Multi-Epoch Multi-Agent Collaborative Localization Using Grid-based 3DMA GNSS and Inter-Agent Ranging

Abstract

GPS navigation in urban environments is prone to error sources such as multipath and signal blockage. In our previous work, we demonstrated a snap-shot multi-agent collaborative 3D-mapping aided (3DMA) GNSS localization algorithm to reduce the impact of these error sources on the position solution. However, a source of error in snap-shot algorithms is jump discontinuities in the position solution due to fluke noise measurements. Therefore, an approach which takes into account temporal correlation between agent positions will further increase the robustness of the system through jump discontinuity error mitigation. In this paper, we present a multi-epoch decentralized collaborative localization algorithm for urban navigation using 3DMA GNSS and inter-agent ranging. The algorithm is a multi-epoch variant of our previous snap-shot multi-agent collaborative 3DMA localization algorithm and uses Intelligent Urban Positioning (IUP) and ambiguity mitigation by constraining an agents probability distribution using its neighbors. The key additions that extend our prior algorithm to a multi-epoch framework are the inclusion of a velocity estimation framework using banks of Extended Kalman filters and a discretized Bayes filter prediction step to propagate a position probability distribution in time. Furthermore, the methodology is applicable to sparsely connected asynchronous networks, and has limited information exchange. The proposed method is validated on simulated datasets in an urban area of Champaign, Illinois with multiple agents in a variety of scenarios. We demonstrate improved performance in terms of positioning accuracy. We also analyze the impact of network connectivity on positioning accuracy.