“RoboSurv” is intended to calculate and optimize a bunch of quantities and metrics that are related to Markov chains. It is motivated by the use of Markov chains in robotic applications where one or a group of robots randomly move on a graph to perform surveillance tasks. These stochastic surveillance strategies for the quickest detection of anomalies or intelligent intruders in network environments. To solve this problem, different algorithms have been proposed. In this package we implement five different algorithms to calculate and optimize related quantities and metrics: mixing time, hittingtime probability, Kemeny constant, entropy rate and return time entropy.
In the optimization part, for solving non-convex optimization problems, we use “fmincon” with SQP solver in Matlab and JuMP with Ipopt solver in Julia. Users also are provided with to solve semidenite programming (SDP) with CVX in Matlab and Convex.jl in Julia. For details of the optimization solver, the users are referred to the above references.
Documentation can be found here