Showing Revision 1 created 11/28/2012 by Amara Bot.

1. Title:
   08-02 Problem Solving Vs Planning
2. Description:

   Unit 8 2 Problem Solving vs Planning

3. 0 0:00 - 0:03
   [Narrator] You remember our problem-solving work?
4. 3000 0:03 - 0:06
   We have a state space like this, and
5. 6000 0:06 - 0:09
   we're given a start space and
6. 9000 0:09 - 0:11
   a goal to reach,
7. 11000 0:11 - 0:13
   and then we'd search for a path
8. 13000 0:13 - 0:16
   to find that goal, and maybe we find
9. 16000 0:16 - 0:19
   this path.
10. 19000 0:19 - 0:21
    Now the way a problem-solving agent
11. 21000 0:21 - 0:24
    would work is first it does all the work
12. 24000 0:24 - 0:26
    to figure out the path to the goal
13. 26000 0:26 - 0:29
    just doing by thinking,
14. 29000 0:29 - 0:31
    and then it starts to execute that path
15. 31000 0:31 - 0:35
    to drive or walk, however you want to get there,
16. 35000 0:35 - 0:37
    from the start state to the end state,
17. 37000 0:37 - 0:39
    but think about what would happen
18. 39000 0:39 - 0:41
    if you did that in real life; if you did all
19. 41000 0:41 - 0:43
    your planning ahead of time, you had the complete goal,
20. 43000 0:43 - 0:46
    and then without interacting with the world,
21. 46000 0:46 - 0:48
    without sensing it at all,
22. 48000 0:48 - 0:50
    you started to execute that path.
23. 50000 0:50 - 0:53
    Well this has, in fact, been studied.
24. 53000 0:53 - 0:56
    People have gone out and
25. 56000 0:56 - 0:59
    blindfolded walkers, put them in a field
26. 59000 0:59 - 1:01
    and told them to walk in a straight line,
27. 61000 1:01 - 1:04
    and the results are not pretty.
28. 64000 1:04 - 1:07
    Here are the GPS tracks to prove it.
29. 67000 1:07 - 1:09
    So we take a hiker, we put him at a
30. 69000 1:09 - 1:11
    start location, say here,
31. 71000 1:11 - 1:13
    and we blindfold him so that he can't
32. 73000 1:13 - 1:15
    see anything in the horizon,
33. 75000 1:15 - 1:18
    but just has enough to see his or her feet
34. 78000 1:18 - 1:20
    so that they won't stumble over something,
35. 80000 1:20 - 1:23
    and tell them execute the plan of going forward.
36. 83000 1:23 - 1:26
    Put one foot in front of each other and walk forward in a straight line,
37. 86000 1:26 - 1:28
    and these are the typical paths we see.
38. 88000 1:28 - 1:30
    They start out going straight for awhile
39. 90000 1:30 - 1:32
    but then go in loop de loops
40. 92000 1:32 - 1:35
    and end up not at a straight path at all.
41. 95000 1:35 - 1:37
    These ones over here, starting in this location,
42. 97000 1:37 - 1:39
    are even more convoluted.
43. 99000 1:39 - 1:41
    They get going straight for a little bit
44. 101000 1:41 - 1:43
    and then go in very tight loops.
45. 103000 1:43 - 1:45
    So people are incapable of walking a straight line
46. 105000 1:45 - 1:48
    without any feedback from the environment.
47. 108000 1:48 - 1:51
    Now here on this yellow path, this one did much better,
48. 111000 1:51 - 1:53
    and why was that?
49. 113000 1:53 - 1:56
    Well it's because these paths were on overcast days,
50. 116000 1:56 - 1:59
    and so there was no input to make sense of.
51. 119000 1:59 - 2:02
    Whereas on this path was on a very sunny day,
52. 122000 2:02 - 2:04
    and so even though the hiker couldn't
53. 124000 2:04 - 2:07
    see farther than a few feet in front of him,
54. 127000 2:07 - 2:10
    he could see shadows and say,
55. 130000 2:10 - 2:12
    "As long as I keep the shadows pointing in the right direction then
56. 132000 2:12 - 2:15
    I can go in a relatively straight line."
57. 135000 2:15 - 2:18
    So the moral is we need some feedback from the environment.
58. 138000 2:18 - 2:21
    We can't just plan ahead and come up with a whole plan.
59. 141000 2:21 - 2:24
    We've got to interleave planning
60. 144000 2:24 -
    and executing.