Le 01/06/2020

Fault detection and diagnosis for safety of off-road mobile robots

Presented by Mahmoud Al Masri and Nicolas Tricot, Irstea.

Robotics is becoming increasingly important in a wide variety of sectors. Its development has been favored by the interest for Man to have autonomous entities capable of helping him to perform actions that are difficult or impossible for human being, either because of their painfulness or because of their dangerousness.

The Mobile robotics, one of the sectors of robotics, pursues this objective thanks to technological and scientific advances in different areas such as mechanics, electronics, automation and computer science. These robots navigate autonomously in complex environments and over long periods of time. Their algorithms guarantee a certain robustness while dealing with foreseen situations.

However, in some situations (malfunctioning of a physical component of the robot, loss in GPS signal while executing a trajectory tracking task, etc.), the appearance of a fault may cause the robot to fail in its mission.

In order to avoid this type of situation, real-time fault diagnosis integrating a human operator in its fault detection and fault accommodation loops appears to be a possible and potentially effective solution. The work described is aimed at this objective. It offers, on the one hand, a fault diagnosis supervision system of a mobile robot and, on the other hand, a set of tools for modeling, parameterizing and applying for diagnosis methods to potential fault of a robot.

The first part of the contributions is composed of two modules. The first module is a hybrid diagnostic method that implements several approaches and types of fault diagnosis approach. This method can diagnose a defined list of faults in real time. The second is a human/robot interaction module allowing the integration of the operator in the diagnostic loop by proposing/validating solutions and by correcting, if necessary, the decisions made by the proposed hybrid method. This module integrates a knowledge database and a case-based reasoning algorithm allowing to save the history of the decisions made by the operator, considered expert, in order to improve his performance continuously.

The second part details the approach done off-line and upstream to the implementation of the hybrid diagnostic method. This approach includes the characterization of the considered faults, the definition and the adaptation of the diagnosis methods making it possible to diagnose the identified faults. The adaptation consists of defining the models, adjusting the parameters and, in some cases, collecting training data. From experimental point of view, the theoretical developments are validated on data coming from a real skid-steering robot and from simulations.

Literally, the presentation shows these works. We first introduce the work context. Secondly, we define the problem statement, the proposed hypotheses and describes the state of the literature of fault diagnosis methods. Theoretical developments concerning fault diagnosis supervision system of mobile robotics and the tools leading to its configuration are described in parts that follow. The application results of the theoretical developments in simulation and on a real robot case are shown and discussed after. Finally, a conclusion and the possible perspectives of this work are presented.

This research project was presented in the Scientific Seminar organised by Robagri during the 2019 FIRA. Watch the video:

Categories : #Labs #Robots