Mobile Vision for Autonomous

1
Mobile Vision for Autonomous

• Navigation and Reconnaissance

Jay Silver Kevin Wortman
Advised by Bill Ross
2
Vision-aided Navigation for Off-road Robotics ACC
Testbed development and DARPA FCS-PerceptOR
program
3
Autonomous Navigation
4
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

5
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

6
Space Variant Resolutions
I(x,y)
I(r,q)
Distorted depth/range perception
Standard Fisheye Res constant
Accurate depth/range perception
Space Variant Res f ( r )
7
Space Variant Resolutions
I(x,y)
I(r,q)
Distorted depth/range perception
Standard Fisheye Res constant
Accurate depth/range perception
Space Variant Res f ( r )
8
Space Variant Resolutions
9
Space Variant Resolutions
Resolution function is the integral of the
angular shift w.r.t. eccentricity
Space Variant Resolution
Approximate with sum and find a least squares
best fit polynomial
10
Space Variant Resolutions
Equal shifts along the radius of expansion for
objects with equal range. Boundary tracking is
simplified.
Birds Eye View of Angular Shift
Red High Angular Shift Blue Low Angular Shift
Space-variant resolution function
Inverse Resolution - degrees/pixel
Sensor limit
Eccentricity - degrees
11
Space Variant Resolutions
Fixed resolution
Space-variant resolution
12
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

13
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

14
Multiple Resolutions
15
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

16
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

17
Path Finding with Probabilistic Obstacle Models
Minimum time to goal heavily rewarded
Danger Avoidance heavily rewarded
18
Ideal Path Finding with Probabilistic Obstacle
Models
19
Navigation

• Resolution f ( polar radius )
• Many resolutions to create new perspectives
• Path finding with probabilistic obstacle models

Access this presentation tomorrow
at www.jaysilver.net/presentations
Next
Reconnaissance