I agree with the first 2 points, but the last doesn't follow. Disaster response isn't a good domain for end users to play around with different configurations

I agree with the first 2 points, but the last doesn't follow. Disaster response isn't a good domain for end users to play around with different configurations to see what works. Few people are first responders, disasters rarely happen, and when they do it's no time to be experimenting with your homebrew robot. Something that's closer to a consumer / small business task, and is tedious, common, and not critical to get right would be a better problem space to explore.