Tony A. Wood

Model-based identification and control of the velocity vector orientation for autonomous kites

In this paper we address a control problem for an autonomous tethered kite system for the purpose of airborne wind energy generation. In particular, we design a tracking controller for the velocity vector orientation of the kite. Motivated by empirical data we model the kite steering behaviour as a delayed dynamical system and explicitly utilise the derived model information for the controller design. We identify the involved parameters from experimental data. To adapt to changes in operating conditions we update the parameter estimation on-line. We present results of the derived approach successfully tested in real-world flight experiments.

Model-based flight path planning and tracking for tethered wings

In this paper we propose a guidance strategy for the flight control of a tethered wing. We control the wing trajectory in a cascaded approach via the velocity vector orientation. In particular, we consider a control-oriented model with an input delay to follow a reference path. To account for the delay we design a predictor and use the predictions to compute a reference for a lower level tracking controller. In a path-planning step we design reference figure-eight paths for the wing to follow. The path design explicitly considers the model parameters used in the tracking controller such that limitations induced by the delay are respected. By estimating the model parameters on-line we enable the adaptation of the tracking controller and also the path-planner to time varying conditions, including the input delay and crucially the line length. We present the derivation of the guidance strategy and demonstrate its performance via simulation results.

Nuclear norm minimization algorithms for subspace identification from non-uniformly spaced frequency data

The nuclear norm is an effective proxy for matrix rank in a range of minimization problems, including subspace identification. Nuclear norm-based methods are implemented via iterative optimization methods and in problems with very noisy data the quality of the nuclear norm-based estimate may warrant the additional computation cost. We present two methods (based on the dual accelerated gradient projection and the alternating direction method of multipliers) for nuclear norm based subspace identification in the case where the data is given as irregularly spaced frequency samples.

Predictive Control of Autonomous Kites in Tow Test Experiments

In this letter, we present a model-based control approach for autonomous flight of kites for wind power generation. Predictive models are considered to compensate for delay in the kite dynamics. We apply model predictive control (MPC), with the objective of guiding the kite to follow a figure-of-eight trajectory, in the outer loop of a two level control cascade. The tracking capabilities of the inner-loop controller depend on the operating conditions and are assessed via a frequency domain robustness analysis. We take the limitations of the inner tracking controller into account by encoding them as optimization constraints in the outer MPC. The method is validated on a kite system in tow test experiments.

Range-inertial estimation for airborne wind energy

An estimation approach is presented for an autonomous tethered kite system for the purpose of airborne wind energy generation. Accurate estimation of the kite state is critical to the performance of automatic flight controllers. We propose an estimation scheme which fuses measurements from range sensing, based on ultra-wideband radios, and inertial readings from an inertial measurement unit. Ranges are measured between a transceiver fixed to the moving kite body and a number of static range beacons scattered on the ground. Estimates are computed using the multiplicative extended Kalman filtering scheme with a sensor-driven kinematic process model using a quaternion representation of the kite attitude. Furthermore, we assume only approximate prior knowledge of the range beacon locations and consider the problem of estimating the kite state and localizing the range beacons simultaneously. We present results of the estimator tested within a simulation environment of an airborne wind energy system and compare performance to an existing estimation scheme based on tether-angles and tether-length measurements.

A stochastic reachability approach to emergency building evacuation

In this paper we introduce a framework for real-time path planning for emergency building evacuation based on the theory of stochastic reachability with random sets. We consider the problem of a single human in a building escaping from a propagating dangerous contamination. We model the human and the hazard as stochastic processes. We formulate the task as a finite horizon stochastic reach-avoid problem in which the objective is to maximise the probability of the human leaving the building while avoiding contact with the hazard within a certain time. We apply dynamic programming to solve the reach-avoid problem and obtain an optimal control policy for the human and a corresponding value for the probability of success. We finally test the results in simulation.

Hybrid Modelling and Reachability on Autonomous RC-Cars

Abstract In this paper we apply reachability analysis to design a controller for an RC-car to drive autonomously on a given circuit. We introduce a hybrid simplification technique to reduce the order of the model; this is crucial for reachability analysis. For a successful implementation on a real system the control problem is divided into two parts: a reachability control strategy is derived for the simplified hybrid model, and a gain-scheduled controller lets the full system track the simplified behaviour. We propose a heuristic algorithm to synthesise a hybrid feedback policy. By considering stochasticity in the model, we improve the performance of the controller which is finally validated on a real physical system.

Control synthesis for stochastic systems given automata specifications defined by stochastic sets

Abstract The problem of control synthesis to maximize the probability of satisfying automata specifications for systems with uncertainty is addressed. Two types of uncertainty are considered; stochasticity in the dynamical system and in the sets defining the specifications. We model the uncertain dynamical sets as stochastic set processes. We show that the optimal control policy can be computed by solving a reachability problem for a hybrid stochastic system, which evolves on product state spaces of the automaton, stochastic sets, and the dynamical system. We derive an approximation to the stochastic set processes to alleviate the complexity of reachability computation. A case study illustrates the framework and the solution approach.

Optimization algorithms for nuclear norm based subspace identification with uniformly spaced frequency domain data

We compare two iterative frequency domain sub-space identification methods using nuclear norm minimization to more commonly used non-iterative methods by means of an artificially created test problem involving very noisy uniformly spaced frequency data. The two corresponding optimization problems are motivated and their first-order algorithmic solutions based on the alternating direction method of multipliers and the dual accelerated gradient-projection method are stated and compared.

Pumping Cycle Kite Power with Twings

Pumping cycle kite power has attracted considerable interest over the last years with several start-ups and research teams investigating the technology. While all these groups produce electrical power with a ground-based generator in a cyclic process, there is no consent about the shape, structure and control of the flying object. In particular the launching and landing strategy has not been settled yet. TwingTec has followed a pragmatic approach focusing on the flying part of the system. The spin-off from Empa and FHNW has developed over the last years in close collaboration with leading research institutes from Switzerland the twing, an acronym for tethered wing. The guiding principle behind the design of the twing was to combine the light weight property of a kite with the aerodynamic properties of a glider plane. Launching and landing was solved by integrating rotors into the structure allowing the twing to hover. Launching, transition into crosswind, autonomous power production, transition into hover and landing has been demonstrated with the current small-scale test system.