Ing. Martin Gurtner, Ph.D.

This paper presents an approach for accurately dropping a relatively large amount of fire retardant, water or some other extinguishing agent onto a wildfire from an autonomous unmanned aerial vehicle (UAV), in close proximity to the epicenter of the fire. The proposed approach involves a risky maneuver outside of the safe flight envelope of the UAV. This maneuver exploits the expected weight reduction resulting from the release of the payload, enabling the UAV to recover without impacting the terrain. The UAV is tilted to high pitch angles, at which the thrust may be pointed almost horizontally. The vehicle can therefore achieve higher horizontal speeds than would be allowed by conventional motion planners. This high speed allows the UAV to significantly reduce the time spent close to the fire. As a result, the overall high heat exposure is reduced, and the payload can be dropped closer to the target, minimizing its dispersion. A constrained optimal control problem (OCP) is solved taking into account environmental parameters such as wind and terrain gradients, as well as various payload releasing mechanisms. The proposed approach was verified in simulations and in real experiments. Emphasis was put on the real time recalculation of the solution, which will enable future adaptation into a model predictive controller (MPC) scheme.

In this paper we document a novel laboratory experimental platform for noncontact planar manipulation (positioning) of millimeter-scale objects using acoustic pressure. The manipulated objects are either floating on a water surface or rolling on a solid surface. The pressure field is shaped in real time through an 8-by-8 array (matrix) of ultrasonic transducers. The transducers are driven with square voltages whose phase-shifts are updated periodically every 20 milliseconds based on the difference between the desired and true (estimated from video) position. Numerical optimization is used within every period of a discrete-time feedback loop to determine the phase shifts for the voltages. The platform can be used as an affordable testbed for algorithms for non-contact manipulation through arrays of actuators as all the design and implementation details for the presented platform are shared with the public through a dedicated git repository. The platform can certainly be extended towards higher numbers of simultaneously yet independently manipulated objects and larger manipulation areas by expanding the transducer array.

Ball and hoop system is a well-known model for the education of linear control systems. In this paper, we have a look at this system from another perspective and show that it is also suitable for demonstration of more advanced control techniques. In contrast to the standard use, we describe the dynamics of the system at full length; in addition to the mode where the ball rolls on the (outer) hoop we also consider the mode where the ball drops out of the hoop and enters a free-fall mode. Furthermore, we add another (inner) hoop in the center upon which the ball can land from the free-fall mode. This constitutes another mode of the hybrid description of the system. We present two challenging tasks for this model and show how they can be solved by trajectory generation and stabilization. We also describe how such a model can be built and experimentally verify the validity of our approach solving the proposed tasks.

In this paper, we propose a novel approach to obtain a reliable and simple mathematical model of dielectrophoretic force for model-based feedback micromanipulation. Any such model is expected to sufficiently accurately relate the voltages (electric potentials) applied to the electrodes to the resulting forces exerted on microparticles at given locations in the workspace. This model also has to be computationally simple enough to be used in real time as required by model-based feedback control. Most existing models involve solving two- or three-dimensional mixed boundary value problems. As such, they are usually analytically intractable and have to be solved numerically instead. A numerical solution is, however, infeasible in real time, hence such models are not suitable for feedback control. We present a novel approximation of the boundary value data for which a closed-form analytical solution is feasible; we solve a mixed boundary value problem numerically off-line only once, and based on this solution, we approximate the mixed boundary conditions by Dirichlet boundary conditions. This way, we get an approximated boundary value problem allowing the application of the analytical framework of Green's functions. The thus obtained closed-form analytical solution is amenable to real-time use and closely matches the numerical solution of the original exact problem.

Various optical methods for measuring positions of micro-objects in 3D have been reported in the literature. Nevertheless, the majority of them are not suitable for real-time operation, which is needed, for example, for feedback position control. In this paper, we present a method for real-time estimation of the position of micro-objects in 3D1; the method is based on twin-beam illumination and requires only a very simple hardware setup whose essential part is a standard image sensor without any lens. The performance of the proposed method is tested during a micro-manipulation task in which the estimated position served as feedback for the controller. The experiments show that the estimate is accurate to within ∼3 μm in the lateral position and ∼7 μm in the axial distance with the refresh rate of 10 Hz. Although the experiments are done using spherical objects, the presented method could be modified to handle non-spherical objects as well.