Indoor Wi-Fi propagation model characterization is crucial to the development of the Wi-Fi access points. The existing characterization methods requires lots of human intervention and a complex sampling process, causing a high labor intensity and low estimation efficiency. We propose to use multiple robot agents to free human workers from the characterization task. We formulate the propagation model parameter estimation problem and the robots path-planning problem into an online Maximum Likelihood Estimation (MLE) problem and an online Markov Decision Process (MDP) problem, respectively. Online MDP is solved with online value iteration while the near optimal MLE solutions are approximated with Particle Swarm Optimization tool. Simulations results proves that online MDP path-pla...[ Read more ]

Indoor Wi-Fi propagation model characterization is crucial to the development of the Wi-Fi access points. The existing characterization methods requires lots of human intervention and a complex sampling process, causing a high labor intensity and low estimation efficiency. We propose to use multiple robot agents to free human workers from the characterization task. We formulate the propagation model parameter estimation problem and the robots path-planning problem into an online Maximum Likelihood Estimation (MLE) problem and an online Markov Decision Process (MDP) problem, respectively. Online MDP is solved with online value iteration while the near optimal MLE solutions are approximated with Particle Swarm Optimization tool. Simulations results proves that online MDP path-planning algorithm can easily outperform a number of traditional algorithms in terms of sampling efficiency and the quality of the sample set, resulting a smaller parameter estimation error and a faster convergence rate for the online estimation iterations.