Sébastien Changey

A mixed extended-unscented filter for attitude estimation with magnetometer sensor

A priori information given by the complete modelling of the ballistic behavior (trajectory, attitude) of a projectile is simplified to give a pertinent reduced evolution model. A new mixed Kalman filter based on Extended and unscented Kalman filter is designed to estimate 3 attitude angles from measures of the magnetic field of the Earth given by a three-axis magnetometer sensor embedded on the projectile. The algorithm has been tested in simulation, using realistic evolution of attitude data for a shot with a 155 mm rotating projectile over a distance of 16 km, with wind and measurement noise. The results show that we can estimate milliradians with non-linear equations and approximations, with good accuracy

Real-Time Attitude-Independent Three-Axis Magnetometer Calibration for Spinning Projectiles: A Sliding Window Approach

This brief deals with attitude-independent three-axis magnetometer calibration in real time. The first part of this note presents a new model used for real-time estimation without any knowledge of the attitude of the projectile. The second part describes the other main contribution, which consists of a new simple and useful filter used for parameters estimation. The latter is formulated within the framework of an extended Kalman filter (EKF), which includes previous measurements in order to increase the performance and stability of the estimator. Finally, numerical simulations support the contribution of the presented algorithm. A good performance is shown through a smart projectile example in a noisy context, without requiring much more computation time than the classical EKF.

A strong tracking extended Kalman observer for projectile attitude and position estimation

In this note we focus on convergence behavior of the Extended Kalman Filter used as a state estimator for projectile attitude and position estimation. We provide first the complete dynamical model, into a non linear state space form, to describe the projectile behavior. Due to strong non linearities and poor observability of the system, very few estimation techniques could be applied, among them the celebrate EKF. The later is, however, very sensitive to bad initializations and small perturbations. The main contribution of this work lies in the use of a modified EKF to assure a strong tracking using magnetometer sensor only. The modified EKF follows from the connection of some instrumental matrices, fixed by the user, and the convergence behavior. Simulation results show the good performances of the proposed approach.

A software based approach for autonomous projectile attitude and position estimation

This work brings a contribution on projectile attitude and position estimation in a noisy context. Due to strong non linearities and poor observability of the dynamical model, describing the projectile behavior, very few estimation techniques could be applied. The main features lie in the use of an EKF based estimator that use a sliding window of output measurements and assures a strong tracking using magnetometer sensor only. Furthermore, we notice that the proposed approach may be applied to a general class of non linear discrete time systems with few computational requirements. High performances, even in the presence of unknown biases, are shown through simulation results.

Projectile attitude and position determination using magnetometer sensor only

A priori information given by the complete modelling of the ballistic behavior (trajectory, attitude) of the projectile is simplified to give a pertinent reduced evolution model. An algorithm based on extended Kalman filters is designed to determinate: • position: x,y,z references in earth frame. • value and direction of the velocity vector; its direction is given by 2 angles (η and θ). • attitude around velocity vector given by 3 angles: roll angle in the range [0, 2π], angle of attack α and side-slip angle β in the range of few milliradians. The estimation is based on the measures of the magnetic field of the earth given by a three-axis magnetometer sensor embedded on the projectile. The algorithm also needs the knowledge of the direction of the earth magnetic fields in the earth frame and aerodynamics coefficients of the projectile. The algorithm has been tested on simulation, using real evolution of attitude data for a shot with a 155 mm rotating projectile over a distance of 16 km, with wind and measurement noise. The results show that we can estimate milliradians with non-linear equations and approximations, with good precision.

Flight Phases With Tests of a Projectile-Drone Hybrid System

A fully functional prototype of a gun-launched micro air vehicle system (GLMAV) is developed for long-distance observation capabilities. A subsonic projectile is launched from an ad hoc tube and transforms into a micro air vehicle (MAV) once arrived over the site to be observed. The ballistic, transient, and operational flight phases of the GLMAV are studied theoretically and experimentally in this paper. Ballistic flight tests prove the stability of the platform in projectile mode and the availability of the inertial measurement unit and GPS measurements for the autopilot. Based on wind-tunnel measurements, a model of the transient phase is built, allowing the design and simulation of control laws for this phase. Indoor and outdoor tests of the system are successfully achieved highlighting the hovering and maneuvering flight capabilities of the GLMAV.

Control of a Gun-Launched MAV for Scene Observation*

Abstract The focus of this paper is on the presentation of a hybrid projectile/MAV concept, namely the Gun-Launched Micro Air Vehicle (GLMAV), with a particular attention on the control strategy used for the hovering flight. The main objective of this concept is the remote observation of hazardous or inaccessible environments, especially for the protection of citizens and vital infrastructures and networks. Prior to its principal use as a standard MAV with two bladed coaxial counter-rotating rotors, the MAV is packaged as a projectile that is launched with the energy delivered by a portable weapon. The first part of the paper gives a description of the GLMAV concept as well as a complete nonlinear mathematical model for hover and near-hover flight. The second part of the paper emphases on nonlinear control strategies for the hover and near-hover flight, particularly backstepping control. The effectiveness of the proposed approaches is illustrated and discussed with some simulations and experimental results.

A Real-Time Sliding Window Filter for Projectile Attitude Estimation

In this paper we propose a sliding window filter for state estimation applied to the problem of spinning projectiles using low cost sensors. The aforementioned filter is formulated in the framework of a standard extended Kalman filter. Behind this particular formulation, the main advantage is that the use of previous measurements can alleviate divergence which can be caused by the highly nonlinear dynamics in the presence of disturbances. We begin the paper with an overview of the kinematics and dynamics relationships of the projectile, in order to show the nonlinearity, which is the motivation behind the proposed filter. Then we propose the formulation of the filter, and finally, simulations show the good performances of the filter especially in presence of disturbances.

STATIC AND DYNAMIC ATTITUDE DECOMPOSITION FOR ESTIMATION WITH MAGNETOMETER SENSOR

Abstract A priori information given by the complete modelling of the ballistic behavior of a projectile is simplified to give a pertinent reduced evolution model. This model is composed of quasi-static and dynamic models. An extended Kalman filter is designed to estimate dynamic part of the 3 attitude angles (roll in [0, 2π], angle of attack and side-slip in the range of few milliradians) from measures of the magnetic field of the earth given by a three-axis magnetometer sensor embedded on the projectile. The algorithm has been tested in simulation, using realistic evolution of attitude data with measurement noise.