| Popis: |
Position, Navigation, and Timing (PNT) is a critical capability that has become indispensable to the United States military, government services, and the commercial sector. This paper presents a method for distributed localization in Inertial Navigation Systems (INS) without GPS. Traditional INS incur integration drift error over operationally relevant times scales, but this drift can be reduced by using a nonlinear filter to compute the joint state p(x|z) of all agents in the network subject to range measurements z between pairs of agents. The goal of this work is to increase the localization accuracy of each agent’s position, velocity, and orientation. To compute the joint state, which includes cross correlations between agents, agents share their inertial measurement unit (IMU) measurements and exchange range measurements at some specified frequency. Tracking the cross correlations between agents allows full utilization of the range measurements via a nonlinear filter, which allows flexibility regarding each agent’s sensor configuration and any additional sensor inputs. In this paper we formulate this method, demonstrate its effectiveness with simulations that use hardware-in-the-loop (HWIL) IMU and range measurements, and investigate the effect of dropped IMU and range measurements on final localization accuracy. Our results show that errors incurred by dropped IMU measurements can be ameliorated by inflating the process noise covariance matrix, and that up to 90% of range messages can be dropped without degrading performance with respect to a baseline. Finally, we also present network communication analyses for IMU measurement broadcasts and range messages. The HWIL data used in these simulations was collected during field experiments conducted by Raytheon BBN Technologies. |
