two software improvements for autonomous navigation by scott triglia [email protected] the overbot...

1
Two Software Improvements for Autonomous Navigation By Scott Triglia [email protected] The Overbot Aim The project’s aim was to improve the Overbot’s navigation abilities using two techniques - obstacle inflation and error tracking. Obstacle inflation - ensures the Overbot successfully avoids obstacles without spending excess processor time making sure the edge of the vehicle does not crash. Error tracking - allows real-time correction to the inherent error of the motors and actuators moving the vehicle. Conclusion Testing of these new features on the Overbot is still early, but so far both features appear to be working. There is a noticeable improvement in speed with obstacle inflation. Although little direct testing has been undertaken, it is certain that if either of these features were missing, their absence would be immediately noticeable through higher processor load and poor correlation between expected and actual navigation respectively. Thanks to the UCSC School of Engineering, the SURF-IT Program, and the NSF for funding. Object Inflation Quickly testing a new path to detect collisions is essential for the Overbot. Object inflation accomplishes this goal. • First, flag all pixels on the map which are within the vehicle’s width of a obstacle. Shrink the vehicle to a single pixel, which now represents its center. • If this center is on a flagged pixel, then the actual vehicle has “hit” the obstacle. Since this only requires checking the pixels on the actual path, collision detection can be performed much quicker than otherwise possible. The time formerly used to check for collisions can now be devoted to other routines to enhance the Overbot’s pathfinding. Object Inflation - An example of object inflation. Before the inflation, we see isolated obstacles and the outline of the robot. After inflation, the robot is reduced to a single pixel and the space around the obstacles is flagged as inflated. Gabriel Elkaim Assistant Professor, Autonomous Systems, UCSC John Connors Graduate Student, Autonomous Systems, UCSC Error Tracking - In this example, the actual robot experiences both random error and a systematic drift to the right, which causes it to deviate from the expected path. If this systematic error is tracked and compensated for, the movement more closely resembles that predicted. Error Tracking An important part of robotics is keeping track of, and compensating for, error in motors and actuators. If ignored, error will cause the projected robot position to have no relation to the actual position. To illustrate the advantage of error tracking, consider a robot which has a systematic error in a motor, causing it to drift right when it is supposed to be driving straight. • First notice the difference in both heading and position between the simulated robot and the actual robot after they have stopped driving their motors. • This results in the robot getting lost, since the simulated heading and position will quickly bear no relation to the actual heading and position. • In this case, tracking the error and compensating for it allows the robot's actual position and heading to nearly match those predicted. Uncorrected Corrected

Upload: bertram-flynn

Post on 19-Jan-2016

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Two Software Improvements for Autonomous Navigation By Scott Triglia striglia@hmc.edu The Overbot Aim The project’s aim was to improve the Overbot’s navigation

Two Software Improvements for Autonomous Navigation

By Scott Triglia

[email protected]

The Overbot

AimThe project’s aim was to improve the Overbot’s navigation abilities using two techniques - obstacle inflation and error tracking.

Obstacle inflation - ensures the Overbot successfully avoids obstacles without spending excess processor time making sure the edge of the vehicle does not crash.

Error tracking - allows real-time correction to the inherent error of the motors and actuators moving the vehicle.

ConclusionTesting of these new features on the Overbot is still early, but so far both features appear to be working. There is a noticeable improvement in speed with obstacle inflation. Although little direct testing has been undertaken, it is certain that if either of these features were missing, their absence would be immediately noticeable through higher processor load and poor correlation between expected and actual navigation respectively.

Thanks to the UCSC School of Engineering, the SURF-IT Program, and the NSF for funding.

Object InflationQuickly testing a new path to detect collisions is essential for the Overbot. Object inflation accomplishes this goal.

• First, flag all pixels on the map which are within the vehicle’s width of a obstacle.

• Shrink the vehicle to a single pixel, which now represents its center.

• If this center is on a flagged pixel, then the actual vehicle has “hit” the obstacle.

Since this only requires checking the pixels on the actual path, collision detection can be performed much quicker than otherwise possible. The time formerly used to check for collisions can now be devoted to other routines to enhance the Overbot’s pathfinding.

Object Inflation - An example of object inflation. Before the inflation, we see isolated obstacles and the outline of the robot. After inflation, the robot is reduced to a single pixel and the space around the obstacles is flagged as inflated.

Gabriel Elkaim

Assistant Professor, Autonomous Systems, UCSC

John Connors

Graduate Student, Autonomous Systems, UCSC

Error Tracking - In this example, the actual robot experiences both random error and a systematic drift to the right, which causes it to deviate from the expected path. If this systematic error is tracked and compensated for, the movement more closely resembles that predicted.

Error TrackingAn important part of robotics is keeping track of, and compensating for, error in motors and actuators. If ignored, error will cause the projected robot position to have no relation to the actual position. To illustrate the advantage of error tracking, consider a robot which has a systematic error in a motor, causing it to drift right when it is supposed to be driving straight.

• First notice the difference in both heading and position between the simulated robot and the actual robot after they have stopped driving their motors.

• This results in the robot getting lost, since the simulated heading and position will quickly bear no relation to the actual heading and position.

• In this case, tracking the error and compensating for it allows the robot's actual position and heading to nearly match those predicted.

Uncorrected Corrected