Lean and robust phenomenon boundary approximation

We present an algorithm for computing a compact polygonal boundary of physical phenomena that nodes in a wireless sensor network can detect. It works strictly local and is robust against measurement errors and packet loss. It can deal with non-convex shapes and assumes nothing but location awareness at the nodes and a loose time synchronization. Other than different schemes described in literature our algorithm does not rely on a hierarchical clustered communication structure. After discussing related approaches and introducing our boundary approximation scheme in detail, we show through an extensive set of simulations that it is robust and accurate.