Stochastic map-aware stereo vision sensor model

Patent metadata

Field	Value
Publication number	US-10613546-B2
Application number	US-201615192603-A
Country	US
Kind code	B2
Filing date	Jun 24, 2016
Priority date	Dec 2, 2015
Publication date	Apr 7, 2020
Grant date	Apr 7, 2020

Abstract

A method for defining a sensor model includes determining a probability of obtaining a measurement from multiple potential causes in a field of view of a sensor modeled based on a stochastic map. The stochastic map includes a mean occupancy level for each voxel in the stochastic map and a variance of the mean occupancy level for each pixel. The method also includes determining a probability of obtaining an image based on the determined probability of obtaining the measurement. The method further includes planning an action for a robot, comprising the sensor, based on the probability of obtaining the image.

First claim

What is claimed is:

1. A method of defining a sensor model, comprising:
determining a probability of obtaining a measurement from a plurality of potential causes in a field of view of a sensor calculating a grid-based stochastic map of an environment, the grid-based stochastic map comprising a mean occupancy level for each voxel in the grid-based stochastic map and a variance of the mean occupancy level for each voxel, an occupancy level indicating whether a voxel is full, partially full, or empty, and each voxel corresponding to a cell in a grid of the grid-based stochastic map;
determining a probability of obtaining an image of an intended target of a robot at a current location of the robot based on the determined probability of obtaining the measurement;
updating the mean occupancy level and the variance for voxels in the field of view based on the probability of obtaining the measurement;
planning an action for the robot based on the probability of obtaining the image; and
navigating, the robot, through the environment based on the planned action.

2. The method of claim 1 , in which determining the probability of obtaining the measurement further comprises, for each potential cause of the plurality of potential causes:
calculating a likelihood of the measurement given the cause;
determining a probability of the potential cause in the grid-based stochastic map; and
combining measurement likelihoods for the plurality of potential causes and a probability of the plurality of potential causes:
to determine the probability of obtaining the measurement, and
to define the sensor model.

3. The method of claim 1 , further comprising inferring a presence of the intended target in the environment based on the determined probability of obtaining the image.

4. The method of claim 1 , in which the field of view is for one pixel of a plurality of pixels from an entire field of view of the sensor.

5. The method of claim 1 , in which the planned action comprises:
a path to the intended target when the probability of obtaining the image is greater than a threshold; and
additional measurements at a new location when the probability of obtaining the image is less than the threshold.

6. An apparatus for defining a sensor model, the apparatus comprising:
means for determining a probability of obtaining a measurement from a plurality of potential causes in a field of view of a sensor calculating a grid-based stochastic map of an environment, the grid-based stochastic map comprising a mean occupancy level for each voxel in the grid-based stochastic map and a variance of the mean occupancy level for each voxel, an occupancy level indicating whether a voxel is full, partially full, or empty, and each voxel corresponding to a cell in a grid of the grid-based stochastic map;
means for determining a probability of obtaining an image of an intended target of a robot at a current location of the robot based on the determined probability of obtaining the measurement;
means for updating the mean occupancy level and the variance for voxels in the field of view based on the probability of obtaining the measurement;
means for planning an action for the robot based on the probability of obtaining the image; and
means for navigating, the robot, through the environment based on the planned action.

7. The apparatus of claim 6 , in which the sensor comprises a stereoscopic camera.

8. The apparatus of claim 6 , in which the means for determining the probability of obtaining the measurement further comprises, for each potential cause of the plurality of potential causes:
means for calculating a likelihood of the measurement given the cause;
means for determining a probability of the potential cause in the grid-based stochastic map; and
means for combining measurement likelihoods for the plurality of potential causes and a probability of the plurality of potential causes:
to determine the probability of obtaining the measurement, and
to define the sensor model.

9. The apparatus of claim 6 , further comprising means for inferring a presence of the intended target in the environment based on the determined probability of obtaining the image.

10. The apparatus of claim 6 , in which the field of view is for one pixel of a plurality of pixels from an entire field of view of the sensor.

11. An apparatus for defining a sensor model, the apparatus comprising:
a memory; and
at least one processor coupled to the memory, the at least one processor configured:
to determine a probability of obtaining a measurement from a plurality of potential causes in a field of view of a sensor calculating a grid-based stochastic map of an environment, the grid-based stochastic map comprising a mean occupancy level for each voxel in the grid-based stochastic map and a variance of the mean occupancy level for each voxel, an occupancy level indicating whether a voxel is full, partially full, or empty, and each voxel corresponding to a cell in a grid of the grid-based stochastic map;
to determine a probability of obtaining an image of an intended target of a robot at a current location of the robot based on the determined probability of obtaining the measurement;
to update the mean occupancy level and the variance for voxels in the field of view based on the probability of obtaining the measurement;
to plan an action for the robot based on the probability of obtaining the image; and
to navigate, the robot, through the environment based on the planned action.

12. The apparatus of claim 11 , in which the sensor comprises a stereoscopic camera.

13. The apparatus of claim 11 , in which the at least one processor is further configured, for each potential cause of the plurality of potential causes:
to calculate a likelihood of the measurement given the cause;
to determine a probability of the potential cause in the grid-based stochastic map; and
to combine measurement likelihoods for the plurality of potential causes and a probability of the plurality of potential causes:
to determine the probability of obtaining the measurement, and
to define the sensor model.

14. The apparatus of claim 11 , in which the at least one processor is further configured to infer a presence of the intended target in the environment based on the determined probability of obtaining the image.

15. The apparatus of claim 11 , in which the field of view is for one pixel of a plurality of pixels from an entire field of view of the sensor.

16. A non-transitory computer-readable medium having program code recorded thereon for defining a sensor model, the program code executed by a processor and comprising:
program code to determine a probability of obtaining a measurement from a plurality of potential causes in a field of view of a sensor calculating a grid-based stochastic map of an environment, the grid-based stochastic map comprising a mean occupancy level for each voxel in the grid-based stochastic map and a variance of the mean occupancy level for each voxel, an occupancy level indicating whether a voxel is full, partially full, or empty, and each voxel corresponding to a cell in a grid of the grid-based stochastic map;
program code to determine a probability of obtaining an image of an intended target of a robot at a current location of the robot based on the determined probability of obtaining the measurement;
program code to update the mean occupancy level and the variance for voxels in the field of view based on the probability of obtaining the measurement;
program code to plan an action for the robot based on the probability of obtaining

Assignees

• Qualcomm Inc

Inventors

• Aghamohammadi Aliakbar
• Agarwal Saurav
• Omidshafiei Shayegan
• Somasundaram Kiran
• Lott Christopher
• Behabadi Bardia Fallah
• Gibson Sarah Paige
• Wierzynski Casimir Matthew
• Reitmayr Gerhard
• Diaz Spindola Serafin

Classifications

• G06T7/579

  from motion · CPC title

• G06T2207/10021

  Stereoscopic video; Stereoscopic image sequence · CPC title

• G06T2207/20076

  Probabilistic image processing · CPC title

• G06T2207/30252

  Vehicle exterior; Vicinity of vehicle · CPC title

• G05D1/0274 Primary

  using mapping information stored in a memory device (navigation using map-matching G01C21/30) · CPC title