Control of Differential Driven Rover by Means of Sensor Fusion
Authors
Thøgersen, Andreas Holst ; Thomsen, Jesper Skovager
Term
4. term
Education
Publication year
2022
Submitted on
2022-06-15
Pages
78
Abstract
This thesis analyzes and controls a differential-drive rover for agricultural use, where steering is achieved by varying the speeds of the wheels. The goal is to determine and control the rover’s position. To estimate position and other states, an extended Kalman filter is designed as a sensor fusion algorithm that combines multiple sensors for higher precision than a single sensor. The sensors are wheel encoders (measure wheel rotation and distance), an inertial measurement unit/IMU (measures motion and rotation), a magnetometer (provides heading relative to Earth’s magnetic field), and GNSS with RTK (satellite positioning with improved accuracy). The estimated states are used to design two controllers: a heading controller that keeps the correct direction at a constant linear velocity, and an inertial-frame controller that drives the rover to predefined positions. The Kalman filter and sensor inputs are handled by a microcontroller running FreeRTOS for real-time operation, while motor control and interfacing are handled by an NVIDIA Jetson Nano running the Robot Operating System (ROS).
Denne afhandling undersøger og styrer en landbrugsrover med differentialstyring, hvor retningen bestemmes ved at variere hjulenes hastigheder. Målet er at bestemme og kontrollere roverens position. For at beregne position og andre tilstande er der udviklet et udvidet Kalman-filter, en sensorfusionsalgoritme der kombinerer flere sensorer for højere præcision end én sensor alene. De anvendte sensorer er hjulenkodere (måler hjulrotation og afstand), en inertimåleenhed/IMU (måler bevægelser og rotation), et magnetometer (giver retning i forhold til jordens magnetfelt) samt GNSS med RTK (satellitpositionering med forbedret nøjagtighed). De beregnede tilstande bruges til at designe to controllere: en heading-controller, der holder den rette kurs ved konstant lineær hastighed, og en controller i den inertielle referenceramme, der får roveren til at nå foruddefinerede positioner. Kalman-filtret og sensorinput behandles af en microcontroller med FreeRTOS for drift i realtid, mens motorstyring og interface håndteres af en NVIDIA Jetson Nano med Robot Operating System (ROS).
[This apstract has been rewritten with the help of AI based on the project's original abstract]