My question is motivated by the definition of electoral districts.
However, this is probably too restrictive and difficult to achieve in face of other restrictions. So, a better solution would be to have districts "as convex as possible". For that to make sense, one would like to have a measure of how close to convex a non-convex set is, and then try to optimize some total non-convexity of a district subdivision. Has any such measure ever been studied?
The application I am interested in is $2$-dimensional, so specialized definitions would be welcome.