Nonlinear speed estimation of a GPS-free UAV

Abstract

International audienceIn this article, the problem of robust state observer design for a class of unmanned aerial vehicles (UAVs) is addressed. A prototype four-rotors helicopter robot for indoors and outdoors applications is considered: the drone is not equipped with GPS related devices, so we describe a strategy to estimate its translational velocity vector based on acceleration, angles and angular speeds measurements only. Since the linearised system is non-observable at the equilibrium point, a nonlinear observability verification is performed for persistently exciting trajectories. A global exponential solution to this open problem is provided in the framework of adaptive observation theory when exact measurements are available. A modified observer is presented to enhance velocity estimation robustness in the realistic case of noisy measurements. The results are compared with a classical estimation strategy based on the extended Kalman filter to test the algorithm's performance

Adaptive relative velocity estimation algorithm for autonomous mobile robots using the measurements on acceleration and relative distance

“…For land-based AMRs, the odometry time-derivative methods can provide satisfactory results for velocity estimations. 10 Normally, for aerial AMRs, well-known Kalman filter is utilized to estimate the velocity vector and for removing the Gaussian white noise from positioning data, by fusing the GPS and inertial measurement unit (IMU) measurements. 11,12 In a Kalman filter, the optimal estimation gain matrix is updated online using a derivative Riccati equation.…”

“…15 Later, the translational velocity vector of an autonomous quadrotor is estimated using the measurements on the local translational acceleration, the values for Euler angles in a global frame, and also the values for angular velocities in a local frame. 10 The dependencies on GPS data are completely removed in the proposed linear time-varying velocity estimation. In that solution, the values for mass of the robot and the constant variable for generating thrust force by the propellers are incorporated into the velocity estimation algorithm.…”

“…Relatively, there are fewer investigations for estimating the velocity vector of AMRs, especially for the aerial mobile robots. For land‐based AMRs, the odometry time‐derivative methods can provide satisfactory results for velocity estimations . Normally, for aerial AMRs, well‐known Kalman filter is utilized to estimate the velocity vector and for removing the Gaussian white noise from positioning data, by fusing the GPS and inertial measurement unit (IMU) measurements .…”

“…That solution is further extended and an unbiased state estimation is provided by utilizing a novel switching‐gain technique . Later, the translational velocity vector of an autonomous quadrotor is estimated using the measurements on the local translational acceleration, the values for Euler angles in a global frame, and also the values for angular velocities in a local frame . The dependencies on GPS data are completely removed in the proposed linear time‐varying velocity estimation.…”