Inertial navigation over a very small range

[PerryGT]

I'm interested in applying inertial navigation over a very small range < 2 meters. Everything I have seen so far that is related to using inertial navigation for dead reckoning is for very large scales and gives accuracy on the order of fractions of miles or several feet. Does anyone know what type of accuracy to expect with only very small movements? The system that I'm envisioning would only apply dead reckoning for a few minutes within a space of 2 meters and then be zeroed before the next application. Any information would be very helpful.

 

[DaveC426913]

I assume you're unable to use other means, such as IR range sensors?

 

[PerryGT]

Very good observation about IR. Infrared is used for the application I have in mind but I believe it will be more costly and involved than something based on accelerometers/gyros. My goal is to implement this in a developing country that has limited resources.

 

[mfb]

The accuracy depends on the quality of acceleration/orientation sensors you have, the way the object moves, and the timescales.

 

[PerryGT]

As an example, let's say I'm tracking the movement of a pencil as a I draw a simple sketch in less than 5 minutes. The pencil is initially zeroed at a location located 2 feet away. Are there sensors accurate enough to track the pencil on the order of millimeters?

 

[berkeman]

As an example, let's say I'm tracking the movement of a pencil as a I draw a simple sketch in less than 5 minutes. The pencil is initially zeroed at a location located 2 feet away. Are there sensors accurate enough to track the pencil on the order of millimeters?

There are certainly mechanical arrangements that can track the writing instrument very accurately. I've seen it done with two spring-loaded thin cables (with rotary encoders on the spring-loaded spools), and with a combination of a linear encoder (rod attached to the pen/stylus) and a rotary encoder (with the rod passing through it).

Also, there are several "graphics tables" techniques that can be used to accurately encode the position of the drawing stylus...

 

[PerryGT]

Thank you Berkeman for the response but I am just using the pencil as an example. I'm trying to get a feel for using inertial navigation over a small range. Other options such as encoders, IR, optical tracking have limitations for my application.

 

[DaveC426913]

There are certainly mechanical arrangements that can track the writing instrument very accurately. I've seen it done with two spring-loaded thin cables (with rotary encoders on the spring-loaded spools), and with a combination of a linear encoder (rod attached to the pen/stylus) and a rotary encoder (with the rod passing through it).

Also, there are several "graphics tables" techniques that can be used to accurately encode the position of the drawing stylus...

If you read between the lines of the OP's posts 1 and 3, it seems that he's looking for something that is both inexpensive as well as inobtrusive. Somebody trying to manipulate a pen might be a little put off (if not completely obstructed) by having attached cables and an armature for recording.

 

[berkeman]

Does the application have wheels? If so, rotary encoders will work for position/movement sensing.

Can an apparatus be placed over the 2m X 2m area to help with the tracking?

 

[berkeman]

If you read between the lines of the OP's posts 1 and 3, it seems that he's looking for something that is both inexpensive as well as inobtrusive. Somebody trying to manipulate a pen might be a little put off (if not completely obstructed) by having attached cables and an armature for recording.

True enough, which is why I mentioned the graphics tablet paradigm. :-)

BTW, good to see you back at the PF!

 

[DaveC426913]

Thanks. Good to be back.Perry, my phone has a motion-sensitive ruler app. I touch my phone to a point, then move it to another point and it's supposed to be able to tell me the distance. In theory that means your idea is sound, though I can't say the app is much use.

 

[PerryGT]

The device is something that will attach to a patient, likely through some type of bite block. The goal is to orient and position the patient's head to a specific location within the room. So no wheels ;)

 

[PerryGT]

Yes thanks Dave. That is exactly the type of application I am considering. I just don't know how accurate these types of sensors can be. For patient setup we need to be on the order say 3 mm or less. Once setup the patient will remain fixed and we no longer need the inertial navigation.

 

[DaveC426913]

Yes thanks Dave. That is exactly the type of application I am considering. I just don't know how accurate these types of sensors can be. For patient setup we need to be on the order say 3 mm or less. Once setup the patient will remain fixed and we no longer need the inertial navigation.

Using multiple redundant sensors would narrow down the margin of error.

BTW, what are the consequences of getting it wrong? Is there human confirmation before firing up the 1.4GW laser?But I still say you're better off with IR sensors. If they're cheap enough to go in a Wii, then you should be able to use them for your purposes.

And rather than range-finding techniques, you could use angle. Have you ever seen the "image" that the Wii bar sees? It tracks the controller movement in a 2D plane.

 

[etudiant]

The needed performance is easily achieved using differential GPS, basically a set of local emitters that are referenced to the overall GPS signal which then monitor the difference in time of the GPS signal from the local emitters. That setup allows millimeter accuracy in 3 dimensions over a large space, city block sized, with no wires, just small self contained sensor plus battery and transceiver units.
The GPS sensors needed do not need to be very large or powerful, as they will only interact with the fixed local emitters.
The method is in use in the aerospace industry, but I don't know of an example where it is in hospital use.

 

[PerryGT]

Thank you all. I will investigate the Wii sensor and also the GPS solution. As noted by Dave, I am looking for an inexpensive and unobtrusive solution. In the USA we us a variety of methods to align patients including optical tracking, IR tracking, stereotactic localization, and electromagnetic beacons. The gold standard is x-ray imaging which confirms proper alignment and is used widely. In developing countries most of these solutions are unavailable and alignment is based on external marks which are aligned to fixed external lasers using a 3 point system. This system is not always robust and provides very little feedback to help align the patient correctly. My purpose is to develop something to aid the laser alignment and improve accuracy.

 

[berkeman]

Nobel work, ,Perry! :-)

 

[rbelli1]

See: http://hackaday.io/post/8803

This project looks like it is seeking to do inertial tracking at the level you need. It seems to be a very rough work in progress though.

With the 9 DOF sensor there is absolute tracking to the Earth's magnetic field too.

BoB