Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile ...

0 downloads 0 Views 651KB Size Report
Approaches and Multi-Agent Fuzzy-Based Approaches. Abdelfetah Hentout1, Mohamed Ayoub Messous1,3,. Saliha Oukid2, and Brahim Bouzouia1. 1 Division ...
Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators: Traditional Approaches and Multi-Agent Fuzzy-Based Approaches Abdelfetah Hentout1, Mohamed Ayoub Messous1,3, Saliha Oukid2, and Brahim Bouzouia1 1

Division of Computer-Integrated Manufacturing and Robotics (DPR) Centre for Development of Advanced Technologies (CDTA) BP 17, Baba Hassen, Algiers 16303, Algeria 2 Research Laboratory for the Development of Computerized Systems (LRDSI) 3 University Saad Dahleb of Blida (USDB) BP 270, Blida 09000, Algeria [email protected], [email protected], [email protected]

Abstract. This paper surveys the different control approaches for autonomous mobile manipulators (traditional control approaches and multi-agent heuristicbased control approaches), and focuses mainly on multi-agent fuzzy-based approaches. Directions are discussed and properties of the state-of-the-art in control approaches are classified and compared depending on the techniques used for controlling the robots. The conclusion of the paper presents our point of view about the current state of designing multi-agent fuzzy-based control approaches for such autonomous robots. Keywords: Traditional control approaches, Fuzzy-based control approaches, Multi-agent control, Autonomous mobile manipulators.

1

Introduction

A mobile manipulator consists of one or more manipulators installed on a mobile base. This type of robots can perform most common tasks of robotics that require, at the same time, locomotion and manipulation capabilities. The mobility, offered by the mobile base, allows the robot to significantly expand its reachable workspace and incrementing, thus, its operational capabilities [1]. Mobile manipulators have applications in several fields such as handling tasks, grasping and transporting objects, forestry, mining, construction, etc. Recently, the environments of such robots have migrated from industrial and factory environments to human environments [2]. Such robots are well adapted to human tasks; they are currently present in hospitals, offices, homes for assisting elderly and/or disabled people, etc. [3]. An autonomous robot must be able to operate in complex, not-completely-known and changing worlds using its limited physical and computational resources with a reduced human intervention [4]. Two axes symbolize the needs related to autonomy [5]:

J. Lee et al. (Eds.): ICIRA 2013, Part I, LNAI 8102, pp. 679–692, 2013. © Springer-Verlag Berlin Heidelberg 2013

680 •



A. Hentout et al.

Ability to act with a reduced human intervention: the task provided to the robot is, in general, of a high-level with a distant time horizon. Therefore, the robot must be able to make any decision to progress its task, and achieve its objective. If the introduced task cannot be accomplished directly, the robot must be able to deduce the intermediate operations to achieve this goal. Variability of the environment: this variability disturbs, in permanence, the operations plan of the robot. Therefore, mechanisms must exist to provide robustness1. An autonomous robot must be able to perform its task properly, even if it encounters unexpected situations, without/with a reduced human intervention.

In an autonomous robotic system, two essential elements can be distinguished (i) the robot that consists of a mechanical structure and actuators, and (ii) the robot controller which is the part that provides the intelligence (by processing information and managing information issued from sensors) so that the robot can perform the tasks for which it was designed. Mataric [7] defines the control of a robot as the process of taking information about the environment through the sensors equipping the robot, processing it as necessary in order to make decisions about how to act, and executing actions in the environment. The control architecture is defined by Mataric [6] as a methodology that supplies structure and imposes constraints on the way that robots are controlled. Caselli and colleagues [8] have defined it as an organized set of basic functionalities that the robot can accomplish. The control is the entity that will handle this architecture. The main objective of a control architecture is to simplify the development of basic modules and their composition, necessary to present complex behaviors. On the other hand, a control approach represents mathematical models and tools, and the manner of how these components are exploited in order to ensure the desired behavior. These last years, there has been considerable research effort devoted to autonomous control of mobile manipulators. The works led to different approaches that have been proposed and developed. This paper reviews the various control approaches for such robots (traditional approaches and heuristic-based approaches). After surveying more than one hundred papers of current research, it was found that heuristic approaches (neural network, genetic algorithm, fuzzy logic, neuro-fuzzy, etc.) gave effective and suitable results for autonomous control of the robots evolving in unknown and dynamic environment compared to the classical techniques [9]. This article provided, first, some definitions related to autonomous robots, control and control architectures. Then, it proposes a classification of the main control approaches for autonomous mobile manipulators illustrated with examples from the literature. After that, the paper focuses on multi-agent fuzzy-based control approaches. This concept allows building an architecture by dividing it into a number of agents that are intelligent and capable of handling operations independently. Finally, the article presents our point of view about the current state of designing multi-agent control architectures for such robots. 1

Robustness is the ability to make a correct service in a non-predictable situation of the environment (unexpected obstacle, closed door, etc.). Fault tolerance is the ability to give a correct service regardless of the faults affecting the components of the system (sensor or software error, etc.) [6].

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

2

681

Properties of Control Architectures

The control of a robot with a certain degree of autonomy and complexity, to achieve goals or to react to environment changes, is determined by the organization of its control architecture. This latter must ensure that the robot will accomplish, in real-time, its tasks despite all the encountered situations. The control is required to be reactively fast but, also, thorough while maintaining some properties (stability, robustness, etc.). A robot, to be usable, is required to meet some behavior specifications and design requirements. The main specifications and requirements are given as follows [4] [10] [11] [12]: •

• •









Reactivity to the environment and Adaptability: as the state of the environment changes very quickly and unpredictably, the robot must be able to react to unexpected situations. Moreover, it must be capable to consider external and asynchronous events with time limits that are compatible with the correct, efficient and safe execution of the current task. In addition, the control architecture must be adaptable to switch smoothly and rapidly between the different strategies of control. Furthermore, the robot must be able to change its behavior, while taking into account all the current encountered circumstances. The robot must, therefore, have capacities to recognize situations and make decisions in order to select the most appropriate behavior. Autonomy: as the environment conditions are variable and imperfect, the robot must be able to perform a task through the management of its sub-systems. Fusion of multiple sensors: the narrow accuracy, reliability and applicability of individual sensor must be compensated by integrating the other sensors of the robot. Resolving of multiple goals: in case of robotic tasks, situations requiring conflicting concurrent actions are inevitable. The control architecture must ensure proper means to be able to fulfill those multiple goals. Robustness, Reliability and Fault-tolerance: the robot must handle imperfect inputs, unexpected events and sudden malfunctions. Furthermore, it must be able to operate without failures or performance degradation over a certain period. In addition, the robot must have some hardware and software redundancy, so that corrective strategies can be applied in case where the acting conditions of the robot change. Programmability: a useful robot should be able to achieve multiple tasks described at some abstraction level, instead of only one precise task. Furthermore, the robot must not be simply directed by external stimuli, but a programmable machine able to execute tasks assigned by the operator. Modularity, Flexibility, Expandability, Scalability and Reconfigurability: the control architecture must be divided into smaller sub-systems (individual components or modules) that can be separately and incrementally designed, implemented, tested, debugged and maintained. It must be mentioned that such phases require a long time and experimental robotics requires continuous changes during these phases (design, implementation, etc.). Therefore, flexible architectures are required to allow the design to be guided by the success or failure of individual elements, and designed to be scalable, since new features can

682







3

A. Hentout et al.

be developed and added progressively to the capabilities of the robot, new control algorithms can be tested, etc. without modifying the existing ones. Global reasoning: a control architecture should have a global high-level decisionmaking agent. This latter is responsible of understanding the overall situation and merging the partial available information while taking into account misinterpreted data issued from the sensors. Intelligent and Consistent behavior: the behavior of the robot and its reactions to external stimuli must be correct with the final purpose, guided by the objectives of its main task, and not only determined by the state that it perceives from the environment. Thus, its instant reaction to external events must be based on the general context (state of the robot and its environment) and the task which is performing. Safety: autonomous robots can interact with humans (museum guides, personal service robots, etc.) or intervene in sensitive areas (nuclear centrals, etc.), which introduces additional security measures.

Main Frames Attached to a Mobile Manipulator

The analysis of a mobile manipulator includes the interaction between its mobile base and manipulator. For modeling the robot, the following assumptions were made: • • • •

The floor and the wheels are indeformable. The contact surface between the driving wheel and the floor is assimilated to a point. There is no slipping between the driving wheels and the floor. The manipulator is fixed rigidly on the mobile base.

The mathematical analysis of a mobile manipulator requires defining the main reference frames and transformation matrices as shown in Fig. 1 where [13]: • • • • • • • • • • •

• •

dof (degrees-of-freedom): number of joints of the robot that may vary independently. , , , : the absolute reference frame attached to the environment. , , , : it is connected to the mobile base of the robot. , , , : this frame is attached to the basis of the manipulator. , , , : RE is fixed to the end-effector of the robot. , , , : it is attached to the joint k (k=1, ..., dof) of the manipulator. M TE: this transformation matrix defines RE in RM. B TM: it defines RM in RB. A TB: the transformation matrix defining RB in RA. A TE: this transformation matrix defines RE in RA. EffectorAFin(xEAFin, yEAFin, zEAFin,ψEAFin, θEAFin, ϕEAFin): it is the final situation of the endeffector of the robot given in RA. The first three values represent the position of the end-effector; the last three are its orientation according to Euler angles. BaseFin(xBFin, yBFin, θBFin): it represents the final situation of the mobile base in RA. ConfigurationFin(q1Fin,…, qdofFin): it is the final configuration of the end-effector in RM.

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

683

ZC C

TI

ZE

YC

XC YE

XE

M

u v

E

TC

R

TE

ZM

A

XM

TE

C

TA

I

TA

YM

B

ZA

TM

ZB XB YB

YA A

TB

XA

Fig. 1. Main reference frames attached to a mobile manipulator and the corresponding transformation matrices

The control architectures should be designed to meet all the requirements given previously. The existing research experience seems to have not yet performed a definitive paradigm for the distribution and/or coordination of the functionalities required for all the autonomous mobile manipulators [4]. In the literature, several studies have focused on proposing control architectures for such robots. Each proposition gives a specific way to solve the control problem. Some methods provide a total decoupling between the two sub-systems (mobile base and manipulator). While executing a task that requires the intervention of the two subsystems, the control is carried out sequentially between the two disjoined entities. There exist, also, models for synchronizing the control of the mobile base and that of the manipulator. To organize the operations that a mobile manipulator must perform and structure its internal functions in order to achieve the objectives assigned by the operator, each designer uses his own approach. The next sections review the different control approaches for autonomous mobile manipulators (traditional and multi-agent heuristicbased approaches) and discuss their major properties and characteristics, while focusing, primarily, on multi-agent fuzzy-based approaches.

4

Classical Control Approaches

This first class, considered as classical, is based on the study of mathematical models for both manipulator and mobile base mechanical sub-systems. Controlling a mobile manipulator consists of computing the motion of the manipulator joints and that of the mobile base. For this aim, the study of both Direct (DKM) and Inverse (IKM) Kinematic models of the robot is needed.

684

4.1

A. Hentout et al.

Direct Kinematic Model (DKM)

The DKM is the mathematical model that relates the variables of the joint space and the task space. It is written as follows where f is a non-linear function of the DKM: EffectorAFin=f(BaseFin, ConfigurationFin)

(1)

The computation of the DKM uses different matrices from the various frames (Figure 1). This means that each frame is considered moving relatively to the precedent one. Thus, RB is considered moving relatively to RA, RE is considered moving relatively to RM which moves relatively to RB. Due to the transformation matrix ATE, the situation of the end-effector is given in RA, by the following relation: A

4.2

A

B

M

TE= TB* TM* TE

(2)

Inverse Kinematic Model (IKM)

The IKM expresses the situation of the robot depending on the situation of its endeffector. It is written as follows: –1

(BaseFin, ConfigurationFin)=f (EffectorAFin)

(3)

The IKM of a mobile manipulator, operating in a three-dimension workspace, has as a purpose to compute the generalized coordinates (xBFin, yBFin, θBFin, q1Fin, ..., qdofFin), according to EffectorAFin. Unlike DKM, there are no specific rules to compute the IKM of a robot. However, it is up to the controller designers to choose specific strategies depending on the type of the robots. In most cases, IKM accept an infinite number of solutions since the robot presents geometric redundancy (the number of generalized coordinates is greater than the number of operational coordinates). EffectorAFin is completely defined in RA. However, BaseFin as well as ConfigurationFin can vary significantly as shown in Figure 2. In addition, the non-holonomic constraints of the mobile base can have an important effect on the form of the trajectory required to reach EffectorAFin. These last constraints cannot be resolved by using the IKM. It requires studying the differential-kinematic model and considering velocities [14]. This will further complicate the development of the control approach and requires considerable computation time, especially for real-time control. A solution was proposed in [13] which partially decouples the robotic system. It begins with the motion of the mobile base, while considering the non-holonomic constraints of the mobile base and those related to the environment (obstacles avoidance), so that the robot can move its manipulator into a new situation where it may reach the imposed situation. Afterwards, the IKM of the manipulator is used, for, finally, placing the end-effector in the desired situation. Korayem [15] used the Extended Jacobian Matrix concept to solve the redundancy resolution and nonholonomic constraints. The authors formulated the problem as a Trajectory

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

685

Fig. 2. Variations of BaseFin and its influence on ConfigurationFin to reach EffectorAFin [14]

Constrained non-linear Optimization Problem, and solved it by using the Iterative Linear Programming method, to find the Maximum Dynamic Load of the robot for a given two-end-point task. Nait-Chabane and colleagues [16] presented a global approach based on human-like behavior to help the disabled operator to understand the action of the robot. When the objective is close to the robot, the mobile base and the manipulator move together and the redundancy is used to maximize a Manipulability criterion [2]. 4.3

Discussion of the Classical Approaches

Using the mathematical models to control mobile manipulators produces good and accurate results and offers a fairly exact control for repetitive tasks in controlled and known environments (industrial robotics). In this case, when the robot is required to repeat a trajectory thousands of times, very complicated calculation of these models is done, in most cases, off-line with the ability to optimize time and/or energy expended for performing a task. However, most robotic environments using mobile manipulators are not completely known but, per contra, evolutionary. Moreover, it should be noted that methods used for computing DKMs represented generic rules, whereas IKMs were constructed according to the mechanical structure of the robot. Also, these models do not tolerate any change in the mechanical structure of the robot (malfunction of one or more joints) without adding specific modes for failures treatment. When a fault occurs, the possibility of offering a minimum service until repairing the breakdown is a significant element of Quality Of Service (QoS) [17]. Furthermore, classical approaches have the disadvantage of computing time which is quite important depending on high number dof of the robot, especially in frequently-changed and notcompletely-known environments. Finally, according to Brooks [10], traditional robotics seems unable to deliver real-time performances in a dynamic world.

5

Multi-Agent Heuristic-Based Approaches

When working with complex problems with large dimensions, the resolution of control problems for mobile manipulators is very difficult using traditional mathematical models. Several approaches have described the process allowing the end-effector of

686

A. Hentout et al.

such a robot to reach a position in its workspace, without using IKM or differentialequation solvers [18]. 5.1

Multi-Agent Approaches

Multi-agent and Distributed Artificial Intelligence (DAI) techniques offer simple solutions [19]. Each joint is implemented as an agent. Every agent tries to align the position of the end-effector with that of the target, while being independent of the motions and positions of the other joints, and with no prior knowledge of the actions of the other agents. By acting recursively, the other agents, controlling the other joints, try to do the same job. A global behavior can emerge, consequently, from all the local agents which satisfy the desired objective. Among the principal works done in this area, we find that developed by Delarue and colleagues [17]. They have built a control architecture adapted to the robotics of service (the ARPH robot), by using a multi-agent approach. This latter consists of two parts (i) the first concerns the manipulator and (ii) the second is assigned to the mobile base. This architecture has exploited the redundancy of the robot to offer tolerance to certain faults and breakdowns. The authors affirm, in their works, that each agent acts independently without any inter-synchronization. Each agent computes, first, the current position of the end-effector and, then, tries to match this position with the purpose to reach. By an approach of type “Virtual movement-Check”, each agent tends, with small local movements, to fulfill this constraint. If the imposed goal is outside the accessible workspace, each agent will align its joint with that objective. A global emergent behavior is then expected, which is to bring the robot to a fully-extended position, trying to get as close as possible to the objective. Fig. 3 shows an extension of this principle with a manipulator controlled by four agents and a mobile base controlled by a single agent.

Agent Base

Fig. 3. Agents associated to a mobile manipulator as proposed in [19]

To solve a complex problem, a Multi-Agent System (MAS) can emerge global behaviors by using several agents. Each of these latter has a knowledge and a limited actions ability. Moreover, fuzzy logic, neural networks and/or genetic algorithms associated with MAS, can provide high-level control for such complex systems (mobile manipulators, etc.) [20] [21]. 5.2

Multi-Agent Fuzzy-Based Approaches

Fuzzy logic is a mathematical formulation that copes with uncertainty in information [22]. Fuzzy control has proven to be a successful approach to many complex nonlinear systems or even non-analytic ones. It has been suggested as an alternative approach to conventional or classical control techniques in many situations [23]. Such a control is characterized by the use of linguistic rules to manipulate and implement

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

687

human knowledge in control systems so as to handle the uncertainty present in the environment [24]. The model presented by Agah and Tanie [25] represents, to our knowledge, the first control architecture which is based on techniques of fuzzy logic with multiple distributed agents. The authors proposed a new approach in which the different agents cooperate and contend to control the system. By utilizing multiple software agents, they showed that the control design of the global system can be divided into small modules, in terms of the task, environment, or system status. These agents cooperate with each other to transform the input information (from sensors) into outputs information (control signals). Each agent was implemented using fuzzy-logic rules, and the agents were able to access all the available input information and, afterwards, suggest actions based on their knowledge and autonomy. The main controller (supervisor) evaluates the proposals of the contending agents according to their relevance, confidence and effect. Then, it selects the winner action that will be executed by the system. Finally, the winner agent controls the whole system alone. This methodology was applied to control an inverted pendulum balanced by a robotic system, while being perturbed by a human force (similar to the system shown in Fig. 4). The practicality and feasibility of the proposed concepts were verified by applying the system to control a mobile manipulator in two-dimension handing over an object to an interacting human operator. Three fuzzy-logic agents were developed for the robot pendulum balancing task (i) Flexed agent, (ii) Mid-range agent and (iii) Extended agent. For low values of robot length, Flexed agent is more relevant; for middle values, Mid-range agent is more relevant; and for high values, Extended agent is more relevant. Each agent is implemented as a fuzzy-logic controller with 21 rules. In these rules, the measured parameters are (i) the rotation angle of the pendulum, (ii) its rotation angle velocity, and (iii) the robot length. The control parameters determined by the system are (i) the magnitude and (ii) the direction of the force to be applied by the robot. The confidence of each action proposed by an agent is based on the fuzzy membership values of the sensory values (measure parameters). In this work, the agents were defined according to the task to be done by the robot. Furthermore, one agent was active at a time. In such a scheme, it is more complicated to design and improve the control architecture for other systems. 5.3

Multi-Agent Fuzzy-Based Approaches Merged with other Heuristic

The most important drawbacks using fuzzy-logic to solve control problems are given hereafter [26]. These entire disadvantages have led to the idea of merging other heuristic approaches such as neural networks, genetic algorithms, etc. and the fuzzy technology. • •



Standard and systematic method does not exist for the transformation of the human knowledge or experience into the rule base of a fuzzy inference system. There is no general procedure for choosing the optimal number of rules, since a large number of factors are involved in the decision (performance of the controller, efficiency of computation, human operator behavior, the choice of linguistic variables, etc.). In designing fuzzy-logic expert systems, a great deal of care and effort is required to obtain the rules.

688 • • • •

A. Hentout et al.

Even when human operators exist, their knowledge is often incomplete and episodic, rather than systematic. It is not possible to show the stability of the controlled system, since the model of the system is not known. There is no guarantee that rules are coherent and a no mismatch exists between the rules. Because of the complex operations (fuzzification and particularly defuzzification), computing time could be very long.

The work presented by Erden, Leblebicioglu and Halici in [27] fits into this context (merging fuzzy-logic and genetic algorithms). It describes a MAS approach to a service mobile manipulator that interacts with human during an object delivery and hand-over task in two dimensions. The required operations to achieve the overall objective are distributed between agents and each one fulfills its specific operation. The agents of the system are controlled using fuzzy control. The functions of the fuzzy controllers are tuned by using genetic algorithms. The agents are identified as the Base, Shoulder, and Elbow of the robot. Each agent controls the movement of the junction referred by the name (Figure 4): •





Base agent: this first agent controls the displacement velocity of the mobile base (vb) depending on horizontal and vertical distances dx and dy between the endeffector of the robot and the hand of the human operator. Shoulder agent: this agent controls the angular velocity of the shoulder (ws) according to dx and dy distances and angle β. To compute this latter, the Shoulder agent needs information on the state of the Elbow agent (θe). Elbow agent: this last agent controls the angular velocity of the elbow (we) function of dx and dy distances and angle α.

In this work, an additional amelioration of the control has been performed by considering a criterion of improvement, with a focus on membership functions of the fuzzy controllers. This criterion is computed with a certain fitness function, which is a combination of (i) the distance traveled by the effector, (ii) the duration of the maneuver and (iii) the energy.

Shoulder w d

Elbow

Human

Base agent x=0

d

Fig. 4. Different agents associated to the system as proposed in [27]

v

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

5.4

689

Discussion of the Multi-Agent Approaches

In contrast to classical control approaches, multi-agent approaches offer methods that use the agent paradigm by proposing a decomposition of the robot control into a set of distinct agents. The MAS approaches benefit of all the advantages of distributed problem solving. The MAS perspective made it possible to consider the architecture as a compound of simpler modules, which gave way to easier design of the whole system. In addition, the need for massy mathematical models, IKM and differential-equationsolutions is overcome. Therefore, there is a considerable decrease in design effort and computation time compared to classical approaches. Moreover, with such a usage of MAS, the control architecture is more flexible to be applied to any robot [27]. The design phase of the development of a control architecture is the most strategic one. However, it is necessary to understand and to identify the needs to design and implement the approach in a proper manner. With the increasing complexity of architectures, using a development methodology is very necessary. Nevertheless, the absence of such a methodology covering the whole life cycle of a MAS makes the task very difficult and complicated [28]. The multi-agent approach poses, moreover, the problem of the management and control of agents and their shared resources. Another constraint that should not be overlooked when designing such a control system is the lack of information. This lack is mainly due to the measurement errors delivered by the physical sensors of the robot. Multi-agent heuristic-based control approaches don’t require a precise mathematical model of the robot to be controlled. However, if the model exists, it can be used for simulating and testing the control strategy. The main advantages of such controllers are given as follows [26] [29] [30]: • • • • •

No need to have a mathematical model of the robot (information is not available, information is not complete, the process is too complex, etc.). It is possible to implement expert-human knowledge and experience using comprehensible linguistic rules. Thanks to dedicated processors, it is possible to control fast processes. Such techniques allow developing robust and smooth controllers starting from heuristic knowledge and qualitative models. Such techniques allow considering imprecise, vague and unreliable information; and integrating symbolic reasoning and numeric processing in the same framework.

It should be mentioned that the aim of all the works presented in this section was to test the performances of the multi-agent heuristic-based design of control approaches in simulation. Moreover, all the studies were applied on a simple case of a service mobile manipulator interacting with human operators during hand-over tasks in two dimensions. Unfortunately, no works were done on real physical robots in three dimensions.

6

Conclusion

More than one hundred works have been surveyed in this paper covering the field of developing control approaches for autonomous mobile manipulators. We have tried to

690

A. Hentout et al.

bring together the major applications of multi-agent paradigm in the literature and come to conclusions about the nature of research in this field. The survey reveals that the application of heuristic methods, and especially fuzzy-based techniques merged with other heuristic approaches (neural networks, genetic algorithms, etc.), has increased due to their success in coping with the problems of combinatorial explosion. This article presented a comparison between two different control approaches for autonomous mobile manipulators. The classical approaches use the mathematical models to control the robot. On the other hand, the second class is based on the DAI techniques for problem solving. Here, the control is mainly distributed on several concurrent agents, combining reactive and deliberative capacities. This class provides an alternative to use mathematical models to control such robots. It offers results that approximate human behaviors, and improves tolerance to certain faults and mechanical failures. Throughout the article, we have discussed some recent research which proposed interesting models for controlling mobile manipulators. As remarked by Simmons [31], in many cases different approaches can be used for the same tasks. Fuzzy control can be adopted to coordinate the various system behaviors in response to the environment [32]. The combined use of fuzzy control with multi-agent approach, also, has another advantage of having a distributed fuzzy control system with smaller fuzzy sub-systems instead of one big centralized fuzzy system. Therefore, adopting an architecture/approach is a technological problem where the designer must consider the required degrees of reactivity, intelligent behavior and the related implementation cost. According to the type of robots and the necessary level of autonomy, different recommendations about the appropriate approach can be found in the literature. Typically, the differences are the ease with which approaches can be developed, and the efficiency with which tasks can be achieved [4]. The future works are dedicated to develop our own architecture for autonomous control of mobile manipulators. For this purpose, a multi-agent fuzzy-based approach is responsible for controlling the robot while sharing the control of the heterogeneous sub-systems.

References 1. Sugar, T., Kumar, V.: Decentralized Control of Cooperating Mobile Manipulators. In: The International Conference on Robotics and Automation (ICRA 1998), Leuven, Belgium, pp. 2916–2921 (1998) 2. Nagatani, K., Hirayama, T., Gofuku, A., Tanaka, Y.: Motion Planning for Mobile Manipulator with Keeping Manipulability. In: The International Conference on Intelligent Robots and Systems (IROS 2002), Switzerland (2002) 3. Alfaro, C., Ribeiro, M.I., Lima, P.: Smooth Local Path Planning for a Mobile Manipulator. In: The 4th Portuguese Robotics Festival Robotica 2004, Protugal (2004) 4. Medeiros, A.A.D.: A Survey of Control Architectures for Autonomous Mobile Robots. Journal of the Brazilian Computer Society 4(3) Campinas (April 1998) 5. Lussier, B.: Tolérance aux fautes dans les systèmes autonomes. Ph.D. Thesis in Computer Science System, Polytechnic National Institute of Toulouse, France (April 2007) 6. Mataric, M.J.: Behavior-Based Robotics as a Tool for Synthesis of Artificial Behavior and Analysis of Natural Behavior. Trends in Cognitive Science 2(3), 82–87 (1998) 7. Mataric, M.J.: Situated Robotics. Encyclopaedia of cognitive science, pp. 25–30. London Nature Publisher Group, Macmillan Reference (2002)

Multi-Agent Fuzzy-Based Control Architecture for Autonomous Mobile Manipulators

691

8. Caselli, S., Fantini, E., Monica, F., Occhi, P., Reggiani, M.: Toward a Mobile Manipulator Service Robot for Human Assistance. In: Robocare Workshop, Roma, Italy (October 30, 2003) 9. Mohanty, P.K., Parhi, D.R.: Controlling the Motion of an Autonomous Mobile Robot Using Various Techniques: a Review. Journal of Advance Mechanical Engineering 1(1), 24–39 (2013) 10. Brooks, R.A.: A Robust Layered Control System for a Mobile Robot. IEEE Journal of Robotics and Automation RA 2(1), 14–23 (1986) 11. Ferraz de Camargo, R.: Architecture matérielle et logicielle pour le contrôle d’exécution d’un robot mobile autonome. Ph.D. Thesis in Robotics, University of Toulouse III, France (1991) 12. Goût, J.: Intégration de la Planification Temporelle dans l’Architecture Décisionnelle d’un Robot Autonome. Ph. D. Thesis in Computer Science, University of Paul Sabatier, Toulouse, France (1999) 13. Hentout, A., Bouzouia, B., Akli, I., Toumi, R.: Mobile Manipulation: A Case Study. In: Lazinica, A., Kawai, H. (eds.) Robot Manipulators, New Achievements, pp. 145–167 (April 2010) 14. Akli, I., Haddad, M., Bouzouia, B., Achour, N.: Trajectory Generation for Operational Task Execution With Manipulability Analysis. In: The International Conference on Advanced Robotics (ICAR 2011). Tallinn University of Technology, Estonia (2011) 15. Korayem, M.H., Ghariblu, H., Basu, A.: Maximum Allowable Load of Mobile Manipulators for Two Given end Points of End-effector. International Journal of Advanced Manufacturing Technologies 24, 743–775 (2004) 16. Nait-Chabane, K., Delarue, S., Hoppenot, P., Colle, E.: Strategy of Approach for Seizure of an Assistive Mobile Manipulator. Robotics and Autonomous Systems 57(2), 222–235 (2009) 17. Delarue, S., Hoppenot, P., Colle, E.: A Multi Agent Controller for a Mobile Arm Manipulator. In: The 4th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2007), Angers, France, May 9-12 (2007) 18. Duhaut, D.: Distributed Algorithm For High Control Robotics Structures. In: International Conference on Artificial Intelligence, vol. 1, pp. 45–50 (1999) 19. Colle, E., Nait Chabane, K., Delarue, S., Hoppenot, P.: ARPH: Comparaison d’une méthode classique et d’une méthode utilisant la coopération hommemachine pour exploiter la redondance de l’assistant robotisé. In: The 4th Conference HANDICAP 2006, France, June 7-9 (2006) 20. Tournassoud, P.: Planification et contrôle en robotique: Application aux robots mobiles et manipulateurs. Hermes Sciences Publications (September 1992) 21. Guessoum, Z.: A Hybrid Agent Model: a Reactive and Cognitive Behavior. In: The 3rd International Symposium on Autonomous Decentralized Systems (ISADS 1997), Germany, pp. 25–32 (April 1997) 22. Klir, G.J., Folger, T.A.: Fuzzy Sets, Uncertainty, and Information. Prentice-Hall, Upper Saddle River (1992) 23. Precup, R.-E., Hellendoorn, H.: A Survey on Industrial Applications of Fuzzy Control. Computers in Industry 62(1), 213–226 (2011) 24. Passino, K.M., Yurkovich, S.: “Fuzzy Control”. Addison-Wesley, Menlo Park (1998) 25. Agah, A., Tanie, K.: Fuzzy-logic Controller Design Utilizing Multiple Contending Software Agents. Fuzzy sets and systems 106, 121–130 (1999)

692

A. Hentout et al.

26. Godjevac, J.: Comparative study of fuzzy control, neural network control and neuro-fuzzy control., Technical report n° 103/95, Federal Polytechnic School of Lausanne, Computer Science Department (February 1995) 27. Erden, M.S., Leblebicioglu, K., Halici, U.: Multi-agent System-Based Fuzzy Controller Design with Genetic Tuning for a Mobile Manipulator Robot in the Hand Over Task. Journal of Intelligent and Robotic Systems 39(3), 287–306 (2004) 28. Hentout, A., Bouzouia, B., Toukal, Z.: Modeling of Agent-based Architecture for Driving Mobile Manipulator Robots. In: The International Conference on Distributed HumanMachine Systems (DHMS 2008), Greece, March 09-12 (2008) 29. Singh, M.K., Parhi, D.R., Bhowmik, S., Kashyap, S.K.: Intelligent Controller for Mobile Robot: Fuzzy Logic Approach. In: The 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG), Goa, India, October 1-6 (2008) 30. Rashid, R., Elamvazuthi, I., Begam, M., Arrofiq, M.: Fuzzy-based Navigation and Control of a Non-Holonomic Mobile Robot. Journal of computing 2(3), 130–137 (2010) 31. Simmons, R.G.: Structured control for autonomous robots. IEEE Transactions on Robotics and Automation 10(1), 34–43 (1994) 32. Vadakkepat, P., Miin, O.C., Peng, X., Lee, T.H.: Fuzzy Behavior-Based Control of Mobile Robots. IEEE Transactions on Fuzzy Systems 12(4), 559–564 (2004)