Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Metric sensing and control of a quadrotor using a homography-based visual inertial fusion method
Li P., Garratt M., Lambert A., Lin S. Robotics and Autonomous Systems76 1-14,2016.Type:Article
Date Reviewed: Jul 20 2016

Is a locust a computer?

A locust is certainly an autonomous rotary-wing micro aerial vehicle (RMAV), and command and control of RMAVs is important for applications. How does one duplicate the locust’s proprioception using a monocular camera and an inertial measurement unit (IMU)? Larger airborne vehicles, that is, airplanes and helicopters, use heavy mechanical gyroscopes or ring laser gyros to determine attitude, that is, the pitch, yaw, and roll angles that specify where the vehicle is pointed. But these require too much power (either due to weight or computational requirements) for an RMAV.

This very wide-ranging paper approaches pose estimation through optic flow (OF), the displacement of features between consecutive frames. This is a sensor fusion problem: OF interacts with a computational model (homography) and inertial measurements to estimate vehicle speed over the ground (requiring a Kalman filter). The result is “adequate” closed loop control.

The bibliography (59 items) includes literature from robotics, circuits, numerical analysis, aircraft guidance, navigation and control, biological (connectionist) models, pattern matching, computer vision, automated cartography, image understanding, cybernetics, and computational physiology. (I was not personally aware that computational physiology was a field.) Concepts from optics also appear. So many fields are involved that the paper could almost have universal interest, except that it presumes a level of expertise in each area.

The original data acquisition was obtained by running around a field with a quadcopter before launching the vehicle on its own.

There are a few interesting surprises. For example, it works out that binary intensity is more robust for OF estimation as well as computationally more efficient. And the paper is terse: “Although the outlier ratio drops as the number of iterations ... increases, the improvement is small.” Was this assertion tested using a Monte Carlo method? (Probably.) How many iterations were performed? How small was the improvement?

Finally, what about locusts? The last sentence of the paper proposes a model based on locust behavior to extend the method to include more vertical data. So include animal computation as one of the paper’s topics.

Reviewer:  J. Wolper Review #: CR144610 (1611-0856)
Bookmark and Share
 
Autonomous Vehicles (I.2.9 ... )
 
 
Geometrical Problems And Computations (F.2.2 ... )
 
 
Motion (I.4.8 ... )
 
 
Sensor Fusion (I.4.8 ... )
 
 
Sensors (I.2.9 ... )
 
Would you recommend this review?
yes
no
Other reviews under "Autonomous Vehicles": Date
Mobile robotics: a practical introduction
Nehmzow U., Springer-Verlag New York, Inc., Secaucus, NJ, 2003.  304, Type: Book (9781852337261)
Aug 20 2003
Evolution of behaviors in autonomous robot using artificial neural network and genetic algorithm
Lee M. Information Sciences 155(1-2): 43-60, 2003. Type: Article
May 19 2004
RoboCup-2003 new scientific and technical advances
Pagello E., Menegatti E., Bredenfel A., Costa P., Christaller T., Jacoff A., Polani D., Riedmiller M., Saffiotti A., Sklar E., Tomoichi T. AI Magazine 25(2): 81-98, 2004. Type: Article
Mar 24 2005
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy