Learning Locomotion at Penn

The learning locomotion project is a DARPA-sponsored research project that aims to develop controllers for high-speed quadrupedal walking on rough terrain.

Machine learning techniques will be applied to a small sized four-legged walking robot walking on precisely mapped rough terrain.

Each leg of the robot has three joints. In addition, the robot also has an inertial measurement unit and foot contact sensors. The terrain that the robot will walk over is precisely scanned and mapped. Feedback will be provided by a very accurate motion capture system.

Investigators

• Daniel Lee (PI)
• Dan Koditschek
• Vijay Kumar
• George Pappas
• Lawrence Saul

Personnel

• Sachin Chitta
• David Cohen
• Roman Geykhman

Videos

• See Penn's code in action (QuickTime, 11MB)
  

---

GRASP Laboratory
Page maintained by Sachin Chitta.
Last modified: Fri Jan 06 14:32:00 Eastern Standard Time 2006