Im trying to develop a system that autonomously navigates a large outside space, where accuracy is vital (GPS is too inaccurate). There are a number of options but have largely been used inside, has anyone tried these or used anything else?
WiFi triangulation, Dead reckoning, RFID landmarks
I would look at the System "PIKSI" from Swift Navigation
It uses one fixed GPS modul and one which is carried on your Robot. Your accuracy is about 5 centimeters, which is pretty good for a GPS based system
I haven't tried it yet, so I cannot share experience, but I think it is worth a try.
Here are some links:
http://swiftnav.com/piksi.html
https://www.kickstarter.com/projects/swiftnav/piksi-the-rtk-gps-receiver
I have tried using Decawave's DW1000 chips. You can now get these on little boards:
These can measure the distance between each other by measuring the time of flight of an Ultra-Wideband message. The precision is about 10cm, and the trueness depends on a lot of factors.
You have a few as base stations, and one or more on each robot. By triangulation, you can work out the location of the robot.
I can report that the chips are insanely complex to program, and the example code they provide is a huge hairball that looks like it was written by an intern. However, they do work, and the other benefit is that you can put a data payload in the messages, so they solve your communications problems too.
You can fuse Laser mesurements (which are very precise) with GPS mesurements to navigate and localize the obstacles. I think this is the solution used by google car, but be careful: Laser finders can be very expensive.
