Let's Make Robots!

Robot Size and Methods of Obstacle Avoidance

I realize that most LMRians are going to have more experience with smaller robots and that I am in the minority here, but for the few that have similar or greater experience with big boys (or relevant experiences and insights from or in related fields) I wondered if I could see if you've noticed the same trends as I have in terms of larger (greater than 25cm in any dimension) Robs and effectiveness of various kinds of near-surface detection.

The conventional wisdom is that environment should decide what is most effective-if you have more black surfaces then use US, if you have more soft surfaces use IR, but if other factors come into play (say motors that have to be large enough to move such a platform that will always cause a drop out no matter how big of a snubber you put across a Sharp) what do you make of that wisdom?  Then there's analog v. digital (Maxbotix v. HCSR04 for example) across either method, etc.

Also, has anyone had any success implementing the use of an accellerometer as a crash sensor on a larger, slower platform?  The MMA on a MIcro Magician works wonders on a Tadpole sized machine but I can't get it to scale up to anything as large as Serv-O or aRDui5x.  Both still need physical connection bump switches for reliable impact detection.

Luckily I'm not talking about anything as massive as a wheel-chair motor machine.  Nothing I've built could knock a three-year-old boy over, so there isn't a lot of danger in the scale I'm considering here.

As I ponder it, I realize that there's no really perfect solution.  Even if you could give a human-sized robot visual, tactile and sonic sensors on a par with what we have evolved as a species it wouldn't be enough to prevent every mishap.  It certainly isn't enough to keep me from bumping into the chair every morning on the way out of the bedroom and knocking over the occasional Ming vase.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

I realize it's relatively expensive, but using a Kinect works great, and with ROS it can be pretty easy to get working. But it doesn't work in sunlight so it's not perfect.

In my (limited) experiance, I have found that a Maxbotics EZ0 does wonders on the front of a larger 'bot. I have a reasonably large outdoor bot that I am working on right now (http://boomsandbots.blogspot.ca/2013/10/a2-some-pictures-and-videos.html) and my EZ0 has been wonderful at picking up all sorts of objects. It will pick up a wall even if the bot doesnt approach it head-on (I have had it see walls at darn-near 45 degrees off of perpenducular) and even picked up a ~3/8ths inch sapling growing in my yard in time to avoid it. My main problem has not been with it seeing things, but with getting the robot to stop in time to avoid hitting them.

Go for an EZ0, they are expensive, but awesome.

Size does matter for some things. For example, you mentioned the Micro Magicians impact detection. As the robot gets bigger and has more momentum it's rate of deceleration may decrease (it's harder to stop) so the algorithm for detecting impacts may need to be changed.

DAGU now has a big robot (about 5ft tall) that is controlled by a T'REX controller with the same impact detection system as the Micro Magician. When I get time I will test it to see if your correct or simply not configuring the parameters correctly.

For other things such as the lamp you show above. Although a sensor near the ground should detect a big base quite easily the small diameter pole is much harder to detect no matter if you use IR or Ultrasound so you might need to use multiple sensors at different heights.

One sensor I have read about but never gotten around to trying uses an RF field to detect objects in a similar manner to a Theremin but unfortunately these sensors, just like IR and US sensors will still detect some materials better than others and cannot measure distance accurately.

If you want a perfect sensor then you will need to invent it.

That's the solution CtC offered when I started coming up against this and if I had the funds to start over on the particular project That got me thinking about it this week I'd probably go that route.

I remember the a post about a Dagu big bot-a mobile one for factories and industrial usage, right? It really intrigued me at the time but it seemed like maybe you stopped updating it on LMR because of the same fact that I initially mentioned (namely that we tend to make desktop bots more here.) if it makes any difference to your posting schedule, I still want to know more about it!

Here is a video of a modified version of my Big Bot robot base with a fibre glass body on top.

http://www.youtube.com/watch?v=TOqqhtjM-9M

Someone else has been in charge of this project so I haven't had a chance to play with it very much. If I was going to make something like this autonomous then it would definitely need US sensours around the base, as low to the ground as possible for detecting peoples feet as well as walls and other objects. It would then probably need another set waist height.

As this robot is intended for public areas I would not bother with IR sensors as sunlight could prevent them working properly.

Max, I am really not getting it :-)

Do you want to have a reliable solution to detect a leg of an chair, say a small and slim object?

That's not the only thing I'm asking but it plays a part. For example, if you have one of these in the environment:
1bf75870-c4d0-4399-bc61-b4e5b19352bd_300.jpg
it's going to look a lot different to a Hero Jr. Vs. a Micro Magician. However, back when robots were as big as Heros, they all seemed to go with US (Heathkits, Androbots, RB5X and Hubots all reached the same conclusion.)

I think I got it. In my opinion US is the better choice since the nature shows us how precise it is. Dolphins and bats are using it to navigate and find food. IR will have a  certain noice vs US. US is also created by nature but i guess far less than US.