Antonios Tsourdos

Contaminant Cloud Boundary Monitoring Using Network of UAV Sensors

In this paper, we describe research work currently being undertaken to detect, model, and track the shape of a contaminant cloud boundary using air borne sensor swarms. A model of the contaminant cloud boundary is first developed using a splinegon, defined by a set of vertices linked by segments of constant curvature. This model is then used in an estimator to predict the evolution of the contaminant cloud. This approach is efficient in that only the vertices and segment curvatures are required to define the cloud boundary, rather than using a distribution function to represent the dispersion density.

Nonlinear Model Predictive Coordinated Standoff Tracking of a Moving Ground Vehicle

This paper proposes a nonlinear model-predictive control framework for coordinated standoff tracking by a pair of unmanned aerial vehicles. The benefit of this approach is to get optimal performance compared with using a decoupled controller structure: heading control for standoff-distance keeping and speed control for phase keeping. The overall controller structure is fully decentralized as each unmanned aerial vehicle optimizes its controller based solely on the future propagation of the pair vehicle states and the target estimates received via communication. This paper uses an acceleration model for sophisticated and realistic target dynamics, which can consider a more reasonable system noise covariance matrix reflecting the target’s motion characteristics. To simplify optimization formulation and decrease computation burden, a new manipulation using the inner product of position vectors of the unmanned aerial vehicles with respect to the target position is proposed for antipodal tracking instead of us...

Robust nonlinear filtering for INS/GPS UAV localization

Unmanned aerial vehicles (UAVs) are increasingly used in military and scientific research. UAVs rely on accurate location information for a variety of purposes including navigation, motion planning and control, and mission completion. UAV INS/GPS localization is generally used based on navigation filter. Extended Kalman filter is largely used to solve the problem of data fusion and localization, however, EKF suffers from the initialization problem and the linearization errors which severely degrade the performance of the UAV localization estimates. In this paper we propose another innovative alternative, which is based on the Hinfin nonlinear filtering to avoid issues linked with classical filtering techniques and getting a significant robustness. This filtering approach is based on the Hinfin robust control theory, results, comparison with the EKF filter and validation on a simulation of a 3D flight scenario are presented.

Direct Intercept Guidance using Differential Geometry Concepts

This paper examines the application of differential geometry to the engagement of both nonmanoeuvring and manoeuvring targets. The kinematics of the engagement for both manoeuvring and nonmanoeuvring target are developed and expressed in differential geometric terms. Two-dimensional geometry is then used to determine the intercept conditions for a straight line target and a constant manoeuvre target. The intercept conditions for both targets are developed for the case when the interceptor missile guides onto a straight line interception. These two cases are shown to have a common set of core conditions such that it enables a unified guidance law to be developed. The guidance law is shown to be globally stable using Lyapunov theory, so that guidance capture is assured for any initial condition. The analysis and guidance law design does not rely on local linearisation and can be shown to produce guidance trajectories that mirror proportional navigation for the straight line interception of a nonmanoeuvring target for which proportional navigation was originally developed. The paper finishes with simulation in two dimensions, illustrating the convergence and solution properties of the approach.

A solution to simultaneous arrival of multiple UAVs using Pythagorean hodograph curves

This paper presents a solution to the problem of simultaneous arrival of a swarm of UAVs by safe and continuously flyable paths. Continuously flyable-paths are generated by satisfying the curvature constraint throughout the path-length. The flyable paths ensure the safety of the UAVs by changing the shape of the flyable-paths by adjusting curvature of the paths. The main idea used in this paper is that a specific type of path is used in the first place for path planning and the shape of the path is varied to meet the multiple constraints. Pythagorean hodograph curves are used for the path planning algorithm. The principle of differential geometry that a planar curve is completely determined by its curvature is used for changing the shape of the path. The multiple constraints are: (i) curvature constraints (ii) minimum-separation-distance and (iii) non-intersection of paths at equal length

3D Path Planning for Multiple UAVs Using Pythagorean Hodograph Curves

This paper presents path planning of multiple UAVs for simultaneous arrival on target. The path planning is divided into phases: firstly design of flyable paths and secondly producing flyable and safe paths achieve the mission. Spatial PH curves are used to produce the paths. The flyable paths are produced by satisfying the maximum curvature and torsion bounds of the UAVs. Latter the flyable paths are tuned to meet the safety conditions and in turn achieve the mission by producing the paths equal in lengths. The curvature bounds and safety conditions and producing paths of equal lengths are all achieved by increasing the curvatures of the paths.

Sensor based robot localisation and navigation: using interval analysis and unscented Kalman filter

Multiple sensor fusion for robot localisation and navigation has attracted a lot of interest in recent years. This paper describes a sensor based navigation approach using an interval analysis (IA) based adaptive mechanism for an unscented Kalman filter (UKF). The robot is equipped with inertial sensors (INS), encoders and ultrasonic sensors. A UKF is used to estimate the robots position using the inertial sensors and encoders. Since the UKF estimates may be affected by bias, drift etc. we propose an adaptive mechanism using IA to correct these defects in estimates. In the presence of landmarks the complementary robot position information from the IA algorithm using ultrasonic sensors is used to estimate and bound the errors in the UKF robot position estimate.

Differential Geometric Guidance Based on the Involute of the Target's Trajectory

This paper presents a novel approach to missile guidance using the differential geometry of curves and not relying on the line of sight information. The target’s trajectory is treated as a smooth curve of known curvature and the new algorithm is based on the involute of the target’s curve. The missile’s trajectory uses the concept of virtual target to generate the correct involute trace. It is shown that the missile is either on the trace immediately or may be able to reach it through an alignment procedure. In general, following the trace requires a three-dimensional maneuver in which the missile flies above the target’s tangent plane. The projection of the three-dimensional trajectory onto the tangent plane coincides with the involute trace, but is traversed in the time-to-go, thus resulting in the intercept. Two air-to-air scenarios of point masses are considered for a maneuvering target of the F-16 fighter class: 1) a two-dimensional engagement with target executing a constant g turn; 2) a three-dimensional engagement with target executing a barrel-roll maneuver. Perfect target information is assumed in simulations. In the first case, intercepts occur both for the involute law and proportional-navigation (PN) guidance; PN based intercepts occur quicker, but the involute-based trajectories are more difficult to evade and always result in a side impact. In the second case, PN fails to intercept the target, while the involute law is successful.

Missile autopilot design using quasi-LPV polynomial eigenstructure assignment

This paper presents the design of a missile autopilot over its flight envelop using quasi-linear parameter-varying polynomial eigenstructure assignment (PEA). The paper describes the extension of PEA to parameter-varying systems using a nonlinear missile model developed by Horton as an example. The autopilot is designed for a single-plane lateral acceleration control and a 5 degree of freedom (DOF) autopilot is also designed. Both lateral acceleration and augmented lateral acceleration outputs are considered. The lateral acceleration autopilot has nonminimum phase characteristics, and it is shown that the quasi-linear parameter-varying PEA approach can handle nonminimum phase systems unlike classic dynamic inversion techniques. Simulation results are presented over fast variations in Mach number and show that the design is robust.

Optimal Impact Angle Control Guidance Law Based on Linearization About Collision Triangle

C ONTROL of missile-target-relative geometry is one of the desired features of guidance in many modern applications. A typical example is to impact a ground target in a direction perpendicular to the tangent plane of the terrain with very high precision both in miss distance and impact angle [1]. Various needs for the maximum warhead effectiveness, and sometimes enhancement of survivability of the missile launch vehicle, in naval applications call for guidance laws that can achieve a specified final direction of approach to the target as well [2]. Also, ensuring a small angle of the missile body relative to the target during the whole engagement process is critical in the case of missiles with strapdown seekers [3]. This necessity of control of terminal engagement geometry has been amajor thrust for much of the researchwork in the area of guidance law design with impact angle constraints. In this Note, a novel method of optimal impact angle control guidance law development based on linear quadratic optimal framework [4] against an arbitrary maneuvering target is presented. Throughout the Note, the missile velocity profile is assumed to be arbitrary. The equation of motion of a missile is often written in terms of the angular variables associated with velocity vectors; in this case, the missile acceleration is computed as the angular rate of its velocity vector multiplied by the magnitude of the velocity, which is directly realizable for aerodynamically controlled missiles. The main problem here is the fact that the kinematics is now nonlinear, defying closed-form solutions of many optimal guidance problems of interest. Thus, it has been common practice to linearize the kinematics (e.g., [5]) or approximate by linear equations [6] to come up with a nice linear quadratic optimal guidance problem. A natural question to follow is then how we linearize the kinematics in a right manner. This question, however, has not been addressed adequately in the literature and linearization has been performed in many cases with the usual assumption of small values of angular variables involved. As a result, the guidance laws often yield poor performance when the associated angular variables get larger. Usually, the collision triangle is defined to be the triangle formed by the initial positions of target and missile, and the intercept point at which the missile hits the target when flown by a straight (with zero effort) line. When no specific requirement on final engagement geometry is posed, the linearization about the usual collision triangle works reasonably well (e.g., [7]; also see [8]). If a specific impact angle between the missile and target velocity vectors is required, however, the usual collision triangle no longer serves as zero-effort collision geometry because the missile trajectory may largely deviate from the collision triangle to satisfy the specific impact angle requirement. This is why some papers just assume before linearization that the end game is initiated with a collision triangle satisfying closely the impact angle requirement [9]. No attempt, however, has been made yet to address specific questions such as what collision trianglewe should be looking for and howwe compute and use it for the linear optimal guidance problem formulation. In this Note, we introduce and use, as the basis of linearization, the perfect (or zero effort) collision triangle for the impact angle control problem, which varies depending on the value of the prescribed impact angle, and solve a linear optimal guidance problem.