Navigation Solution for Marine Applications using MEMS-based Sensors and GPS
Author
Eskildsen, Bjørn
Term
10. term
Publication year
2012
Submitted on
2012-02-17
Pages
110
Abstract
Specialet undersøger, hvordan man kan beregne position, hastighed og orientering (attitude) ved at kombinere data fra inertiessensorer, magnetometer og GPS. Arbejdet er udført i samarbejde med CDL – inertial engineering, og den primære hardware er MiniSense2 (MS2) fra CDL, som indeholder lavpris, MEMS-baserede 3-aksede gyroskoper, accelerometre og magnetometre. Metoden bygger på en løst koblet integration mellem MS2-sensorerne og GPS, hvor et modelbaseret udvidet Kalman-filter bruges til at sammenflette sensordata og et fysisk bevægelsesmodel for at estimere de ønskede tilstande. Der udledes bevægelsesligninger for bevægelse på den roterende jord og sensor- og feltmodeller, herunder antagelse om additive bias i MS2-sensorerne samt modeller for jordens tyngdefelt og magnetfelt. Da filteret er et udvidet Kalman-filter, behandles også en lineær små-perturbationsmodel. Løsningen evalueres i to testopsætninger. I en simuleret test bruges et punktsystem, der påvirkes af kræfter og momenter, så det udfører skibslignende bevægelser; her er de sande tilstande kendt, og sensormålinger forstyrres med støj og bias. Der indlægges desuden simulerede GPS-udfald for at teste robusthed. I en virkelig test indsamles data under en bådtur til søs fra MS2 og GPS, og der findes reference for orientering fra et højkvalitets Attitude and Heading Determination System (AHDS); også her indgår simulerede GPS-udfald. I simuleringen estimerer navigationsløsningen korrekt alle sensorbias, og ydeevnen er acceptabel både med og uden GPS-dækning, hvilket bekræfter de grundlæggende principper i filterdesignet. I felttesten er den samlede ydeevne derimod ikke acceptabel, selv om gyrobias tilsyneladende estimeres. Estimerede magnetometerbias peger på, at ikke-medtagne effekter påvirker målingerne. Som mulige årsager peges der på en udeladt lever arm mellem GPS og MS2 samt en “soft iron”-effekt, der kan forvride magnetfeltmålinger.
This thesis examines how to estimate position, velocity, and orientation (attitude) by combining data from inertial sensors, a magnetometer, and GPS. The work was carried out in collaboration with CDL – inertial engineering, using CDL’s MiniSense2 (MS2) as the main hardware platform. The MS2 includes low-cost, MEMS-based 3‑axis gyroscopes, accelerometers, and magnetometers. The navigation solution uses a loosely coupled integration of MS2 sensors and GPS, and applies a model-based extended Kalman filter to fuse sensor readings with a physics-based motion model to estimate the states of interest. Differential equations for motion on the rotating Earth are derived, along with sensor and field models, including additive sensor biases, and models of the Earth’s gravity and magnetic fields. Because an extended Kalman filter is used, a linear small-perturbation model is also addressed. The design is evaluated in two tests. In a simulation, a point-mass model is driven by forces and moments to produce ship-like oscillations; ground truth is known, and synthetic measurement noise and biases are added. Simulated GPS outages are included to test robustness. In a real-world sea trial, data were collected from an MS2 and a GPS, with a high-grade Attitude and Heading Determination System (AHDS) providing reference attitude; simulated GPS outages are also included. In the simulation, the navigation solution correctly estimates all sensor biases and performs acceptably with and without GPS, supporting the basic filter concept. In the real data, overall performance is not acceptable, even though gyroscope biases appear to be estimated. The magnetometer bias estimates indicate unmodeled effects in the measurements. Likely causes mentioned are an unmodeled lever arm between the GPS and MS2 and a “soft iron” magnetic effect that distorts magnetometer readings.
[This abstract was generated with the help of AI]
Documents