The focus of this project is to design an autonomous vehicle following a desired path by a RTK (Real Time Kinematic) GPS, including designing prototype PCB(Printed Circuit Board) layout, finding sensors and components for the system, designing and simulating the control system and making software.
The system is based on a robotic vehicle with three wheels where two of them are driving. An ARM CORTEX A9 based controller is running the system.
In software, in order to do the sensor fusion to reduce noise influence in the system, a Kalman filter is implemented. System stability is achieved by implementing a PID controller.
There are a lot of Linux versions available, but after 2.6 all versions are similar despite of their version numbers. Thus, we use Linux Kernel 3.2.31 since it is not the latest and it is not very old. Compared with other versions, it contains less reported bugs.
But to be able to have the system work in the Pandaboard, we also need to do some configuration before we really can run our applications. A specific compiler is used in ARM processor instead of X86 C compiler, such as GCC, or Visual C. Thus we use the CodeSourcery ARM GNU/Linux tool chain. It uses a mainline gcc that has stable ARM support.
The study concludes that an autonomous vehicle can reach an accuracy of ±5cm by using the RTK-GPS.
To read this external content in full, download the paper from the author archives .