Extended Preintegration for Relative State Estimation of Leader-Follower Platform

TL;DR

This paper presents an extended IMU preintegration method for relative state estimation of leader-follower platforms, improving real-time tracking accuracy and robustness in challenging environments.

Contribution

It introduces an analytic extended preintegration constraint for two-platform IMU data, integrated into a unified graph optimization framework for enhanced relative state estimation.

Findings

Outperforms existing approaches in simulations

Enables accurate 6DoF visual tracking in VR applications

Provides a real-time, robust estimation method for challenging scenarios

Abstract

Relative state estimation using exteroceptive sensors suffers from limitations of the field of view (FOV) and false detection, that the proprioceptive sensor (IMU) data are usually engaged to compensate. Recently ego-motion constraint obtained by Inertial measurement unit (IMU) preintegration has been extensively used in simultaneous localization and mapping (SLAM) to alleviate the computation burden. This paper introduces an extended preintegration incorporating the IMU preintegration of two platforms to formulate the motion constraint of relative state. One merit of this analytic constraint is that it can be seamlessly integrated into the unified graph optimization framework to implement the relative state estimation in a high-performance real-time tracking thread, another point is a full smoother design with this precise constraint to optimize the 3D coordinate and refine the state for the refinement thread. We compare extensively in simulations the proposed algorithms with two existing approaches to confirm our outperformance. In the real virtual reality (VR) application design with the proposed estimator, we properly realize the visual tracking of the six degrees of freedom (6DoF) controller suitable for almost all scenarios, including the challenging environment with missing features, light mutation, dynamic scenes, etc. The demo video is at https://www.youtube.com/watch?v=0idb9Ls2iAM. For the benefit of the community, we make the source code public.