6

I would like to have a better understanding of work in the field of "Navigation Among Movable Obstacles". I started off with Michael Stilman's thesis under James Kuffner, but that has not yet sated my appetite.

I am currently trying to simulate a scenario where debris (Tables and Table parts) from a disaster scenario block pathways. The debris forms part of a movable obstacle. The robot which will be used is a bipedal humanoid.

The thesis describes an approach to define the search space of possible actions leading from the start point to the goal. However, it assumes a mobile robot which works via gliding.

I think the state space definitions would change for a bi-pedal robot. Why is why I wonder what other work is being done in this field. Perhaps the work of other research groups could give me clues as to how to design and perhaps reduce the search space for a bipedal humanoid robot.

An implementation of Navigation among Movable Obstacles would also aid me in understanding how to reduce the search space of possible actions.

So does anyone know of a working implementation of Navigation among movable obstacles?

Any supporting information about other professors or research groups working on similar problems would also be very useful.

I hope this edit is sufficient for the problem description.

ThomasH
  • 773
  • 2
  • 8
  • 25
Naresh
  • 589
  • 1
  • 5
  • 13
  • This question seems to be rather broad and not too clear. Could you elaborate? – Manishearth Nov 20 '12 at 06:51
  • 1
    Link to thesis added for reference – Andrew Nov 20 '12 at 07:17
  • Hi @Naresh, what are you wanting to achieve? At the moment, it's difficult to know where to start, other than regurgitate a PhD thesis :) – Andrew Nov 20 '12 at 07:18
  • Hi Naresh, welcome to Robotics. Whilst there is the makings of a good question here (possibly two 8') it needs a narrower scope and more details of what the problem you are trying to solve is. Could I suggest that you re-read the FAQ before asking your next question, in particular "You should only ask practical, answerable questions based on actual problems that you face. Chatty, open-ended questions diminish the usefulness of our site and push other questions off the front page". Thanks. – Mark Booth Nov 20 '12 at 09:12
  • 1
    Naresh, Thanks for the additional information, it helps a great deal. I've tried to tidy it up a little and hope that this means we can get some of those down votes reversed for you. – Mark Booth Nov 20 '12 at 23:47
  • Are some of the obstacles small enough for the robot to just walk over them, if he was bipedal? If not, i'm not sure that the robot's mode of locomotion matters to the implementation. – Daniel Eberts Nov 21 '12 at 09:17
  • Possibly. Some of the other reasons about a bi-pedal's state space may matter is unlike a mobile robot, stopping the bi-pedal robot at an exact space from the obstacle might be dependent on the obstacle's height.

    Linear motion can no longer be approximated as well.

    Uneven territory cannot be traversed by a mobile robot. Bipedal can traverse in uneven territory upto a certain degree.

    The modes of grasping a fewer due to obvious balance issues.

    A lot of little things that could change the search space for bipedal robots in my opinion.

     – Naresh Nov 21 '12 at 09:34
  • Are the obstacles things that would interrupt the movement of the legs or things that would be unsafe to step on? I ask because an alternate way to handle this would be to sense when the leg bumps something, and return to the last stable state. Then, just re-plan the next step in light of the new information on the obstacle. – Ian Nov 21 '12 at 13:17
  • Obstacles in context of this problem are defined as 'things' which can be moved by the robot and are sufficiently large enough to typically block a path(or a doorway).

    So I doubt that 'something small enough to interrupt foot movement would be counted as one.' That would just be worked through via re-balancing and searching for a better path in a 'non-flat' environment.

     – Naresh Nov 21 '12 at 15:40
  • Perhaps the bipedal robot is able to use one of its legs to move obstacles away or "kick" them away. Implementing such an ability would be interesting in terms of control theory and trajectory planning for the legs as well as in terms of perception: How do I recognise obstacles that can be moved or "kicked"? – Daniel Eberts Nov 21 '12 at 17:34

1 Answers1

2

Speaking about navigation among moving obstacles, look at Jur van den Berg thesis. Also Lavalle provided a tutorial at ICRA12 http://msl.cs.uiuc.edu/~lavalle/icra12/ and the videos are here http://techtalks.tv/events/105/ but neither mention the problem of bipeds.

amine23
  • 154
  • 1