March 7, 2007. Walked across Terrain Board G with raised start and end flat boards at 3.1 cm/sec.

November 14, 2008: Bounding at 16cm/s

March 3, 2007. Video demonstrating learning a Terrain Cost Map by walking the dog on various terrain boards and then generalizing to an uncharted terrain board.

Feb 06, 2007. Walked across the terrain shown in the picture. The total time to calculate the plan was about 30sec and the average walking speed from start to goal was about 1.35cm/s.

Feb 06, 2007. Crossed the terrain shown above (flat with a 2x4, terrain G, and step rotated and tilted). Average speed was approximately 1.5cm/s.

Feb 06, 2007. This run is titled "I'm not dead yet". It was pretty sloppy, but shows the persistence of the controller.

October 30, 2006. Climbed over Test07 Terrain with choreographed steps chosen by a human. The CoM shifts and the leg swing trajectories are automatically computed such that they avoid body collisions with the terrain, swing foot stubbing, shin stubbing, and result in stable support polygons. The walk in this video is currently extremely slow (about 6 minutes to cross). This file is large (300MB) so download to your hard drive first. We will be putting a smaller one up...

October 30, 2006. Climbed to high point of Test07 terrain board and got two legs on one of the tall rocks. Footsteps were choreographed (chosen by a human). The CoM shifts and the leg swing trajectories are automatically computed such that they avoid body collisions with the terrain, swing foot stubbing, shin stubbing, and result in stable support polygons.

October 30, 2006. Biologically Inspired Penguin Slide and recovery from tall rock.

October 30, 2006. Biologically Inspired Penguin Slide, recovery, fall, and recovery.

August 30, 2006. Climbed up step 7.5 cm high using new pitch and height algorithm.

August 30, 2006. Demonstration of force control. The knee joint is under open loop torque control. The two hip joints are under position control, with their desired positions such that the foot stays along a curve. The location on the curve is governed by the actual position of the knee joint. This technique will be useful for soft foot placement with guaranteed contact with the ground.

Test 04, August 14, 2006, at U Penn. Rocks tilted at 6.5 degrees and mesa tilted at 6.3 degrees in the opposite direction. Our third run shown here successfully completed in 38 seconds.

Video demonstrating walking over a step that achieves the Phase I metrics (6.4 cm height, 1.6 cm/sec), August 18, 2006, at IHMC. Mesa was raised to provide a 7.5cm obstacle. Rocks added to provide 1.2m run. Run complete in 41 seconds (2.9cm/sec).

Test 03, July 6, 2006. Rocks tilted at 6.5 degrees and mesa tilted at 6.3 degrees in the opposite direction. Our 3 official trials all failed due to IMU pitch offset. After fixing that problem we took this unofficial run that successfully completed in 43 seconds.

GUI view of showing foot placement. Video shows actual foot position, target foot position, and desired foot position.

Overhead view of gait. View shows foot positions, body position, support polygon, and COM position.