Planning Robust Sensor Relocation Trajectories for a Mobile Robot with Evolutionary Multi-objective Optimization

Wireless sensor networks provide a method for monitoring a region of interest. Incorporating a mobile robot within the sensor network allows various types of functionality to be added. One example of this is the replacement of risky and/or damaged sensors with other functional, passive ones. Using a specially designed risk management framework (RMF), we can proactively detect sensors that are at a high risk for failure and replace them before any network coverage is lost. The problem of optimizing the robot trajectory while picking up passive sensors and dropping them at the locations of the damaged sensors in the field has been studied as the “Robot-Assisted Sensor Relocation” (RASR) problem. One shortcoming of existing RASR methods is that the chosen robot trajectory is the one with the shortest length; however, no regards as to the durability of the passive sensors in the relocation chain are taken into consideration. We propose a more robust manner to come up with these trajectories by taking into account the current energy levels of the participating passive sensors as well as the ideal locations for their deployment. We resort to multi-objective optimization (MOO) to handle the tradeoffs among the different decision objectives that are part of this new formulation, named here as “Reliable Robot-Assisted Sensor Relocation”. We outline the RRASR problem as well as the RMF used for detecting risky sensors in the wireless sensor network before the calculation of the sensor relocation trajectory takes place. We also evaluate the performance of six state-of-the-art evolutionary multi-objective optimization (EMOO) algorithms with sensor networks of varying sizes, inflicted damage levels, and passive sensor densities. The empirical results confirm the feasibility of utilizing EMOO approaches to suggest multiple sensor relocation trajectories to the network manager.