Northstar Evaluation
Last Modified: 2006-08-15
find:

basket

Acroname Robotics PDF webpage version Northstar Evaluation PDF

Related
Products

Product image for NorthStar Developer Bundle
NorthStar Developer Bundle
Product image for NorthStar Detector Kit
NorthStar Detector Kit
Product image for NorthStar Projector Kit
NorthStar Projector Kit

Contents

Expectations

This is a new product type for us.  We had very few expectations to start with beyond being able to determine position and orientation within a room with good accuracy.  We vaguely understood the notion that there was an emmitter device and a sensor that identified a unique pattern of IR light on the roof to gain orientation and pose information.  Beyond that, we expected:

  • Position within centimeters
  • Failure in our warehouse ceiling
  • Orientation information
  • Interface Simplicity
  • Easy Experience out of the Box
  • Simple serial protocol to access all data
  • Roughly 10 fps (frames-per-second)

Some of these expectations were random, without any reading of the documentation or specifications.  This is just what we expected.  We aso wanted to learn some things we were curious about:

  • How vertical displacement affects performance?
  • How pitch and roll affect performance?
  • How different roof structures change performance?
  • Effects of lighting?
  • Interference from other IR in room?
  • Dropout or false readings?
  • Stepping in ceiling height?

Out of the Box

This is a higher cost sensor product and for that, you get very nice packaging and presentation.  The two main components (projector and detector) come "individually wrapped" and everything is included to get things up and running quickly for evaluation.  All documentation is included in printed form and there is a CD-ROM for the software. 

Image of both the projector and detector units come professionally packaged in protective cases.
Both the projector and detector units come professionally packaged in protective cases.

The first part to get set up is the projector which is made from two identical IR light projecting boards.  These are little more than a PIC chip that modulates the LED signal and the lenses to focus the light.  These need to be arranged on a small frame to get the correct spacing on the ceiling.  A drawing is inclued for the basic table top setup. 

Image of setting up the angle for the projector.
Setting up the angle for the projector.

One thing that would have been nice would be a table of angles and measurements for common scenarios (floor, table top, etc.) Once the angle setting is dialed in, the frame can be tightened. 

Image of tightening the set screw to lock in the projetor angle.
Tightening the set screw to lock in the projetor angle.

The detector portion consists of a small camera in an IR shield with and absurd connector to a daughter board that offers RS-232 level shifting, power, and some minimal configuration.  This tiny connector is the sort of thing that can doom a product.  It is likely that Evolution found a high-volume customer that specified the very thin and fragile ribbon cable.  This has been maintained for a version for robotics that is tough to work with.  It also nearly requires you to build a board with that connector or use the board they provide which is bigger than the actual detector. 

Image of us fighting with the mini connector.
Fighting with the connector.

A better approach to this would have been a standardized serial connector like that specified in the Acroname Serial Interface Standard.  This would allow users to select between USB and standard Serial ports and the connectors are easy to work with.  This is not dooming, but for the money, we wished it were easier and cleaner. 

Image of the assembled detector assembly.
The assembled detector assembly.

Wall wart power supply is used for both (provided in the kit) the detector and projector.  When plugged in, the projector can be tested with a provided nifty little IR card.  This we found tough to use without first turning off the lights in the room. 

Image of the IR detection card held over the projector's LED lense.
The IR detection card over the projector's LED lense.

Finally, there was some software.  We are cross-platform snobs so it was a bit disapointing to see only a Windows version of the software.  MacOS X and Linux would have been nice, depending on your religion.  Otherwise, the provided software install was simple and fast. 

Test Rig

So, with everything set up, we placed the projector on a table top and the detector nearby to get started.  We then moved both around to various locations to determine accuracy and reliability.  We also used a measuring tape from known reference points to establish some idea of accuracy. 

We tested this device place on a Garcia robot and shot various IR inputs at it to see how it was at signal rejection, etc. 

Garcia's IR controller being used to hammer the detector.
Garcia's IR controller being used to hammer the detector.

Performance

In our test, the sensor is very impressive in an office environment.  With consistent ceilings and some field of view (from the perspective of the robot), this sensor performs very well.  We particularily liked the display of the signal strength in the Windows UI as we moved the sensor with respect to the projectors.  It helped give an intuitive sense of how the orientation and field-of-view affect the performance.  If you were saturating an area with projectors, this would be an invaluable tool. 

Our expectations with regard to our "pathological case" warehouse ceiling bore out.  The roof pretty much confused the detector completely.  We got occasional readings but they were not correct.  This warehouse roof is beyond the capabilities of this localization system.  It may be possible with careful placement to get this worst-case scenario to give meaningful results, but we only had on projector to test with.  All our other tests worked very well with the exception of this one test. 

Image of our "pathalogical case" warehouse ceiling
Our "pathalogical case" warehouse ceiling

Conclusion

This is a fantastic sensor.  We found it to be very reliable and accurate in our office environment and other testing areas where the ceiling is somewhat consistent.  Our pathological case of the warehouse area was tough for the sensor, but that makes absolutely good sense when you consider how the sensor actually works. 

Since the transmitters (projectors) can be keyed in frequency, we can envision a group of projectors being employed allowing navigation throughout a space by a robot where the robot "hops" between projectors, much like your cell phone as you move around in the cell network. 

Cost is somewhat high for this detector but that is because it is a completely integrated solution.  If you are interested in making some widget to do localization, this is not likely the tool for you.  On the contrary, if you are looking to benefit from localization and do more work on that foundation, this is a slam dunk product.  You can configure it, plug it in, and you are up and running. 

Our only complaint was the connector which is always a sore spot in robotics, regardless of the choice made.  This particular connector is for manufacture and high-volume, not for research and one-of-a-kind applications.  It is not a show stopper, just tough to work with for us. 

 
 
voice: 720-564-0373, email: sales@acroname.com, address: 4822 Sterling Dr., Boulder CO, 80301-2350, privacy
© Copyright 1994-2008 Acroname, Inc., Boulder, Colorado. All rights reserved.