JAMES SVACHA
AERIAL ROBOTICS

Projects

The following is a list of major projects, with downloads for the corresponding papers, that I have worked on over the course of my graduate studies at Penn.

Vision-Aided Inertial State Estimation For Quadrotors [Download PDF]

The purpose of this project was to develop an estimator that only needs an IMU to estimate the tilt and velocity of the quadrotor when yaw is unknown. Optionally, we can add a camera to the robot and use feature-based methods to improve the yaw estimates. The tilt and velocity estimates from the filter can be used to reject outlying feature correspondences in a computationally efficient way. The video below shows our system in action.

Inertial State Estimation for Quadrotors [Download PDF]

In this project, we use the developed drag and thrust models from the drag compensation project below to estimate the orientation and velocity of a quadrotor UAV with only the IMU as sensory input. The tilt component of attitude can be estimated, as well as the body-frame linear velocity and the yaw can be dead-reckoned, although it will drift. The tilt and velocity are all that are needed to stabilize the quadrotor, and may be estimated if the yaw is known. In addition, our method makes it possible to estimate the thrust coefficient and first-order drag coefficients of the quadrotor's dynamic model. The video below shows the system in action.

Modeling and Compensation for First-Order Drag Effects on Quadrotors [Download PDF]

The contributions of this project were to 1) develop a simple lumped-parameter dynamic model for drag effects in the horizontal and vertical directions and 2) develop a controller to compensate for these effects. The video below shows the controller in action for the case of forward flight at a speed of around 5 m/s indoors. The lagging quadrotor doesn't compensate for drag in the control loop and the leading one does. Even at these moderate speeds, a significant difference is seen in the position error. At high speeds (near 20 m/s), the position error can be around one meter without drag compensation.

Newtonian-Eulerian State Estimation For Quadrotors

The full system dynamics can be used to estimate additional parameters such as moment of inertia and the offset of the IMU from center of mass. This work is in progress and nearing completion.

Wind Estimation For Quadrotors

By fusing vision with the filtering frameworks described below, the quadrotor can be used as an anemometer. This work is in progress.