Localization determination for vehicle operation

Patent metadata

Field	Value
Publication number	US-11112259-B2
Application number	US-201716758538-A
Country	US
Kind code	B2
Filing date	Oct 24, 2017
Priority date	Oct 24, 2017
Publication date	Sep 7, 2021
Grant date	Sep 7, 2021

Abstract

Traversing a vehicle transportation network, by a vehicle, may include determining vehicle operational information, determining a metric location estimate for the vehicle using the vehicle operational information, determining operational environment information of a portion of the vehicle transportation network, determining a topological location estimate for the vehicle within the vehicle transportation network using the metric location estimate and the operational environment information, and traversing the vehicle transportation network based on the topological location estimate for the vehicle. The operational environment information can include sensor data of a portion of the vehicle transportation network that is observable to the vehicle. To determine the metric location estimate, a non-linear loss function with a Kalman filter may mitigate effects of un-modeled sensor error(s). Techniques using Hidden Markov Models and the Earth Mover's Distance to determine the topological location estimate are also described.

First claim

What is claimed is:

1. A method of traversing a vehicle transportation network, the method comprising:
determining vehicle operational information of a vehicle, including sensing a position of the vehicle in a global coordinate system, wherein the vehicle is an autonomous vehicle or a semi-autonomous vehicle;
determining a metric location estimate of the vehicle using the vehicle operational information, including determining the metric location estimate using the sensed position as input to a non-linear loss function with a Kalman filter to mitigate effects of un-modeled sensor error, wherein the metric location estimate is an output of the Kalman filter;
determining operational environment information of a portion of the vehicle transportation network, the operational environment information including sensor data of a portion of the vehicle transportation network that is observable to the vehicle;
determining a topological location estimate of the vehicle within the vehicle transportation network using the metric location estimate and the operational environment information, wherein the topological location estimate includes a lane membership of the vehicle within the vehicle transportation network; and
traversing, by the vehicle, the vehicle transportation network using the topological location estimate of the vehicle as input to perform a vehicle operation.

2. The method of claim 1 , wherein the operational environment information comprises lane line data.

3. The method of claim 2 , wherein the operational environment information comprises remote vehicle location data, and determining the topological location estimate of the vehicle within the vehicle transportation network using the metric location estimate and the operational environment information comprises:
generating lane hypotheses for the vehicle using the lane line data and the remote vehicle location data; and
determining the topological location estimate as whichever of the lane hypotheses is more likely based on the metric location estimate.

4. The method of claim 3 , wherein generating the lane hypotheses for the vehicle using the lane line data and the remote vehicle location data comprises:
defining multiple lanes of the portion of the vehicle transportation network using the lane line data and the remote vehicle location data; and
generating the lane hypotheses based on a cardinality of the multiple lanes, the lane hypotheses comprising at least two of a first lane hypothesis wherein the topological location estimate is a left lane of a multi-lane road, a second lane hypothesis wherein the topological location estimate is a center lane of the multi-lane road, or a third lane hypothesis wherein the topological location estimate is a right lane of the multi-lane road.

5. The method of claim 3 , further comprising:
removing, from the remote vehicle location data data for any remote vehicle traveling in a direction different from the vehicle before determining the topological location estimate as whichever of the lane hypotheses is more likely based on the remote vehicle location data.

6. The method of claim 1 , wherein the operational environment information comprises remote vehicle location data, and the remote vehicle location data comprises a relative location of one or more remote vehicles to the vehicle.

7. The method of claim 1 , wherein the lane membership comprises a lane within a multi-lane road of the vehicle transportation network.

8. The method of claim 1 , wherein determining the topological location estimate of the vehicle comprises determining the lane membership within a multi-lane road of the vehicle transportation network over a series of temporal points by:
modeling the multi-lane road using a Hidden Markov Model (HMM), a solution of the HMM producing a probability distribution over lanes of the multi-lane road.

9. The method of claim 8 , wherein
states of the HMM comprise a left lane, a right lane, and a position between the left lane and the right lane,
state transition probabilities of the HMM comprise:
a first probability that the lane membership of the vehicle will remain in the left lane from a current temporal point to a subsequent temporal point;
a second probability that the lane membership of the vehicle will change from the left lane at the current temporal point to the position between the left lane and the right lane at the subsequent temporal point;
a third probability that the lane membership of the vehicle will change from the position between the left lane and the right lane at the current temporal point to the left lane at the subsequent temporal point;
a fourth probability that the lane membership of the vehicle will remain in the position between the left lane and the right lane from the current temporal point to the subsequent temporal point;
a fifth probability that the lane membership of the vehicle will change from the position between the left lane and the right lane at the current temporal point to the right lane at the subsequent temporal point;
a sixth probability that the lane membership of the vehicle will change from the right lane at the current temporal point to the position between the left lane and the right lane at the subsequent temporal point; and
a seventh probability that the lane membership of the vehicle will remain in the position between the left lane and the right lane from a current temporal point to a subsequent temporal point, and

observation probabilities of the HMM comprise respective probabilities of possible output values of the sensor data for the states.

10. The method of claim 8 , wherein a highest probability value of the probability distribution indicates the lane membership of the vehicle.

11. The method of claim 8 , wherein determining the lane membership comprises:
using the HMM within a Variable Structure Multiple Hidden Markov Model comprising multiple HMMs, each of the multiple HMMs modeling a respective possible topology for the lanes of the multi-lane road.

12. The method of claim 8 , wherein a cardinality of the lanes of the multi-lane road changes from a first temporal point to a second temporal point of the series of temporal points, and
determining the lane membership of the vehicle within the multi-lane road of the vehicle transportation network at the second temporal point comprises:
mapping the probability distribution over the lanes of the multi-lane road produced using the HMM at the first temporal point to an updated probability distribution over the lanes of the multi-lane road at the second temporal point.

13. The method of claim 12 , wherein mapping the probability distribution comprises:
mapping the probability distribution using an Earth Mover's Distance (EMD) metric.

14. The method of claim 1 , wherein:
sensing the position of the vehicle comprises sensing a global position using a global positioning system (GPS) sensor, and
determining the metric location estimate of the vehicle within the vehicle transportation network comprises:
using the sensed global position as the input to the non-linear loss function,

wherein the non-linear loss function weights sensed values of the GPS sensor; and
using an output of the non-linear loss function within the Kalman filter, wherein the Kalman filter outputs the metric location estimate.

15. An autonomous or semi-autonomous vehicle comprising:
a processor configured to execute instructions stored on a non-transitory computer readable medium to:
determine vehicle operational information of the vehicle, including sensing a position of the vehicle in a global coordinate syst

Assignees

• Nissan North America Inc
• Renault Sas

Inventors

• Nashed Samer
• Ilstrup David

Classifications

• B60W60/0011 Primary

  involving control alternatives for a single driving scenario, e.g. planning several paths to avoid obstacles · CPC title

• G01C21/3461 Primary

  Preferred or disfavoured areas, e.g. dangerous zones, toll or emission zones, intersections, manoeuvre types or segments such as motorways, toll roads or ferries · CPC title

• G06F18/295

  Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models · CPC title

• G06V20/588

  Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road · CPC title

• G01C21/3492

  employing speed data or traffic data, e.g. real-time or historical (traffic control systems for road vehicles involving transmission of navigation instructions to the vehicle G08G1/0968) · CPC title