Trainee Project
Fault tolerant control architecture of bi-directional AGVs in presence of sensors' failures.
2022/04/01 - 2022/08/31
Factory 4.0 is based on the convergence between industry and the digital world and aims to achieve new gains
in productivity and competitiveness. Logistics is a key element of this connected industry. In production
logistics, Automated Guided Vehicles (AGVs) are considered as the most flexible means of transportation,
especially in large-scale manufacturing systems.
An AGV is an automated guided vehicle that moves along guidepaths that can be materialized in different
ways. When AGVs are allowed to move on both directions within a same lane of the guidepath the AGV system
is said to be bidirectional. Bi-directional AGVs systems provide several advantages but are complex to control
because of the several blocking situations that must be avoided. Indeed, when few AGVs are moving on the
same layout, conflicts may result during their movement when they try to enter some sections of their paths
simultaneously. Therefore, it is necessary to supervise the fleet and coordinate the vehicles' movement to
allow them achieve their transportation tasks while avoiding conflicts and deadlocks. A deadlock is an
undesirable situation that causes interruption of traffic and transport missions. This problem is known as
conflict-free routing [1].
The control of the AGVs is generally based on a division of the circuit into independent and non-shareable
zones. The controller supervises the access of vehicles to these zones by ensuring that it will not cause an
undesirable state (imminent or future deadlock). Thus, such a system can be viewed as a discrete event system
(DES) where events are produced as AGVs move along the guidepath. The control methods are therefore
discrete in nature [2]. Most of them are based on informations about the position of AGVs on zones which are
known with certainty. In reality, this position is known thanks to the instrumentation of the system (position
sensors). In practice, sensors are subject to faults. Therefore, their associated events are subject to loss of
observability, making the supervisor inefficient. Nevertheless, the non observability of an event following a
sensor failure is not inevitable. It is sometimes possible to recover this information thanks to the other
operating sensors. This is kown as the property of diagnosability [3].
The purpose of this master project is to study the supervision of an AGV system in presence of sensors'
failures, and to incorporate the diagnosis of some unobservable events into the control architecture (based on
supervision according to Ramadge and Wonham theory [4]) to make it safer (fault tolerance). For this purpose,
we will use diagnostic techniques for DESs based on automata and chronicles [5]. After modeling this
architecture by a DES formalism, we will evaluate its efficiency from the operational and reliability points of

This project focuses on the problem of conflict-free control of a fleet of AGVs, taking into account the
constraints of operational safety and fault tolerance. Unlike previous studies in AGV systems, namely the
operational aspect, we will focus on the dysfunctional aspect and its impact on the efficiency and reliability of
the AGV system's control. This type of study is not addressed in the literature of AGVs where the control is
based on informations available with certainty. In this project, we are interested into the failures of the
instrumentation of the AGVs system which allows to locate the vehicles on the guidepath (positions' sensors).
The supervision of the AGVs system is based on events provided by these sensors, so the failure of the sensors
may be critical for AGVS' safety. Indeed, if the supervisor does not observe some events, it may authorize
forbidden sequences of events. It is therefore important to introduce a diagnostic function that allows, when
possible, to diagnose unobservable events from other operating sensors. This information will be then used by
the supervisor to avoid forbidden state. Thus, the supervision would become more safer and fault tolerant.
The subject of the Master's project is focused on the analysis, control and diagnosis of DESs applied to the AGV
transport system.
AGV, Discrete Event Systems, Supervisory control, Diagnosis, Simulation, Fault-tolerance
Eco-Technic systems engineering