Visual-inertial self-calibration on informative motion segments

Abstract

Abstract-Environmental conditions and external effects, such as shocks, have a significant impact on the calibration parameters of visual-inertial sensor systems. Thus longterm operation of these systems cannot fully rely on factory calibration. Since the observability of certain parameters is highly dependent on the motion of the device, using short data segments at device initialization may yield poor results. When such systems are additionally subject to energy constraints, it is also infeasible to use full-batch approaches on a big dataset and careful selection of the data is of high importance.In this paper, we present a novel approach for resource efficient self-calibration of visual-inertial sensor systems. This is achieved by casting the calibration as a segment-based optimization problem that can be run on a small subset of informative segments. Consequently, the computational burden is limited as only a predefined number of segments is used. We also propose an efficient information-theoretic selection to identify such informative motion segments. In evaluations on a challenging dataset, we show our approach to significantly outperform state-of-the-art in terms of computational burden while maintaining a comparable accuracy.

Observability-Aware Trajectory Optimization: Theory, Viability, and State of the Art

“…The problem of trajectory optimization has received substantial attention in the literature and a wide variety of optimization techniques exist. One the most common methods for trajectory optimization in robotics is to maximize a norm of the Fisher information matrix (FIM) [6]- [9]. The aim of such approaches is to find a trajectory that increases the amount of information available from exteroceptive sensor measurements about the relevant system states and parameters.…”

“…Additionally, the algorithm in [12] filters out trajectory segments for which the state is unobservable, based on the conditioning of the FIM. Schneider et al develop a similar approach in [6], utilizing differential entropy to quantify the information content of each trajectory segment. Usayiwevu et al [13] minimize the posterior covariance of the extrinsic transform parameters for a lidar-IMU self-calibration task through informative path planning.…”

On the Redundancy Detection in Keyframe-Based SLAM

“…Information-based measures have been widely used for various tasks throughout the computer vision and robotics communities, all the way from showing how mutual information can guide the feature matching process in visual SLAM in [4], to using positional covariance to construct skeletal graphs for large Structure-from-Motion problems, in order to speed up the expensive computation process, in [30]. Schneider et al [29] evaluate trajectory segments regarding their entropy with respect to the calibration parameters of the sensor suite in order to estimate this calibration, while Mu et al [21] use entropy to select the most important landmarks for collision avoidance. In [11], information is quantified in order to to calculate the optimal path of a UAV for 3D reconstruction of a scene of interest.…”

Simultaneous self-calibration and navigation using trajectory optimization

“…The method is applied to a 2D simultaneous localization and mapping (SLAM) problem with landmarks. A similar approach was applied in the visual–inertial odometry (VIO) setting in Schneider et al (2017) using an approximated, more computationally efficient information metric. Other works focus on detecting changes in the self-calibration parameters and directly modeling the parameter drift in the estimation stage, such as Nobre et al (2017).…”