Relative Pose Estimation using Range-only Measurements with Large Initial Uncertainty

Cooperative Relative Localization Using Range Measurements Without a Priori Information

“…In our recent work, [33], we have proposed a Multi-Hypothesis Extended Kalman Filter (MHEKF) to estimate relative position and relative heading for two vehicles using range-only measurements. Although the Multi-Hypothesis Extended Kalman Filter sounds similar to the Multi Hypothesis Kalman Filter (MHEKF) or more commonly referred to as the Multi Hypothesis Object Tracking EKF (MHOT-EKF), both these techniques have different structures as well as applications.…”

“…The MHOT-EKF is a technique used for tracking multiple objects and enables efficient tracking even when data association cannot be made, specially in crowded environments [34,35]. The MHEKF proposed in [33] and further improved upon in this paper is a technique developed to address the issue of initialization of estimation algorithms in the presence of large initial uncertainty or low/none apriori information, especially for platforms with low processing power and real-time implementation. The initialization problem for the relative position was solved using a-priori noisy heading information in [33].…”

“…The MHEKF proposed in [33] and further improved upon in this paper is a technique developed to address the issue of initialization of estimation algorithms in the presence of large initial uncertainty or low/none apriori information, especially for platforms with low processing power and real-time implementation. The initialization problem for the relative position was solved using a-priori noisy heading information in [33]. We used a simple χ 2 elimination technique to identify filters that provided the "best" estimate and prune out filters with large covariances.…”

“…In this paper, we extend the 2-vehicle relative localization work [33] to an N-vehicle cooperative relative localization problem using the concept of MHEKF to initialize an Nvehicle joint filter with no a-priori information (no position or heading information is present). The goal of the Nvehicle filter is to cooperatively estimate the relative pose and orientation of the vehicles with respect to the central vehicle using range measurements between vehicles.…”

“…• Improvement of the Multi-Hypothesis Extended Kalman Filter (MHEKF) technique introduced in [33] by removing a-priori heading information. A rule of thumb while choosing the initial number of filters for the MHEKF has also been provided in the paper, which is a function of the overlap between the 2σ uncertainty ellipses around the different initial points.…”