Navigational precision of an autonomous ground vehicle using multiple sensors

Date
2016-03
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: This thesis investigated the navigational precision of an autonomous ground vehicle by fusing different sensors as a means of localization and navigation. Different GPS (Global Positioning System) modules (regular, RTK (Real Time Kinematic) and differential GPS) in conjunction with a digital compass and optical encoders were used as sensors for capturing data regarding the robot’s position. The Arduino Mega 2560 with an 8-bit Atmel microcontroller was used to control all robot functions while MATLAB was used to plot all navigational output data. To implement the localization and navigation, background information had to be gained regarding the functioning of the GPS, motor speed control, fusion of sensor data and algorithms used by the sensors. After this was done all the hardware required to implement navigation was purchased, compatibility between all the components was ensured, housings for the sensors were manufactured, the current platform was modified and a power source sufficient to power everything was selected. Next software was implemented to: control the hardware, capture all the data from the sensors, fuse sensor data, map the environment, establish localization and navigate between waypoints and finally display all the captured data to the user. Before determining the navigational precision of the robot, it needed to be confirmed whether the Piksi RTK GPS could be used as a benchmark for precision comparison of the other sensors. Next case studies tested the navigational precision when: doing multiple runs of the same map, using different complimentary filter values, enabling differential GPS, altering the robot’s speed, introducing wheel slippage, magnetic interference and GPS drift is present and when sensors fail. The high precision with which the Piksi RTK GPS is able to locate the robot gives it the ability to be implemented in various other autonomous and navigation scenarios. Multiple runs of the same map concluded that the consistency of the navigational precision was good enough that data between different runs could be compared. The optimal complimentary filter constant was found experimentally, it was seen that differential GPS resulted in more precise navigation and that the lowest robot speed resulted in the most precise navigational results. Wheel slippage and magnetic interference had a large effect on the robot’s position estimation while GPS drift had little effect. Finally it was seen that any single sensor failure resulted in the robot being unable to navigate. Future work that affects the navigational precision can include: use of different data fusion algorithms, fusion of Piksi RTK GPS data with odometry data, more stable or different robot platform, additional sensor to detect wheel slippage, algorithm to detect magnetic interference and the use of stronger Piksi RTK GPS direct communication antennas.
AFRIKAANS OPSOMMING: Hierdie tesis ondersoek die navigasie presisie van ’n outonome voertuig deur die integrasie van verskillende sensors as ’n wyse van lokalisering en navigasie. Verskillende GPS (globale posisioneringstelsel) modules (gewoon, intyds kinematies en differensiële GPS) in samewerking met ’n digitale kompas en optiese enkodeerders is gebruik as sensors vir die insameling van data aangaande die robot se posisie. Die Arduino Mega 2560 met ’n 8-bis Atmel mikrobeheerder is gebruik om al die robot funksies te beheer terwyl MATLAB gebruik is om die navigasie uitset data te vertoon. Om die lokalisering en navigasie te implementeer het hulle eerstens agtergrond kennis aangaande GPS, motor snelheid beheer, integrasie van sensor data en algoritmes wat deur sensors gebruik word ingesamel. Na afloop daarvan is al die nodige hardeware om navigasie te implementeer aangekoop, versoenbaarheid tussen al die komponente verseker, omhulsels vir die sensors vervaardig, die huidige platform aangepas en daar is besluit op ’n voldoende kragbron om alles aan te dryf. Daarna is die sagteware geïmplementeer wat: al die hardeware beheer, al die data van die sensors ontvang, die sensor data saamsmelt, ’n kaart van die omgewing skep, tussen koördinate navigeer en uiteindelik al die ingesamelde data aan die gebruiker vertoon. Voor hulle kon kyk na die navigasie presisie van die robot, het hulle eers bepaal of die intyds kinematiese GPS gebruik kan word as ’n maatstaf vir die vergelyking van presiesheid van die ander sensors. Volgende is daar deur gevallestudies die navigasie presisie getoets wanneer: herhaaldelike lopies van dieselfde kaart gedoen is, verskillende komplimentêre filter waardes gebruik is, die differensiële GPS aangeskakel is, die robot se snelheid verander is, wielglip ingesluit is, magnetiese inmenging en GPS dryf teenwoordig is asook wanneer enige van die sensors faal. Die hoë presisie waarmee die Piksi intyds kinematiese GPS in staat was om die robot te lokaliseer gee dit die vermoë om in verskeie ander outonome en navigasie verwante situasies geïmplementeer te word. Verskeie lopies van dieselfde kaart het gewys dat die konsekwentheid van die navigasie presisie voldoende was om data tussen verskillende lopies met mekaar te vergelyk. Die optimale komplimentêre filter konstante is eksperimenteel gevind, dit is waargeneem dat differensiële GPS tot meer presiese navigasie gelei het en dat die stadigste robot snelheid die mees presiese navigasie resultate gelewer het. Wielglip en magnetiese inmenging het ’n groot invloed op die robot se posisie vasstelling gehad, terwyl GPS dryf ’n klein effek gehad het. Uiteindelik is waargeneem dat ’n enkel sensor faling veroorsaak het dat die robot nie kan navigeer nie. Toekomstige werk wat die navigasie presisie affekteer kan die volgende insluit: die gebruik van verskillende data integrasie algoritmes, die integrasie van die Piksi RTK GPS data met verplasingsmeter data, meer stabiele of ander platform, addisionele sensor om wielglip waar te neem, algoritme om magnetiese inmenging waar te neem en sterker Piksi RTK GPS kommunikasie antennas.
Description
Thesis (MEng)--Stellenbosch University, 2016.
Keywords
Autonomous navigation, Sensor fusion, Robotics, Autonomous ground vehicle, UCTD
Citation