Collaborative Geolocation Based on Imprecise Initial Coordinates for Internet of Things

Non-linear minimum optimization methods, like maximum likelihood estimation method, have been heavily employed in range-based localization problems in plenty of research. However, conventional optimization methods require precise initial values. When the initial coordinate is imprecise or even significantly different from feasible solution, these methods can easily get stuck in local optimum solutions. For localization applications in open outdoor areas, these initial values can be provided by Global Navigation Satellite System (GNSS) with acceptable error. On the contrary, in areas with poor satellite signal reception, no precise initial value is available. To solve this problem, we proposed a generally effective approach, degrading the sensitivity of non-linear optimization algorithms on initial value accuracy, which can be applied for collaborative localization in areas with poor or no GNSS signal coverage. Penalty strategy was involved to restrict imprecise initial values to get close to both feasible region and global optimum solution. Both simulations and field tests were carried out to verify the feasibility and efficiency of the proposed approach. The results show that the sensitivity of algorithms on initial value accuracy is reduced efficiently, and accurate location estimation can be obtained even with random values as initial coordinates.