Transcript Temporal Motion Control

Temporal Motion Control
CMPUT 610
Martin Jagersand
Administrative:


Schedule paper/project presentations
Lecture on practical Hand-Eye systems
–
–
–
Tracking demo
In robotics lab
Ok to have with robotics group meeting?
(Thu 12-1pm)
Temporal motion


How do robots move in films, TV, industry?
How is it different or inferior to humans?
Traditional temporal control:
1. None
2. Piecewise polynomial trajectory, derivative
boundary conditions
Cooperative robotics

Better, human like motion control needed if
robots are to assist humans
Human-like model

Minimize “jerk” (acceleration change)


Leads to polynomial trajectory

T = [0..1]
Double step:
Exact optim criterion not important

Goodman:

Integrate:

Double step:
Double step
Multistep

Pos:

Vel:
Model and world state

Where we/the robot believes it is is not
necessarily where it really is.

Need to add correction movements
Kinematic errors
Integrating feedback

Sees hand near goal (milner)

Attention focused on target
Feedback system


Fast internal feedback
Slower external trajectory corrections
Adding correcting movements
Multiple corrections
Minerva


Modular link arm
Robotic and human
performance
compared
Coordinate frames

X is in what frame?
Execution in muscle/joint space!
– Planning in world space? Retinal/image space?
(plan along straight lines, execute slightly curved traj)
–

Natural coordinate system?
–
–

Gravity, direction of movement, horizon
But Georgopoulos coding in polar coordinates
(direction and extent)
Coordinate transforms