Endeavour Drive - Drivability analysis |
Endeavour Drive - Drivability analysis |
Sep 18 2008, 11:05 AM
Post
#1
|
|
Founder Group: Chairman Posts: 14434 Joined: 8-February 04 Member No.: 1 |
http://www.sciencefriday.com/program/archives/200809191
You can listen via NPR, or via one of the web feeds that are listed on the site, but make sure you do listen if you can. |
|
|
Oct 13 2008, 05:12 AM
Post
#2
|
||
Senior Member Group: Moderator Posts: 2785 Joined: 10-November 06 From: Pasadena, CA Member No.: 1345 |
Another example of an Aqua terrain. This time Oppy traversed the dunes in a perpendicular direction (with little problem).
-Mike -------------------- Some higher resolution images available at my photostream: http://www.flickr.com/photos/31678681@N07/
|
|
|
||
Oct 13 2008, 10:42 AM
Post
#3
|
|
Member Group: Members Posts: 236 Joined: 5-June 08 From: Udon Thani Member No.: 4185 |
Thanks Mike,
tremendous job you did, but it sure helps s lot to get a perspective of what we can expect looking at the terrain on the HiRISE images! I guess all together we have produced enough data with this analysis for a book ! Regards, Geert. |
|
|
Oct 13 2008, 04:22 PM
Post
#4
|
|
Senior Member Group: Members Posts: 1018 Joined: 29-November 05 From: Seattle, WA, USA Member No.: 590 |
I guess all together we have produced enough data with this analysis for a book ! I think you still need to make some predictions. There are now ample examples of terrain where rovers had no problems as well as terrain where they did have problems. That should be enough to define an algorithm to estimate risk for any particular drive direction. One could do that by hand, or one could train a neural net to do it. (Or perhaps an SVM might work better, if you can settle for answers like "Yes, drive this way" and "No, don't go that way.") Anyway, given an algorithm that gives hazard ratings to different directions, it should be a simple dynamic programming problem to pick an optimal path from any given starting point. Or has someone tried this already? --Greg |
|
|
Oct 13 2008, 09:42 PM
Post
#5
|
|
Member Group: Members Posts: 236 Joined: 5-June 08 From: Udon Thani Member No.: 4185 |
Anyway, given an algorithm that gives hazard ratings to different directions, it should be a simple dynamic programming problem to pick an optimal path from any given starting point. Or has someone tried this already? I'm working on that more or less, but it is not as easy as it sounds Presently I'm able to produce "optimal" routes given a certain starting point and a destination, more or less by letting the software analyze all possible routes and then selecting the most promising one. Selecting factor is then the standard deviation in brightness of all pixels on the selected route, assuming a higher stdev value will indicate a rougher route. The route shown in my earlier message is an example of such a computer calculated route. Problem is that there is far more to it then just the stdev value, and making the software smart enough to recognize the various traps takes a lot of work and to get a more reliable answer you will need a lot more data (preferably some accurate method to measure heights of ripples and type of surface material), I doubt whether you can ever extract that from only the HiRISE imagery. Presently I'm working on a method to incorporate the various other analyze-methods into this route-selection, however problem with this is that many values are 'multi-directional' and what I need is a directional-value. In other words, they show you 'this position is good/moderate/bad' but what I need is a value that indicates 'driving in this direction from this position is good/moderate/bad', there are many examples of situations in which you can easily drive in one direction but get problems if you drive into an other direction. Most of all however what I need is simply time to work on these types of algorithms, and presently I simply don't have enough free spare time |
|
|
Oct 13 2008, 10:29 PM
Post
#6
|
|
Senior Member Group: Members Posts: 1591 Joined: 14-October 05 From: Vermont Member No.: 530 |
Presently I'm working on a method to incorporate the various other analyze-methods into this route-selection, however problem with this is that many values are 'multi-directional' and what I need is a directional-value. In other words, they show you 'this position is good/moderate/bad' but what I need is a value that indicates 'driving in this direction from this position is good/moderate/bad', there are many examples of situations in which you can easily drive in one direction but get problems if you drive into an other direction. You need a function that given a position and a direction, returns a value. For any given position, you call the function for all directions and choose the best one. That function is going to be calling itself recursively, until you get to the position that neighbors your destination, at which point the function is going to return happily (zero), or until it hits a red spot, at which point it returns unhappily(1). Good routes would have low values-- all green equals zero. Between green and red is a floating-point value, but you needn't assign it linearly. How you determine what value to return is the fun part of experimenting, because the weighting given to green vs. blue vs step size is going to cause big differences in whether it goes out of its way to stay green versus plowing straight through blue. (ploughing strait threw blew-- gotta love engrish) So, you give it a starting point, a destination point, a step size, a function of assigning floating-point values to colors, a field of play and you get one most-optimal route. You then vary the step size and value function and use your eyeballs to choose the ones that look like they have the best tradeoff between distance and being conservative. I'm pretty sure that's dynamic programming, but I'm a bit hazy on the details. Prof. Cormen wouldn't be proud, but that class was eight years ago. (Edit: Ah yeah, reading wikipedia makes it clearer that dynamic programming would build a solution set to avoid recursion calculating the same information a gagillion times. So if the lazy try seems to lock your system up, I guess you start building a data structure, and have the function also return from points that are already in the data structure.) |
|
|
Lo-Fi Version | Time is now: 22nd September 2024 - 07:04 PM |
RULES AND GUIDELINES Please read the Forum Rules and Guidelines before posting. IMAGE COPYRIGHT |
OPINIONS AND MODERATION Opinions expressed on UnmannedSpaceflight.com are those of the individual posters and do not necessarily reflect the opinions of UnmannedSpaceflight.com or The Planetary Society. The all-volunteer UnmannedSpaceflight.com moderation team is wholly independent of The Planetary Society. The Planetary Society has no influence over decisions made by the UnmannedSpaceflight.com moderators. |
SUPPORT THE FORUM Unmannedspaceflight.com is funded by the Planetary Society. Please consider supporting our work and many other projects by donating to the Society or becoming a member. |