Documentation : Hardware | Software | Tutorials    Wardrobe : Jackets | Applications
Papers : Notes | Proposals | Research    Administration : Stock | Minutes | Progress

Next Generation Ultrasonic Positioning

This page outlines a few research ideas/goals for the next generation of ultrasonic positioning, as discussed by Paul, Henk, and Mike at the end of October, 2002.

Goals

We would like our indoor positioning system to have the following properties:

Cliff's current system has achieved the first two, cheap and simple, but reliability and scalability are still an issue.

Obstacles

Attaining reliability means that we must overcome two main obstacles, interference and reflections.

Interference

The current ultrasonic system that we employ uses narrow-band transmitters and receivers operating in the 40kHz band. The receivers use a simple level sensitive, one bit protocol for recognising ultrasonic pulses from the transmitters. As a result, any noise in the 40kHz band with large enough amplitude easily confuses the system. This noise is produced by various things in the environment, like rustling crisp packets or jingling keys.

A reliable positioning system must be able to differentiate noise and interference from authentic ultrasonic pulses. Some ideas for tackling this problem are discussed in the Proposals section below.

Echoes and Reflections

Each transmitter is allocated a time slice in which it must transmit an ultrasonic pulse. The receiver listens for the first pulse within this slice and assumes that it is from the designated transmitter. However, if there are reflections in the environment from the previous time slice, then the system will record an invalid distance reading.

The current solution to this problem is to make the time slices large enough to let all reflections dissipate before the next transmitter is activated. The draw back to this is that large time slices result in slower response times and large latencies for the system. Latency between distance readings also causes errors in position calculations since they assume that a set of distances correlate to the same point in time and space. Of course, this is not the case for objects in motion where the position changes as time passes. It is also important to note that this solution degrades the scalability of the system. Larger rooms, for example, require more transmitters - each increasing the latency further.

Alternative solutions to this problem is discussed in the next section.

Proposals

  1. It may be possible to reduce the effects of noise by changing the ultrasonic pulse from its one bit protocol to a several bit protocol. Giving bit patterns to each transmitter may also allow us to increase the frequency of the system since the receiver could differentiate between pulses from different transmitters. This may be a cheap way to improve reliability.
  2. Broadband ultrasonic systems get around noise and reflection problems by utilising a larger portion of the available ultrasonic spectrum. See Mike Haza's paper in the proceedings of UbiCom 2002. This is heavy duty stuff and is expensive in terms of computing power and component cost.
  3. We have talked about using multiple microphones on a receiver to measure the angle of incidence of incoming pulses. With this information we may be able to eliminate reflected signals or noise that originate from invalid positions (we know where the transmitters are supposed to be). This can be done in two ways:
    1. by placing the mics very close together (within one wavelength ~ 10mm) and measuring the pulse phase difference between them. Three microphones will give the angle at which the pulse is incident.
    2. by placing the mics slightly further apart and measuring the time difference between receipt of the same pulse. Instead of phase we measure time to achieve the same result.
    Of course, we have to be certain of the orientation of the receiver.
  4. It is also possible to eliminate erroneous readings by using the spatial relationship between the distances given by the system. The relationship gives a set of valid distances for each transmitter - given the distance to other transmitters. See Henk's 3D hull diagram.
  5. Kalman filters may provide a nice way to model the state of the system and allow us handle data in a stochastic manner - taking care of those nasty outliers. Henk has started on this one.
  6. Is there a way to use signal strength to gain information?
  7. Is is possible to get rid of the synchronising RF pinger?
  8. Cliff and Amoss have begun work on examining probability densities of pre-recorded location plots. I think that they plan to use this data to build a probability model for locating mobile users. Can you guys add more to this?
  9. I believe that Cliff plans to do some reading over the next couple of weeks on sensor fusion. Anything to add here Cliff?

It will probably be possible to combine a number of these techniques to improve on the reliability of the ultrasonic system.

Existing Systems and Reading

Below is a list of some of the current papers and projects relating to Ultrasonic Positioning.


This page last updated November 4, 2002
The material displayed is provided 'as is'. Contents may not be reused without prior permission.
For problems or questions regarding this web contact Cliff Randell.