GeoBot
      1


      
Sonar
      Array Ground Plane and Final Testing


      
Updated 4/3/06

      
This will be a
      follow up article, on the quadrate sonar of the Geobot. Last time, we detailed the theory
      of operation of our multi zone pattern, and how we tailor the
      robots response depending on what zone the closest danger is
      in. This works very well as you will see at the end in a small
      movie clip. One additional problem remained to be solved however
      for the system to work in the outdoor environment.

      
Seeing the Ground

      
While the drawings
      below show a nice 30 degree cone pattern for the sonar beam,
      in fact it is a spray in every direction in front of the emitter
      of a loud super sonic chirp. The actual shape is very wide -
      in fact 180 degrees next to the emitter, and narrows down to
      25 degrees at 4 feet. Since the sonar is 6 inches off the ground,
      it receives a very strong signal from the outdoor gravels, rocks
      and sand that lies flat on the ground. A constant 1 to 3 foot
      reflection swamps the data in which we are trying to see obstructions
      in the path of the robot.

      
Shaping the Beam

      
We did a lot of
      research on sonar beam shapes, and discovered that the few individuals
      that tried to change the shape of the beam to narrow it always
      failed or gave up. Such things as fuzzy tubes around the emitter,
      shrink tubing, cork pads, rubber tubes, and a whole array of
      other materials were tried by us with only a very tiny effect
      on the beam pattern. We still could always see the ground. But
      we did find a way to change the beam, and it works very well.

      
Shaping by Reflection

      
This first graphic
      here illustrates the problem well: The bottom of the sonar beam
      strikes the ground, and the nearest edge where it hits bounces
      back and is reported as the obstacle distance. Even tilting the
      beam up did little for us - except to detect the trees and ceiling!
      

      

 The problem did not exist inside the house
      on the work bench or smooth floors, since they have no coarse
      structure to reflect the beam back. It was discovered however,
      that we could DIVERT the beam exactly like a mirror reflects
      the light of a flashlight. By using a grazing incidence reflection,
      we were able to shadow the ground from the view of the emitter
      and receiver. 

      

 After
      some math to determine the size reflector to shield out the ground
      to 4 feet, a 4 inch deep by 10 inch wide proto type was constructed.
      It works like this drawing here, mounted just below the quad
      array, it creates a sonar shadow on the ground in front of the
      robot. Yes a few rocks may slip by, but the bumper will catch
      them if they are tall enough. 

      

 This
      prototype taped on with red cloth tape proved the concept, the
      math was spot on. Field tests in a very rocky environment proved
      it worked fine, and the ground signal was almost always past
      the maximum sonar zone.

      

 The
      final version seen here is made from .09 polycarbonate sheet.
      Its mounted on the bumper so if it hits a rock or tree, the bumper
      will take over. So far, this hasn't happened - the sonar stops
      it long before it hits.

      

      mpg Movie


      
(Click
      Here)
 
      This 500k movie clip will show our outdoor test of the Geobot
      quadrate sonar in action. I placed a array of large blocks of
      petrified wood and repeatedly put the Geobot on a collision course
      with it. After many trials, it never once touched any rock, and
      drove around all of them.


      
 Conclusions

      
This sensor array
      moves us one step closer to our final goal of the Geobot remotely
      collecting rock and soil samples and returning it to the base.
      By including the sonar avoid level in the Priority Arbitration
      Architecture, the robot will continue on with its task after
      avoiding major dangers. Our next step will be to add the Devontech
      digital compass module, to keep the robot going a specific direction.


HOME

FastCounter by bCentral