Uncertainty-aware Sensor Deployment Optimization for Radio Map Construction

Radio maps find a wide range of applications in wireless communications. Constructing radio maps involves deploying sensors to derive spectrum data, based on which the overall spectrum situation is estimated. Due to cost and other practical constraints, only a limited number of sensors can be deployed and thus how to reasonably deploy sensors becomes one of the crucial issues. The challenges of sensor deployment mainly lie in the fact that both environment and signal source information affect the spatial pattern of the radio map, which is difficult to model and exploit. This study presents a novel deployment strategy that harnesses the model-estimated radio map as prior information, thereby enhancing radio map construction efficiency and accuracy. By pre-sampling to acquire real spectrum data and assessing global spectrum error, the proposed method introduces error uncertainty and formulates an optimization function that adheres to submodularity to minimize error uncertainty. Experiments indicate that our method achieves an average reduction in root mean square error (RMSE) of 22.6% and 16.3% against uniform deployment and random deployment, respectively, across a 1% to 10% sampling rate range.