Rochdi Merzouki

Robust Monitoring of an Electric Vehicle With Structured and Unstructured Uncertainties

This paper deals with a robust fault-detection and isolation (FDI) technique, which is applied to the traction system of an electric vehicle, in the presence of structured and unstructured uncertainties. Due to the structural and multidomain properties of the bond graph, the generation of a nonlinear model and residuals for the studied system with adaptive thresholds is synthesized. The parameters and structured uncertainties are identified by using a least-square algorithm. A super-twisting observer is used to estimate both unstructured uncertainties and unknown inputs. Cosimulation with real experimental data shows the robustness of the residuals to the considered uncertainties and their sensitivity to the faults.

Bond Graph Model Based on Structural Diagnosability and Recoverability Analysis: Application to Intelligent Autonomous Vehicles

A controlled system can be subjected to faults, which may cause unexpected system dynamics that prevent the system objectives to be achieved. This paper introduces the problem of structural fault-tolerance analysis of actuator, sensor, and plant faults with respect to diagnosability and recoverability conditions. Existing methods in the literature usually investigate the fault tolerance of systems separately from fault diagnosis, even if they are strongly related. The bond graph (BG) tool is an adequate tool for dynamic modeling of complex systems and performing fault detection and isolation. The innovative interest of the proposed approach is to extend the use of this tool for verification of structural recoverability conditions by exploiting the behavioral, structural, and causal properties of the BG tool. The obtained structural technique is then applied to an intelligent autonomous vehicle named RobuCar with decentralized inputs and outputs.

Reconfiguration of Directional Handling of an Autonomous Vehicle

This article concerns reconfiguration of an autonomous vehicle, called RobuCar, with four independently driven wheels and two independently adjustable steering angles. A bond graph model of the system is constructed for generating the Analytical Redundancy Relations (ARRs) which are evaluated with actual measurements to generate residuals and to perform structural fault isolation. Once the fault list is updated in the equipment availability database, an automaton selects the next best option to reconfigure the system such that the given control objectives are achieved. The developed methodology is validated by considering two different fault scenarios.

Force control in a parallel manipulator through virtual foundations

An overwhelming controller provides robustness against uncertain parameters, disturbances and un-modelled dynamics. A simplified inverse dynamics model is used in the present paper to develop an overwhelming controller for a parallel manipulator (a Stewart platform), where the controller accommodates modelling uncertainties such as the simplifications made for development of the inverse model in order to improve the computational efficiency. Such a control strategy leads to good trajectory tracking accuracy in the presence of unknown disturbances. However, in addition to trajectory tracking performance, the controller for a Stewart platform should also be able to control or limit the interaction forces in applications such as robot assisted surgery and low-impact docking. The environmental forces can be accommodated during the interaction period by modulating the impedance at the interface of manipulator and environment through virtual flexible foundations. The positional error induced during the force control phase can be recovered during the free flight or idle phase. While this approach has been used successfully in the past to control serial manipulators, the closed loop kinematic architecture in a parallel manipulator introduces many difficulties. This paper proposes a modified overwhelming control scheme for parallel manipulators with compensations for interaction force control and positional error recovery. Bond graph modelling is used as an integrated model and controller development tool. Force controlled machining on a spherical surface, which is akin to a surgical operation, is considered as an example application of the developed control strategy. The simulation results from the bond graph model of the controlled Stewart platform are presented to demonstrate the performance of the developed hybrid position force controller.

Design and validation of a reconfiguration strategy for a redundantly actuated intelligent autonomous vehicle

This paper deals with the reconfiguration of an intelligent autonomous vehicle that utilizes an automatic navigation method and online supervision system and thus can be used to improve the safety, traffic management and space optimization inside the confined space of a port. The bond graph model of the vehicle’s dynamic system is developed in a modular and hierarchical modelling environment. An over-actuated intelligent autonomous vehicle with redundant actuators has four independent driven wheels, four independent braking wheels and a four-wheel steering system. This vehicle can be safely operated with appropriate control law restructuring even when some of its actuators are unusable due to a fault. For actuator fault detection, analytical redundancy relations, which are constraint relations that describe nominal system behaviour and are written in terms of the measured system variables, are derived from the bond graph model. Analytical redundancy relations are continuously evaluated to generate residual signals and the symptoms in these signals are monitored for actuator fault detection and isolation. Once one or more actuator faults are isolated, the system is reconfigured via the selection of an appropriate operating mode to prevent critical or accidental situations. This procedure is validated by considering a fault scenario with two reconfiguration options.

Neural Networks based approach for inverse kinematic modeling of a Compact Bionic Handling Assistant trunk

A common approach to resolve the problem of inverse kinematics of manipulators is based on the Jacobian matrix. However, depending on the complexity of the system to model the elements of the Jacobian matrices may not be calculated. To overcome intrinsic problems related to Jacobian matrix based methods, a new inverse kinematic modeling approach capable to approximate the inverse kinematics of a class of hyper-redundant continuum robots, namely Compact Bionic Handling Assistant (CBHA) is proposed in the present work. The proposed approach makes use of Multilayer Perceptron (MLP) and Radial Basis Function (RBF) Neural Networks as approximation methods. A validation using a rigid 6 DOF industrial manipulator demonstrates the effectiveness and efficiency of the proposed approach.

Adaptive navigation of an omni-drive autonomous mobile robot in unstructured dynamic environments

One of the challenges of Autonomous Systems navigating in real world is to deal with the large amounts of uncertainties which are inherent in such environment while maintaining stability. Higher order Fuzzy Logic Systems (FLS), such as Interval Type-2 Fuzzy Logic Systems (IT2FLS), that use type-2 fuzzy sets, can model and handle such uncertainties, and give good performances that outperform their Type-1 counterparts. However, the complexity and computational time of type-reduction process which is strongly related to Membership Functions (MFs) structure and the number of fuzzy rules limit their applications to simple cases in real-time. Artificial Potential Field (APF) approach due to its elegant mathematical analysis, simplicity and its possibility to take into account the dynamic of the system is widely used for autonomous mobile robots navigation. However, the potential field introduced exhibits local minima other than at the goal position of the robot. In this paper, a new real-time navigation approach where we combine the APF and IT2FL approaches is developed and implemented for an omnidrive mobile robot navigating in dynamic unstructured environments. The novelty of the approach is the association of IT2FL to APF and the way in which the two approaches are hybridized (rule base size reduction). The experiments carried out on an omnidrive mobile robot named Robotino show the effectiveness of our approach.

Qualitative approach for inverse kinematic modeling of a Compact Bionic Handling Assistant trunk

Compact Bionic Handling Assistant (CBHA) is a continuum manipulator, with pneumatic-based actuation and compliant gripper. This bionic arm is attached to a mobile robot named Robotino. Inspired by the elephants trunk, it can reproduce biological behaviors of trunks, tentacles, or snakes. Unlike rigid link robot manipulators, the development of high performance control algorithm of continuum robot manipulators remains a challenge, particularly due to their complex mechanical design, hyper-redundancy and presence of uncertainties. Numerous studies have been investigated for modeling of such complex systems. Such continuum robots, like the CBHA present a set of nonlinearities and uncertainties, making difficult to build an accurate analytical model, which can be used for control strategies development. Hence, learning approach becomes a suitable tool in such scenarios in order to capture un-modeled nonlinear behaviors of the continuous robots. In this paper, we present a qualitative modeling approach, based on neuronal model of the inverse kinematic of CBHA. A penalty term constraint is added to the inverse objective function into Distal Supervised Learning (DSL) scheme to select one particular inverse model from the redundancy manifold. The inverse kinematic neuronal model is validated by conducting a real-time implementation on a CBHA trunk.

Microscopic traffic dynamics and platoon control based on bond graph modeling

Modeling of traffic dynamic is important for the good traffic management which leads to sustainable transport. Traffic models are classified based on level of details they provide as microscopic models and macroscopic models. A microscopic model of traffic flow describes the behavior of individual vehicle in response to motion of the vehicle preceding it, while, a macroscopic model describes the behavior of the traffic as a whole, but the behavior of individual vehicle is not described. In the present work, we develop a microscopic model of car-following behavior of the vehicles and introduce the sub-microscopic model of traffic, in which the dynamic model of each vehicle is developed, which is not considered in most of the existing microscopic models. Then, a model based control strategy is proposed for the local control of the platoon of the intelligent autonomous vehicles (IAVs). This model based control strategy analytically provides the calculation of necessary effort for the follower IAV to maintain the safe inter-distance with the leader IAV.

Road Vehicle Driving Simulator

This chapter concerns the design of a human-in-the-loop road vehicle driving simulator. The key components of the driving simulator are the vehicle dynamic model, Stewart platform (parallel manipulator), real-time controller, visual animation system, and human–machine interface. In this chapter, the vehicle model developed in Chap. 6 has been used. The Stewart platform is modeled with its hydraulic actuators. Overwhelming control strategy, developed in Chap. 9, is used to control the platform motion so that it mimics the motion of the virtual vehicle (numerical model) driven through the human–machine interface (driver controls and visual feedback). For implementing the overwhelming controller, a lean inverse dynamics model of the platform has been constructed. It is shown that the overwhelming control strategy is an efficient method of system inversion. It further makes the control system robust against parametric uncertainties such as the platform payload. A simple graphics interface has been developed to render the driver display based on the computed vehicle position, orientation, and terrain data. The results from the developed basic simulator are presented. This chapter introduces the fundamental concepts needed to develop practical vehicle simulator systems.