In this early experiment, Metalman is given the task of stacking three randomly placed blocks into a tower. The objects are detected using coloured markers on both the top an bottom surfaces. This demonstration appeared in [Taylor and Kleeman, 2001].

• MPEG-1 stacking coloured blocks (10.4 MB)

This experiment combines object classification and visual servoing with wind and odour sensing. Metalman is given the task of finding the cups and predicting which one contains ethanol. The cup is then grasped and brought closer to the nose to validate the prediction. This demonstration first appeared in [Russell et al.].

• MPEG-1 Multi-Sensory synergies in humanoid robotics (13.6 MB)

This early experiment demonstrates interaction with Metalman using motion and colour detection. When motion (caused by a tapping finger) is detected in the vicinity of a randomly placed coloured marker, the robot "taps" (using visual servoing) on the selected target.

• MPEG-1 motion-based interaction (6.85 MB)

This video shows the complete process of grasping a yellow box, including light stripe scanning, segmentation, planning and visual servoing. Both the segmentation and visual servoing in this video are earlier versions of the methods presented in the thesis. This demonstration appeared in [Taylor and Kleeman, 2002].

• MPEG-1 grasping a yellow box (4.66 MB)

This video details the robust stereo light stripe scanning process for the scene shown in Figure 5.9: raw image data from the right camera, candidate stripe pixels, and reconstructed depth map.

• MPEG-1 robust light stripe scanning (2.15 MB)