Re: More Info on the GuideCane

From: David Andrews (dandrews@visi.com)
Date: Thu Oct 09 1997 - 20:16:13 PDT


Sounds like a bad idea to me. What does it do about drop offs, and it
sounds like following this thing around, you could get disoriented. don't
they get it??? Thgere is a reason we touch things with our canes, we then
know where they are, get landmarks, and often know what they are.

David Andrews

On 1997-10-09 nfb-rd@lothlorien.nfbcal.org said:
 LR>The following is copied from IEEE Spectrum, October 1997. It gives
 LR>us a better idea of the capabilities of the GuideCane from the
 LR>University of Michigan.
 LR>I didn't find very much useful information on the inventor's web
 LR>page: http://www-personal.engin.umich.edu/~johannb/
 LR>but maybe I didn't look in the right place.
 LR>[INLINE
 LR>________________________________________________________________
 LR>Seeing-eye cane steers the blin
 LR>A new type of electronic navigational aid for the blind may soon joi
 LR>white canes and seeing-eye dogs. Developed by scientists at th
 LR>University of Michigan in Ann Arbor, the GuideCane is an array o
 LR>ultrasonic obstacle-sensors attached to a wheeled handle, or cane
 LR>Operation is simple. The user selects a direction of travel--say
 LR>ahead--by using a joy stick on the cane. He or she then pushes th
 LR>GuideCane forward with one hand, much as if it were a golf cart. T
 LR>turn right or left, the person presses the joy stick to the right o
 LR>left, whereupon a servo turns the wheels in the appropriate
 LR>direction When the sensors detect an obstacle, an on-board 25-MHz
 LR>486 compute determines the best path around it and directs the
 LR>servo to turn th guide wheels. Through the cane, the user feels the
 LR>change in directio and follows. After the obstacle is passed, the
 LR>cane brings the use back to the desired direction of travel
 LR>The GuideCane weighs only 4 kg, because it does not have to supply
 LR>it own driving power. As a result, it may be picked up and carrie
 LR>upstairs or down, if necessary, explained the unit's inventor, Johan
 LR>Borenstein, head of the Mobile Robotics Laboratory in the
 LR>university' department of mechanical engineering and applied
 LR>mechanics The prototype GuideCane, built by graduate student Iwan
 LR>Ulrich, has 1 sensors and can sense any obstacle within a
 LR>120-degree angle in fron of the user [see figure]. The sensors,
 LR>made by Polaroid Corp. Cambridge, Mass., are the same as those used
 LR>to focus camera automatically, and they operate in much the same
 LR>way. They each emit pulse of ultrasound and wait for the echo. The
 LR>interval between th two events determines the distance to the
 LR>obstacle, while th orientation of the particular sensor that
 LR>detects the reflected puls determines the angle. Whenever a sensor
 LR>detects an obstacle, th information is passed along to the computer
 LR>for it to enter the new on a two-dimensional occupancy grid of
 LR>10-by-10-cm squares. Thus a all times the computer possesses a map
 LR>of the obstacles in the are around the cane
 LR>"But what good is a map of obstacles," asked Borenstein, "if th
 LR>GuideCane doesn't know where it is on the map?" So the uni
 LR>incorporates odometry to keep track of its location. Encoders on th
 LR>guide wheels measure their rotation to a fraction of a revolution
 LR>This is translated into the distance traveled from a point of origin
 LR>When the GuideCane swivels, one wheel moves farther than the other
 LR>and the information is related back to a direction. To eliminate th
 LR>errors that accumulate in this type of measurement, Borenstein plan
 LR>to add a compass so that all measurements can be tied to magneti
 LR>north
 LR>The inventor has added a couple of his own patented algorithms to th
 LR>GuideCane's arsenal of obstacle avoidance techniques. One of the
 LR>calculates the best path around obstacles, based on information fro
 LR>the occupancy grid. The other addresses the uncertainty that
 LR>reflection received by a sensor echoes its own pulse. In a typica
 LR>environment, the pulses undergo multiple reflections. Other thing
 LR>being equal, the uncertainty could force a wait to fire a pulse unti
 LR>the echoes from a previously fired sensor have died away. Bu
 LR>Borenstein's method attaches signatures to the pulses so that eac
 LR>sensor can determine whether an echo came from its own or another'
 LR>pulse. Because the sensors can now fire more often, much faster
 LR>trave speeds are open to the user
 LR>At the moment Borenstein is grappling with more immediate issues,
 LR>suc as the GuideCane's ability to detect different surfaces. The use
 LR>should be able to recognize that he is veering off onto someone'
 LR>lawn, for example. To solve this problem, Borenstein plans to fring
 LR>the sides of the GuideCane with whiskers, which will sense the edge
 LR>o the grass. Normally, the tips of the whiskers will be a couple o
 LR>centimeters above the cement; but if the Cane starts to veer off th
 LR>sidewalk, the user will feel and hear the whiskers brushing along th
 LR>lawn
 LR>A second problem is less tractable and is due to specular reflection
 LR>When ultrasound pulses from a sensor bounce off smooth surfaces lik
 LR>windows or mirrors, they do not reflect back to the sensor unless th
 LR>surface is directly ahead. So if the person is approaching, say,
 LR>glass wall at an oblique angle, the wall may be invisible to th
 LR>GuideCane, and a collision may result. This is not the worst thin
 LR>that can happen, Borenstein believes, since it is the GuideCane tha
 LR>is colliding with the surface and not the blind user. Nevertheless,
 LR>i is inelegant. "There is no good solution for this," Borenstein tol
 LR>IEEE Spectrum. "The most practical solution is a good bumper aroun
 LR>the GuideCane.
 LR>"A preliminary version of the prototype has been tested by visuall
 LR>impaired individuals and the reaction was extremely positive," sai
 LR>Borenstein. "But more development will be required before the devic
 LR>is ready for widespread commercial use.
 LR>________________________________________________________________
 LR>Linda Geppert, Edito
 LR>________________________________________________________________
 LR>contents feedback search hom
 LR>IEEE Spectrum October 1997 Volume 34 Number 1
 LR>(c) Copyright 1997, The Institute of Electrical and Electronic
 LR>Engineers, Inc

David Andrews (dandrews@visi.com)
or BBS: (telnet to nfbnet.org) or call (612) 696-1975

Net-Tamer V 1.10 Beta - Registered



This archive was generated by hypermail 2b29 : Sun Dec 02 2012 - 01:30:04 PST