An Incremental Robust Underwater Navigation with Expectation-Maximisation
- Publication Type:
- Conference Proceeding
- Citation:
- Australasian Conference on Robotics and Automation, ACRA, 2022, 2022-December
- Issue Date:
- 2022-01-01
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
This paper presents a robust navigation solution using low-cost visual-inertial sensors in a 6-Degree of Freedom (DoF) environment. That is an incremental/online navigation solution using the nonlinear least-squares optimisation with classification expectationmaximisation (EM). In this problem, weights are assigned to each measurement observation using the Cauchy function that are iteratively computed from the errors between predicted robot poses and the observed robot measurement. However, the computational cost is quite high in solving the full-batch estimation via Gauss-Newton. By implementing the sliding window filter (SWF), we introduce an incremental EM based robust navigation where the computational cost is shown a significant reduction compared to the full robust batch estimation. The impact of window size on the navigation performance is studied given the dataset is unknown to predict the optimum window gating. This allows a robust constant-time estimation of the robot pose. Such a capability is desirable in underwater navigation applications such as intervention missions. We verify this work using the experimental dataset collected by the UTS submersible pile inspection robot (SPIR).
Please use this identifier to cite or link to this item: