Revision history [back]

Question 1: Is it feasible to use ROS on a platform that will operate outside and potentially a few miles away from a computer running a "master" ROS node?

This is really going to depend on network connectivity. If you can guarantee connectivity it might work but it will be tough. Since you have a large vehicle that may interact with humans I would be a bit concerned. You'll want to have a plan for what to do when you have network drop outs. For what it is worth ROS 2 has better QoS support and might make this approach easier.

Question 2: I have been seeing an awful lot of navigation using lasers and maps. What is your trying to operate in the confines of a private property with various terrain (Woods, Ditches, Small Marshes, etc) and the only boundary is chain link fences. Can a map really be made using the LIDAR?

Generally speaking LIDAR is your go-to tool for navigation as it is much faster to process. I am not super familiar with the RPI lasers but they may be a little under spec'd for your application. My primary concern would be your ability to detect and stop / avoid obstacles while traveling at speed. Something like a human, animal, or large log you're not going to be able to roll over and you'll want to be able to navigate around. Maps won't give you that sort of data.

Question 3: Is it worth the time to try and build a URDF model and somehow get terrain data from Google to simulate this project?

Simulation is always worth it. It will let you test concepts faster and without having to get dirty. Since it sounds like you are moving around a lot of metal you'll want to be fairly confident of your abilities before field testing. As to an URDF you are going to get from it what you put into it. An approximate URDF (i.e. something that captures the rough layout without the fine detail) should at least let you visualize your sensors in situ using RVIZ. Terrain certainly couldn't hurt and would probably make your simulation and navigation more reliable.

Question 4: If you were in my shoes. how would you approach this solution?

I would actually start with simulation and use it for hypothesis testing first. You'll have a lot more development time if you aren't out in the field up to your elbows in robot. I would also consider going with a vendor like ClearPath for your AGV vehicle. Having a vehicle will give you a known-good configuration to start with, and address most of your initial needs so you can focus on application development.