Natural Environment Modeling and Fault-Diagnosis for Automated Agricultural Vehicle

This paper presents results for an automatic navigation system for agricultural vehicles. The system uses stereo-vision, inertial sensors and GPS. Special emphasis has been placed on modeling the natural environment in conjunction with a fault-tolerant navigation system. The results are exemplified by an agricultural vehicle following cut grass (swath). It is demonstrated how faults in the system can be detected and diagnosed using state of the art techniques from fault-tolerant literature. Results in performing fault-diagnosis and fault accomodation are presented using real data.