The robotization of agricultural tasks is booming globally, made possible thanks to the development of robotic platforms equipped with agricultural tools. These technological tools (sensors, effectors, etc.) can cause faults which compromise productivity. This chapter presents a review of state of the art of fault diagnosis methods followed by presentation of a Robot Fault Diagnosis Supervision System (S2D2R). S2D2R which is composed of a hybrid diagnostic method (MHD) and a human robot interaction module (MIHR) aims to detect faults as quickly as possible and then inform an operator in order to resolve the problem. Evaluation of the system in real conditions for faults such as wheel locks and in simulation for GPS faults or IMU faults is presented.

• 1 Introduction
• 2 Principles of fault diagnosis
• 3 Methods for fault diagnosis
• 4 Designing a robust fault detection and isolation system
• 5 Hybrid diagnostic method
• 6 Human/robot interaction module
• 7 Implementing the proposed fault detection and isolation system: key components
• 8 Modelling robot movement and potential faults
• 9 Testing fault diagnosis using Kalman filters
• 10 Testing fault diagnosis using deep learning
• 11 Building the hybrid diagnostic method and human/ robot interaction module
• 12 Conclusion and future trends
• 13 Abbreviations
• 14 Acknowledgements
• 15 Where to look for further information
• 16 References