University of Twente Student Theses


Geometric state estimator tightly-coupling force and pose estimation for interaction in vision-denied environments

Bongertman, M. (2020) Geometric state estimator tightly-coupling force and pose estimation for interaction in vision-denied environments.

[img] PDF
Abstract:In recent years, UAVs showed promising capabilities for different applications. In the public, UAVs proved to be useful in applications such as logistics and inspection. As new advancements are being made by the robotics community, UAVs can also be used for interaction tasks. For such tasks, the employment of UAVs establishes 'safe for humans' working environments in critical operations. Examples of such physical tasks and environments are urgent or regular maintenance tasks on wind turbines, assistance at nuclear plants in case of a catastrophe, safety assessments of newly-blast generated voids into mines, or even assistance during natural disasters such as earth-quakes. To accomplish stable and controlled interaction, both a controller guaranteeing stability and knowledge about the interaction is needed. The interaction force and torque could be measured using a sensor. However, this limits the applicability by size, costs, and weight, which consequently limits the battery time. Besides, force and torque sensors can only measure at the point where they are mounted. A common alternative is to use observers based on the nonlinear dynamics of the robot. This allows for estimating external forces and torques. Nonetheless, these methods only work well for practical situations if the forces are large and the noise is small. State estimators suffer less from this problem as they are designed to consider process and measurement noise. Recently a work has been published which proposes to tightly-couple dynamics and pose estimations into a state estimator. Intuitively, this adds information to the pose estimation resulting in an accuracy increase for the estimated pose. This thesis proposes a tightly coupled pose wrench estimator intended for physical interaction using an aerial robot. The state estimator can estimate the pose and external force by using known actuation inputs and the robot's dynamics. In previous work, the benefits of such an approach have been shown for disturbances instead of controlled contact interactions. This thesis uses the estimated external force for interaction scenarios where a UAV engages in physical contact with its environment. Furthermore, the state estimator exploits measurements from the Global Positioning System (GPS) and Inertial Measurement Unit (IMU). Also, the estimator is able to estimate states belonging to the Lie group SO(3). The estimator is validated through realistic Gazebo simulations and a dataset of an experiment where a UAV physically interacts with a static object. The estimator has been tuned and estimations have reached an accuracy in the same order of magnitude as the simulated accelerometer and gyroscope. Also, the simulations and experiment show that the UAV is capable of accurately estimating the contact force while simultaneously estimating the UAVs pose. Recommendations on future work are to integrate Visual Inertial Odometry (VIO) to realize GPS-Aided VIO and to make use of an automated tuning scheme, as the quality of the estimation is highly dependent on the set of tuning parameters given to the estimator.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:30 exact sciences in general, 53 electrotechnology
Programme:Systems and Control MSc (60359)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page