Message-Passing-Based Distributed Cooperative Simultaneous Localization and Synchronization in Dynamic Asynchronous Networks

Location awareness is a key enabling technology for many applications and services of the Internet of Things (IoT). Since densely deployed heterogeneous agents in IoT typically have mobility and different qualities of internal clocks, as well as limited computation and communication capabilities, high-precision network localization is a challenging problem. Existing methods do not compensate the position variation caused by the agent mobility during a measurement phase, which will result in estimation error, and have high-computational complexity. In this article, we propose a cooperative, distributed, and low-complexity algorithm for network simultaneous localization and synchronization (SLAS), which is suitable for large-scale network consisting of heterogeneous agents with mobility, time-varying clock and time-varying connectivity. We first propose a new measurement model based on the asymmetric time-stamped communication scheme, which compensates for the position variation of each agent within a measurement phase. Second, we construct a factor graph (FG) to represent the underlying Bayesian estimation problem, and apply belief propagation to obtain the marginal distribution of each agent's state. To deal with the complex nonlinear measurements, we extend the posterior linearization technique by using iterative statistical linear regression with respect to the joint posterior of neighboring agents. All the messages on FG are derived in Gaussian form and the computational complexity at each agent is linear in the number of neighboring agents, which has significant advantages in large-scale networks. Simulation results demonstrate that the proposed algorithm has better estimation performance and lower average running time compared to existing methods.