Oct 09, 2010

Meeting Date: 
Saturday, October 9, 2010

Class - Martin Laroque - Navigation

 
Martin gave us a lecture on Robot navigation. You can find his presentation slides here.
 
Some noteworthy tidbits:
 
1) when the antenna's of 2 GPS units are separated more than 3 ft, you can get heading information.
 
2) calibrate your compass once PER robot whenever something on the robot changes (wiring, components, ...)
 
3) keep all wires parallel and twist wire-pairs wherever possible.
 
4) once a gyrocompass is up and running, it is not affected by anything.
 
5) a dip needle is used on the North pole to find the Magnetic North.
 
6) The Variation is the angle between the True North and the Magnetic North. For California, this angle is currently 13.4 degrees.
 
7) The heading information produced by a GPS is not correct. It becomes more accurate when the GPS unit travels faster than 1 mph which is already reasonably fast for a robot. On robots, it is suggested to add a compass sensor. Interesting anecdote: in the DARPA Grand Challenge, contestants that used the GPS compass heading drove in the K-rail at start because of the initial incorrect heading.
 
8) interesting to know: when you use the Python programming language, when you use the math.atan2 function
beware that the parameters are (y,x) and not (x,y).
 
9) a GPS can guide a robot to within 10 ft of a target. After that, vision techniques is the most common technique to get it on target, e.g. to find the orange cone in a RoboMagellan contest.
 
10) for mobile robotics, it is recommended to use odometry for steering but regularly re-align with GPS for location-correction and compass for heading-correction.  gyroscope can help with stabilized heading.
 
11) VisualGPS is a software utility to help with GPS navigation.
 
 

Business

next months contest: Mini-Sumo

next months class: Paralax Propeller 

December Class: Bruce Weimer - navigation using vision

December contest: Talent Show

There was a short discussion on making the Sumo contest more fair by abiding by the weight and size rules more. This gives everyone a fair chance. However, everyone agrees that this excludes some people that have a robot could compete but the limitiations would exclude them. It was agreed upon to have future sumobot contest with 2 categories: strict weight/size limit and an open class to allow other robots to compete.

Our sister club, the Riverside Robotics Club organizes a Robot Expo on Saturday November 6th between noon to 4pm at 16625 Krameria in Riverside at the Woodcrest Library.

Show & Tell

John Davis gave us an update on his RoboMagellan bot:

1) he switched to a Netbook computer instead of the Beagleboard. This because his Devantech I2C compass did not work with his Beagleboard's USB port. This problem is fixed in later models of the Beagleboard but unfortunately not for his.

2) He calibrates his compass by pointing his robot to the 4 Magnetic directions and pushing a button each time. Very simple. This only has to be done once, until the robot configuration changes in the future.

3) The heading values he receives from his compass are a BYTE, which gives values between 0 (NORTH) and 255 (NORTH, 360 degrees later). You have to make provisions in your software for the sudden jump from 255 to 0.

 

 
 
 
 
 
 
 
 
 
 
 
 
Martin Mason is working on a very similar RoboMagellan architecture:
 
Martin also showed us his Compass GUI app which he developed in Python:
 
The compass is a DC-504 dual axis magentic field sensor and the serial-to-USB interface is the USB-to-TTL serial converter CP2102.
 
Martin calibrates his compass by rotting it 360 degrees a couple of times.
 
 
Here is another screenshot of his GPS simulator:
 
 
Remarkable fact is that all members currently working on Robomagellan platforms have chosen the USB interface as their preferred interface between the sensors and the computer/microcontroller.
 
A very popular servo controller turns out to be the Pololu Maestro Servo controller.
 
 
 
Martin also showed us the remote he intends to use with his RoboMagellan: the Logitech Cordless Rumblepad 2. This is connected to your PC wireless via a USB dongle. On the PC, he uses PyGame's joystick interface. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
He also showed a freeware program called: Serial Chart. It takes an incoming stream of data from the COM port and visualizes this in real-time.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
The data that you see in the screenshot is captured from a 5DOF IMU from Sparkfun. It interfaces to the PC via an Arduino board that filters the data using a low-pass filter. The algorithm works something like: 
 
loop
  Read NewValue
  FilterOutput = FilterOutput * FilterTimeConstant  + NewValue * (1 - FilterTimeConstant)
  Pause to make loop time a pretty accurate value.
end loop
 
For better accuracy, you can combine it with a gyroscope and filter it with a Kalman filter, but that is already pretty advances.
 

Bruce showed us an interesting add-on to RoboRealm, to recognize objects.

It is based on AVM - Associative Video Memory Library. A license costs $9.

RoboRealm has integrated an interface to AVM.

It also has a mode to train a robot to follow a path. The Roborealm NOVA Gate module will take snapshops along the way and try to guess its location when detecting the snapshots later on. The robot will try to drive through the virtual gates that were defined during the training pahse.

Here is a video, Bruce posted where he demonstrates his Leaf robot navigating a hallway using the AVM/Roborealm module.

Here is a video showing Leaf robot navigation using AVM Nova Gate module.