- UCMs in Pistol Shooting
- The Task
- Mapping from 7-dof input to 2-dof output
- {Show analagous setup in 2-d}
- Mapping is known a priori
- {Iso-therm diagram}
- For each successful trial, take joint angle snapshot at trigger-time.
- cluster lies along iso-threm {diagram}
- That’s not so unexpected. Choosing successful trials means the gun angles are guarnateed to be near zero.
- Instead, take joint-angle snapshot 50% of the way through the motion.
- Resulting gun-angles could be anything. Yet they tend to lie in clusters running parallel to the iso-therms.
- {diagram}
- This remains true throughout entire trajectory.
- Expert Envelope in Team-of-Teams Task
- Room clearing task with 4 people.
- Recorded data for each trial is a 64-d signal:
- team-size x (position + head-angle + weapon-angle) x velocities
- 4 x (2 + 3 + 3) x 2 = 64
- Hypothesis: Expert behavior is restricted to a low-dimensional subspace.
- How to resolve UCM and Expert Envelope ideas?
- Pistol shooting and Team trajectories are different kinds of data.