|
[DPRG] DPRG LEVEL II Robo-Magellan Extensions
Subject: [DPRG] DPRG LEVEL II Robo-Magellan Extensions
From: Pete Miles
hoardsofminions at yahoo.com
Date: Sun Jan 8 17:09:58 CST 2006
If the final destination cone is not in the line of site of the robot at the start of the contest then how does a robot home in on a orange traffic cone? If a robot can't see the cone, but knows its coordinates relative to where it starts, again how would it find that cone? As far as I understand it, some sort of a navigation technique has to be employed until the robot acquires a line of site vision of the cone. If this line of site doesn't occur until the robot reaches within 40 feet of the cone, then how is this any different than using any other form of navigation (or just random miandering) to get to this general area?
The Robo-Magellan rules now state the robot must travel
A------------B-------------C
Points A and C can be the same point. The rules also say there may be additional intermediate points. If A and C are the same then the contest becomes how well the robot can return to the starting point.
Now, for arguments sake, lets assume there is no cone at the destination point C which is the same as point A (again no cone). Now if a robot must reach point B, isn't the entire accuracy question in each robot's GPS and the organizers equipment still present? Whose equipment was more accurate in defining point B. If the argument about navigating accuracy about reaching the end point which is the same as the start point being eliminated because they are the same point, but then this argument starts all over again in defining the intermediate waypoint B.
In a contest, the same method for quantifying each target point should be the same. Otherwise there will be arguments about one point being defined as being more or less accurate than other points, and different robots can respond differently to these errors.
By having a cone at each target point, and using time averaging (and/or Earth Google) to define the points, every robot will have the exact same set of location information.
One thing that is missing in all of these arguments is the actual path a robot has to make to go from one point to another. They are not straight line paths. There are lots of challenging obstacles between each target point. There are hills, trash cans, curbs, grass, asphalt, beauty bark, huge rocks, cars (that move), people, 3 year old kids that want to pick up the robot, buildings, benches, trees, bushes, wet leaves, sand, water, etc, and the worse - intermittent GPS reception.
The robot has to go around, above, over, under, and through all of these obsticals. Every single one of these will mess up the ability of a robot to be able to keep track of where it is relative to where it needs to go. If a robot looses GPS reception, then how does it know where it is? If the robot's inertial navigation system doesn't agree with its GPS information, then how does it know where it is? All of these problems occur from the time the contest starts to the time it ends. Regardless of whether a cone is at each target point or not, the robot will have a great challenge just to get to an approximate location of each target destination point. What is even worse, this contest will run regardless of what the weather is. Rain, Sun, Snow, Wind, Hail, etc. The robot needs to be able to handle any one of these environments and obsticals.
In order for a robot to succeed, it must implement multiple navigation methods get from point A to point B, and so one. No one single method will get a robot close to the final destination. Pure vision, pure dead reconing, and pure GPS, will not work. If a robot can solve the course using only one of these, then the course was poorly layed out.
The course should be a challenge to a remote control robot. It is funny though, most of the walk up spectators that watch the Robo-Magellan think it is a remote control contest and are puzzled why the big deal about it.
Now because there are no straight line paths between any of the target points, the robot has to choose/follow some course around the obsticals. The robot builder can let the robot autonomously choose its own path, of the robot builder can enter a path the robot must follow to get around all the obsticals. For any naviaging contest, it is a challenge to get a robot to follow any specific path, let alone reach any specific final target point.
When looking at this contest as a whole, navigation accuracy and negotiating the terrain, it is a very difficult contest.
We can argue all day long about navigation accuracy and the true measure of accuracy. But it is a severe challenge for the robot even to get to the point where it becomes an argument. If the course was properly laid out, no one would argue the fact that the robot autonomously navigated from the start point to the end point whether it touched a cone or stoped within a few inches of the destination point (with no cone). Just getting to this point required a lot of navigating skills.
Having to touch an orange cone is just part of the many challenges.
At this year's Robothon I got to see an amazing feat with a robot. This robot was running along an asphalt path. Then all of a sudden a toddler started running up to the robot. When the robot got to about 6 feet of the toddler, it slammed on its brakes and screached to a stop. The toddlers dad was able to grab the toddler before she grabbed the robot and pulled her to the side. Then the robot's ultrasonic sensor array looked to the left, then right, then straight ahead, and the robot started to go forward again. Now this was an impressive demonstration on how well the builders anticipated moving obsticles and safety.
When I was in the Marine Corps, our land navigation testing consisted of finding a set of numbers writting on the side of some trees located out in the swamps/forests at Camp Lejeune, NC. We were given a set of coordinates to each tree, a compass, and a map, and given 8 hours to find them. Success was measured on 1st, how fast we did it, and having all the number correct. We all had to use vision to servo in on a tree to read the numbers we had to get. But we all had to use the compass and counting paces to get us close to where the target was. If we miss read the compass, misscounted our paces, or the more common problem, not properly taking into account how our pace distance changed when climbing over trees, going up/down steep hills, or crawling under brush, we wouldn't even be close to use vision to servo in on the destination tree. Thus many times we had to go back to a new point, or refind where we were using the map and looking for common terrain features. This was
before GPS, and this is true navigation, and the measure of error was if we read the numbers on the tree correctly, a tree that we touched with our hands.
The Robo-Magellan contest if very similar to this. Yes it has to touch an orange cone, but the robot must employ multiple navigation techniques just to get close enough to see the cone inorder to touch it. A robot that is very accurate will be able to find and touch the cone. All other robots won't even come close to touching the cone.
Pete
---------------------------------
Yahoo! DSL Something to write home about. Just $16.99/mo. or less
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.dprg.org/pipermail/dprglist/attachments/20060108/177e2df1/attachment.html
More information about the DPRG mailing list
|
