Last Christmas I was given a Raspberry Pi for Christmas and dived into building a robot to roam around my large garden.
I found :
- the GPS was highly inaccurate due to the large trees blocking the view of the sky
- the Accelerometer compensated compass worked well away from the robot but I never managed to get it calibrated properly to take account of the varying Magnetic fields of the two 500W motors driving the thing. Placing the compass on a stick a metre above the robot worked to an extent but the constant accelerations from the swaying motion drastically reduced the accuracy.
- My basic attempts at odometry worked to an extent but were stymied by the lack of accurate compass and rough ground.
Summer came and I had other things to do.
Now I am coming back to the project with new enthusiasm and new ideas and I would like to hear from anybody who has attempted to fix position using triangulation of either radio transmissions or visual reference points.
I vaguely have in mind three Xbees transmitting from three known locations. If I could measure the angle between two pairs of signals I could calculate the position. I suppose I could have three rotating directional areials but has anybody made a better system.
In the same way if I had three recognisable targets at known positions I could centre the target in the camera and then measure angles. If I could use the size of the image to estimate distance we could go down to two targets.
Has anybody done any of this?
How accurate was it?
What are the pitfalls?
What is the best equipment?