Carnegie Mellon University
Advanced Search   
  Look in
       Title     Description
       Inactive Projects

Automated Turf Management
This project is no longer active.
Head: Sanjiv Singh
Contact: Sanjiv Singh
Mailing address:
Carnegie Mellon University
Robotics Institute
5000 Forbes Avenue
Pittsburgh, PA 15213
Project Homepage
To achieve complete automation on golf courses, we are developing capabilities for:
  • Reliable obstacle detection.
  • Must be able to find obstacles as small as a golf ball, while not generating false positives.
  • Also must detect all true obstacles to keep the vehicle safe.
  • Precise navigation. Must be able to operate with cm level precision to create the cross-hatch patterns seen on premier golf courses.
  • Effective coverage. Must be able to create patterns that cover the entire fairway in an efficient manner.

The key benefits of this technology are

  • reduction in reliance on skilled operators, and
  • expanded hours of golf course availability.
Automated mowing has enormous further commercial potential, since it would be potentially attractive to any lawn owner.

The first approach to obstacle detection is

  • color segmentation, which is the ability to distinguish obstacles based on color. We are developing a method that is robust to changes in color due to shadows and atmospheric conditions.
The second approach we are investigating is
  • stereo based homography, to detect objects which lie above the ground plane.
We are also using a
  • SickTM Laser scanner as a backup detection system, for instances where vision-based methods fail.


  • As of December 2000, we have implemented path tracking and obstacle detection on the mower. We are doing runs of over 1 km in length. The obstacle detection system has been enhanced with the addition of a Sick laser sensor, for use in validating obstacles in situations where vision cannot perform.
  • In August 2000, we held a successful demonstration of this system in which we performed over 50 iterations of our demonstration. Each iteration consisted of mowing six rows, each 10 meters long.