Mini challenge Stairs The following is a thread of discussion from the SRS listserver about adding a new competition to the SRS Robothon; a stair challenge for autonomous robots. The discussion end up talk about whegs.

From:  "Pete Miles"
Date:  Mon Mar 8, 2004  3:36 pm
Subject:  Mini challenge Stairs
            
So, does anyone have any good suggestions as to how to sense stairs as opposed to it being an obstacle to avoid?
From:  "David Hunt"
Date:  Mon Mar 8, 2004  3:38 pm
Subject:  RE: [SeattleRobotics]
   Mini challenge Stairs
            
From the bottom, perhaps some sort of tiered infrared sensor array. It could calculate the distance form an obstacle at several different heights. If it found the right pattern, then stairs. Have the sensor for the second step throw an interrupt to the processor to analyze whether it is stairs, or just an odd object.
From: Tom Capon 
Date: Mon Mar 8, 2004 5:50 pm
Subject: RE: [SeattleRobotics]
   Mini challenge Stairs
            
If you have a vertical scanning sonar sensor or something, you could map the profile of the object. If it looks like a couple of steps, then it would be steps, if it looked all squiggly, or not very high, then you would know it was not steps.
From: Larry Gieb
Date: Mon Mar 8, 2004 9:04 pm
Subject: Re: [SeattleRobotics]
   Mini challenge Stairs
            
You could try measuring the obstacle with rangefinders. and compare to known stair parameters

Commercial and residential codes vary- residential codes allow steeper stairs, but by most codes, stairs will be between 6 and 8 inches in riser height, and greater than 9.5 inches for tread width (a 1" nosing is required for tread width under 11 inches). Some local variants of commercial code demand a 12 inch tread minimum. By definition, the changes will be fairly abrupt, so that would be another clue.

8 inches is now a fairly universal limit on riser height, though some older codes allowed 8 1/4 " maximum, so an abrupt change in distance at less than 8 1/4 inches intervals is another clue

As a matter of practice for ergonomics, if you add 1 riser and 1 tread width, the number will be within a couple of inches of 17. Two risers and one tread width should be near 25. This generally leads to stairs that slope at numbers between 32 -39 degrees. you can be fairly certain not to encounter stairs steeper than 40 degrees. You folks in Seattle could measure stairs at the science center. if you do, publish your findings for the rest of us.

An exception is spiral stairs. these usually have a riser height of around 9 1/2 inches to provide enought headroom as the stairs wind overhead. Building a robot to climb spiral stairs would be a real challenge! Detecting a down flight of stairs will be harder, but a downward trending slope of 32-39 degrees should be a clue.It should be easier to run/fall down stairs

Of course, there are always exceptions to these measurements. Since I have spent over 30 years in the building industry, I tend to notice when I see the exceptions in commercial buildings. Some of the worst errors I found was building that held one of the county code enforcement departments near me. In fairness to the code people, that building was torn down. I think it was an embarrassment to them

just building a tracked or whegged robot that can negotiate a 40 degree slope might be all you need it should probably be long enough to span three stair nosings.

From:  "Kenneth Maxon"
Date:  Mon Mar 8, 2004  9:14 pm
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
A laser ranger finder would be ideal for detecting stairs. Applying a structured light algorithm to a vertical plane would identify them in a single pass, no scanning or other.

A bit harder of a problem would be to pull the repetitive but predictable pattern out of a video image.

From:   Mr. S.
Date:  Mon Mar 8, 2004  9:35 pm
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
I am thinking of experimenting with modulated IR arrays with vertical differential. That is to say that the array of transmitters are arranged vertically, and they are modulated so that the higher emitter will be modulated differently than the lower emitter... this should give an averaged reading that indicates there is a change in distance when facing a wall that is no higher than 8 or so inches. Thanks for the code info. The modulation allows detection of which transmitter is emitting the IR, thus some ability to determine a difference according to height of the transmitter. The conical radiation pattern of IR transmitters gives some value at given distances, and the radiation pattern has to be accounted for in the calculations of what is being "seen"

This thought says that the closer the robot is to the stairs, the more differential there should be. So from some feet away, there is little difference. At point blank is the maximum differential in readings. The problem of being too close for the sensors to work efficiently is encountered when actually climbing the stairs, so a set of sensor arrays would be needed, or a _LOT_ of good programming.

At some point, it would be good for the robot to know what the depth of the stair step is, even if covered in carpet. A physical arrangement could measure the first step, and this might work well enough.

Although its not a stair step, this is the approach I'm experimenting with for curbs. It is most definitely an intersting problem :)

Structured light solves both the stairs going up, and the stairs going down detection problem.

Interestingly, I've come across the thought problem of how to stop the robot from climbing everything, such as fireplaces, people, dogs, a desktop computer on the floor... you name it, there are lots of things you don't want the robot to climb on top of.

I'll let you know what my experiments work out to.

From:  "Kenneth Maxon"
Date:  Tue Mar 9, 2004  2:05 am
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
We all spend a lot of time working indoors here in the Northwest. I think a lot of us would really get a kick out of finding the day of the competition soaked in sunshine to really throw those sensors for a loop. It might be worth while taking a few of the sensors you are planning to use outdoors on a bright (even if overcast) day and experimenting with the changes to their readings if they are still usable at all.
From: Mr. S.
Date: Tue Mar 9, 2004 5:39 am
Subject: Re: [SeattleRobotics]
   Mini challenge Stairs
            
:) Yes, natural light sources are indeed the thing to beat. The modulation is hopefully going to allow the recievers to overcome that prickly problem too. Not to mention the number of man made modulated IR signals in the air around us, especially indoors. My laptop and pda have been known to try talking to each other from across the room, well out of usable range, close to 12 feet. This tells me it is possible for modulation to increase the usability of the IR signal.

Will this actually work? Just have to experiment to see.

If anyone has specific information that they feel a need to share, it is welcomed. :)

From:  F. Thompson
Date:  Tue Mar 9, 2004  5:55 am
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
What is a whegged robot?
From:  Larry Geib 
Date:  Tue Mar 9, 2004  6:44 am
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
The October or November Nuts & Volts had an article on robots with wheels made up of legs arranged in a rotary pattern.

You can get the idea at: http://biorobots.cwru.edu/projects/whegs/whegs.html

From:  Michael Jones
Date:  Tue Mar 9, 2004  8:09 am
Subject:  Re: [SeattleRobotics]
   Mini challenge Stairs
            
Think wheel studs on massive steroids.
From: Alan King
Date: Tue Mar 9, 2004 2:12 pm
Subject: Re: [SeattleRobotics]
   Mini challenge Stairs
            
Also they've been mentioned several times on this and the legged robots lists in the past. Some other ideas for those who might be interested in these from my previous thinking on whegs.. http://home.nc.rr.com/alan69/whegs/whegs.gif

Note that these aren't drawn accurately, just to give the general ideas. I have better drawings but can't locate them in my 120 gigs of junk at the moment so these will do..

First, is a hollow ring wheel with 3 holes for the legs to go through. Out from the axle and up for some displacement is the hub for the 3 legs. (Note the holes should really be 120 deg apart, the upper legs would be more out..) This arrangement will give variable leg speed, the leg on bottom spending more time supporting and the top legs going over faster.

The second is a cam. The legs can all move in easily, only light spring tension holding them out. Still lock and climb over, but when supporting there is a cam that a peg on the leg rides against, that holds the leg out. Note the cam instead of being straight as drawn, should really curve in towards the center to change the leg length to compensate for the angle as it rolls. This would allow the foot path to be straight, and the gait would be smooth.

Last is a guide plate instead of a single cam. Still should be curved, but gives the idea. Depending on what the leg is on, different ride heights are traveled through the guide plate. So if one leg lands on a rock, it could ride over it almost as though it weren't there and still give some support. Has to have progressive tension for the leg extension springs, you wouldn't want it to simply start riding along with all legs at a lower height. Slightly higer spring force on the front legs should cause it to climb out of this condition. Done well you may be able to litter your driveway with rocks, and still have a bot move over them as though they weren't there.

End goal is an infinitely deformable wheel that gives support underneath and still catches a forward edge to climb. Think 12 legs on a wheel. With rigid legs the forward ones will keep it from climbing. But if they move in and only the ones hitting the edge of the stair and experiencing a rotational force lock out, then it'll still work fine. Needs a progressive change over in required force to lock, hitting the vertical of a step will still give some rotational force to the feet above the centerline, even though they aren't over the edge and shouldn't lock out.

With enough legs per wheel, you'd have feet at the front and back of each support area. With the flat support area that means a two whegged bot would be stable, so the extra parts per wheg may pay off in a bot with only two wheel drive.

There are also some easy mechanical arrangements that let you use the wheels with legs retracted for speed, and then extend them when needed for climbing or rougher terrain. Overall I think the whegs could be worked into something extremely versatile with a little experimentation.