication of a fourth road segment, an indication of a fourth lane associated with the fourth road segment, and an indication of a fourth waypoint associated with the fourth lane, such that the fourth waypoint is immediately adjacent to the third waypoint;

generate an updated action cost probability distribution including a plurality of updated action cost probabilities, each updated action cost probability from the plurality of updated action cost probabilities representing a probable cost of transitioning from the third routing state to the fourth routing state; and

generate an updated route cost probability distribution based at least in part on the updated action cost probability distribution, the updated route cost probability distribution including a plurality of updated route cost probabilities, each updated route cost probability from the plurality of updated route cost probabilities representing a probable cost of traveling from the second routing state to the destination using the updated candidate route,

identify an updated optimal route from the plurality of updated candidate routes based at least in part on the updated route cost probability distribution, the updated optimal route having a minimal probable route cost; and

wherein the trajectory controller is configured to operate the autonomous vehicle to travel from the second routing state to the destination using the updated optimal route.

17. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate, in response to successfully transitioning from the first routing state to the second routing state and in response to receiving a different destination, a different plurality of candidate routes from the second routing state to the different destination based on the transportation network information, wherein each candidate route from the different plurality of candidate routes indicates a distinct combination of road segments and lanes,

for at least one candidate route from the different plurality of candidate routes;

identify a third routing state, the third routing state including an indication of a third road segment, an indication of a third lane associated with the third road segment, and an indication of a third waypoint associated with the third lane;

identify a fourth routing state, the fourth routing state including an indication of a fourth road segment, an indication of a fourth lane associated with the fourth road segment, and an indication of a fourth waypoint associated with the fourth lane, such that the fourth waypoint is immediately adjacent to the third waypoint;

generate an different action cost probability distribution including a plurality of different action cost probabilities, each different action cost probability from the plurality of different action cost probabilities representing a probable cost of transitioning from the third routing state to the fourth routing state; and

generate an different route cost probability distribution based at least in part on the different action cost probability distribution, the different route cost probability distribution including a plurality of different route cost probabilities, each different route cost probability from the plurality of different route cost probabilities representing a probable cost of traveling from the second routing state to the destination using the different candidate route,

identify a different optimal route from the plurality of different candidate routes based at least in part on the different route cost probability distribution, the different optimal route having a minimal probable route cost; and

wherein the trajectory controller is configured to operate the autonomous vehicle to travel from the second routing state to the different destination using the different optimal route.

18. The autonomous vehicle of claim 1, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution based on a temporal context.

19. An autonomous vehicle comprising:

a processor configured to execute instructions stored on a non-transitory computer readable medium to;

identify transportation network information, the transportation network information including road segment information representing a plurality of road segments, the road segment information including lane information representing at least one lane for each respective road segment, the lane information including waypoint information representing at least one waypoint for each respective lane,

identify an origin,

identify a destination,

generate a plurality of candidate routes from the origin to the destination based on the transportation network information, wherein each route from the plurality of routes indicates a distinct combination of road segments and lanes,

for at least one candidate route from the plurality of candidate routes;

identify a first routing state, the first routing state including an indication of a first road segment, an indication of a first lane associated with the first road segment, and an indication of a first waypoint associated with the first lane;

identify a second routing state, the second routing state including an indication of a second road segment, an indication of a second lane associated with the second road segment, and an indication of a second waypoint associated with the second lane, such that the second waypoint is immediately adjacent to the first waypoint;

generate an action cost probability distribution including a plurality of action cost probabilities, each action cost probability from the plurality of action cost probabilities representing a probable cost of transitioning from the first routing state to the second routing state; and

generate a route cost probability distribution based at least in part on the action cost probability distribution, the route cost probability distribution including a plurality of route cost probabilities, each route cost probability from the plurality of route cost probabilities representing a probable cost of traveling from the origin to the destination using the candidate route,

identify an optimal route from the plurality of candidate routes based at least in part on the route cost probability distribution, the optimal route having a minimal probable route cost;

receive, from an off-vehicle sensor, current transportation network state information indicating a state of at least a portion of at least one road segment from the plurality of road segments;

generate an updated action cost probability distribution including at least one updated action cost probability representing an updated probable cost of transitioning from the first routing state to the second routing state;

identify an updated optimal route based on the updated action cost probability distribution; and

a trajectory controller configured to operate the autonomous vehicle to travel from the origin to the destination using at least a portion of the optimal route and at least a portion of the updated optimal route.

20. The autonomous vehicle of claim 19, wherein the processor is configured to execute instructions stored on the non-transitory computer readable medium to:

generate the action cost probability distribution based on a temporal context; and

generate the updated cost probability distribution based on the temporal context.