SLIDE 1
CS 730/830: Intro AI
Control
EOLQs
Control
CS 730/830: Intro AI 1 handout: slides Control Wheeler Ruml (UNH) - - PowerPoint PPT Presentation
CS 730/830: Intro AI 1 handout: slides Control Wheeler Ruml (UNH) - - PowerPoint PPT Presentation
CS 730/830: Intro AI 1 handout: slides Control Wheeler Ruml (UNH) Lecture 6, CS 730 1 / 12 EOLQs Control Wheeler Ruml (UNH) Lecture 6, CS 730 2 / 12 Control Problems MPC Break P Control PD Control PID Control
SLIDE 2
SLIDE 3
Control
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 3 / 12
SLIDE 4
Planning Problems
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 4 / 12
Observability: complete, partial, hidden State: discrete, continuous Actions: deterministic, stochastic, discrete, continuous Nature: static, deterministic, stochastic Interaction:
- ne decision, sequential
Time: static/off-line, on-line, discrete, continuous Percepts: discrete, continuous, uncertain Others: solo, cooperative, competitive
SLIDE 5
Model Predictive Control
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 5 / 12
used with ‘receeding horizon’ (≈ real-time search) simulate a bunch of controls (near nominal), pick best!
- r steer to a bunch of states (near nominal), pick best!
flexible, dangerous
SLIDE 6
Break
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 6 / 12
■
asst3
■
projects
■
wildcard class
SLIDE 7
P Control
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 7 / 12
u = KP (xr − ˆ x) responsiveness vs smoothness = spring model unstable with inertia!
SLIDE 8
PD Control
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 8 / 12
u = KP (xr − ˆ x) + KD d(xr − ˆ x) dt dampen correction if error is changing a lot = dampened spring model! does nothing if persistent error balances P component
SLIDE 9
PID Control
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 9 / 12
u = KP (xr − ˆ x) + KI
- (xr − ˆ
x)dt + KD d(xr − ˆ x) dt removes any persistent error however, ‘wind-up’ widely used. not optimal or necessarily stable. tune by hand, or Thrun says coordinate-wise bisection search
SLIDE 10
Bisection Search
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 10 / 12
given f and initial guesses l and r 1. bracket a local minimum (a) try guess m in middle (b) if m smallest, done! (local min between l and r) (c) if l smallest, r ← m, m ← l and move l left move l by at least original r − l (double interval) (d) if r smallest, m ← r and move r right 2. refine estimate (a) try lm between l and m. (b) if smaller than m, r ← m and m ← lm (c)
- therwise, try mr between m and r.
(d) if smaller than m, l ← m and m ← mr (e)
- therwise m is smallest, l ← lm and r ← mr
(f) until range small or values close
SLIDE 11
See Also
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs
Wheeler Ruml (UNH) Lecture 6, CS 730 – 11 / 12
- ptimal control: eg, Linear-Quadratic-Gaussian (LQG)
discrete control: eg, Markov decision processes state estimation aka filtering: eg, Kalman filter, particle filter
SLIDE 12
EOLQs
Control ■ Problems ■ MPC ■ Break ■ P Control ■ PD Control ■ PID Control ■ Bisection Search ■ See Also ■ EOLQs