Presentation is loading. Please wait.

Presentation is loading. Please wait.

Complexity of STRIPS SHAKEY the Robot October 8, 2003.

Similar presentations


Presentation on theme: "Complexity of STRIPS SHAKEY the Robot October 8, 2003."— Presentation transcript:

1 Complexity of STRIPS SHAKEY the Robot October 8, 2003

2 1.Backwards planning 2.Why planning is hard 3.SHAKEY the robot

3 Backwards Planning Node = set of states h’(n, i) = MIN{ h(i,n’) | n’ instantiates n }

4 Sussman’s Anomaly An example from the blocks world in which divide and conquer fails Interacting goals A B C goal A B C Initial state On (A,B) On (B,C)

5 Sussman’s anomaly 1.Assume we want to satisfy On (A,B) first A B C Initial state A B C A B C B C A But now we cannot satisfy On (B,C) without undoing On (A,B) 2.Assume we want to satisfy On (B,C) first. But now we cannot satisfy On(A, B) without undoing On(B, C)

6 Planning is PSPACE-Complete PSPACE = can be computed in polynomial space PSPACE-Complete = as hard as anything in PSPACE Proof: Reduce ANY (space bounded) Turing machine to a STRIPS planning problem

7 Proof Propositions: in(i,x)symbol x in tape position i at(i,q)head is at i in state q trans(q,x,y,q’)transition table Actions: action do(i,q,x,y,q’) precondition: in(i,x), at(i,q), trans(q,x,y,q’) effect: at(i+1,q’), in(i,y),  in(i,x),  at(i,q)

8 Is Planning Possible? SHAKEY the robot –Brains: STRIPS style planner –Stanford Research Institute, 1969 –Funded by DARPA (as was the Internet!) VIDEO

9 What Else Did Shakey Have? Sensing Error Recover –Regress goal backwards to latest state that matches “real world” Learning –“Macro operators” – MACROPS –Potential pitfall?

10 courtesy JPL Started: January 1996 Launch: October 15th, 1998 Experiment: May 17-21


Download ppt "Complexity of STRIPS SHAKEY the Robot October 8, 2003."

Similar presentations


Ads by Google