A Hierarchical Architecture for Formation Control ... - ScienceDirect.com

30 downloads 0 Views 179KB Size Report
Hu Zhi-weia*, Liang Jia-honga, Chen Linga, Wu Bingb. aCollege of Mechatronics Engineering and Automation, National University of Defense Technology, ...
Available online at www.sciencedirect.com Available online at www.sciencedirect.com

Procedia Engineering

ProcediaProcedia Engineering 00 (2011) Engineering 29 000–000 (2012) 3846 – 3851 www.elsevier.com/locate/procedia

2012 International Workshop on Information and Electronics Engineering (IWIEE)

A Hierarchical Architecture for Formation Control of MultiUAV Hu Zhi-weia∗, Liang Jia-honga, Chen Linga, Wu Bingb a

College of Mechatronics Engineering and Automation, National University of Defense Technology, Changsha, 410073, China b

The Research Institute of Chemical Defense, Beijing, 102205, China

Abstract Unmanned Aerial Vehicles (UAVs) often fly in formation owing to their lightness, flexibility and versatility, meaning that the distances between individual pairs of UAVs stay fixed. A four-level hybrid architecture is presented in this paper, with mission planning level, formation manage level, formation control level and UAV control level. with this architecture, we can complete a typical scenario of formation control of multi-UAV, which consists of formation reaching, formation keeping, collision avoidance and formation reconfiguration.

© 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Harbin University of Science and Technology Open access under CC BY-NC-ND license. Keywords: Unmanned Aerial Vehicles; Formation Control; Finite State Machine; Information architecture

1. Introduction Unmanned Aerial Vehicles (UAVs) are gaining an increasing interest in several important areas, such as surveillance, rescue, replacing men in hazardous or difficult to reach environment. The cooperative capability of UAVs fleet is a natural extension of a single UAV control problem. A formation of UAVs may constitute a much more effective system than a single vehicle, and the formation flight is a precondition for the cooperation among UAVs[1]. Roughly speaking, there are three approaches to the formation control for multi-agent system reported in the literature, namely leader-following, behavioral, and virtual structures. Each of them has its own strengths and weakness. In [2], the authors introduce a coordination architecture that subsumes leader-

* Corresponding author. Tel.: +086-151-7311-5288. E-mail address: [email protected].

1877-7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. doi:10.1016/j.proeng.2012.01.582

2

HuAuthor Zhi-weiname et al./ /Procedia ProcediaEngineering Engineering00 29(2011) (2012)000–000 3846 – 3851

following, behavioral, and virtual-structural approaches to the multi-agent coordination problem. However, this architecture has limitation in generating geometric formation patterns. Ren and Sorensen propose a unified, distributed formation control architecture that accommodates an arbitrary number of group leaders and arbitrary information flow among vehicles[3]. A approach of hybrid supervisory control of UAVs for a two-dimensional leader-follower formation scenario is presented in [4], which is able to capture internal relations between the path planner and the decision making unit of the UAVs. However, the main problem for the formation control is probably how to define practical architectures, or the dependencies associated with communications and control. A key requirement is that the architecture should be scalable. The number of communication links required for a single agent should not grow linearly with the number of agents in the formation for example. From the viewpoint of control, it is clear that there are tasks at both the level of the whole formation, determining waypoints for a path which the formation should follow, as well as control tasks for the individual agents of the formation, such as maintaining their relative positions, or switching from one formation shape to another[5]. In addition though, we shall consider various change scenarios, such as splitting, merging, and closing ranks. The architectures need to be able to sustain these tasks. Now certainly in formations, there is no single agent exercising control over every other agent. Control tasks in some way have to be handled in a decentralized manner. Our objective in this paper is to develop a hybrid architecture for formation control of multi-UAV, which should be able to accommodate the most requirements from the UAVs formation flight, be scalable for the number of the UAVs and flexible for different control algorithms. 2. Four-Level Formation Control Architecture The architecture of the formation control for multiple UAVs determines the overall performance of the system, such as efficiency, stability, scalability, modularity, etc. Thus, the architecture should be organized in hierarchical layers to accommodate requirements as much as possible. In order to formalize the framework of the UAVs fleet, we propose the four-level hybrid architecture shown in the Figure 1, with the mission planning level, formation manage level, formation control level and UAV control level. The architecture is flexible enough to enable the future integration of additional intelligent attributes at the mission planning level, for instance the introduction of a mission re-planning algorithms.

Fig. 1. Four-level Formation Architecture

3847

3848

Hu Zhi-wei al. / Procedia Engineering 29 (2012) 3846 – 3851 Author name / et Procedia Engineering 00 (2011) 000–000

All UAVs are primarily identical both in the configuration of hardware and software. In the formation control application, one can be assigned as a leader while others are assigned as followers. However, it is important to distinguish between the leader and followers. During the formation flight, mission planning level and formation manage level are activated in the leader, but are not activated in the followers. 2.1. Mission Planner Level The mission planner level is the top of the formation architecture, which permits the supervisor to specify the mission plan, including the formation pattern for the whole team, no-flying-zones, waypoints, etc. It’s character can be the supervisor, such as human beings. Noted that the mission order is preinstalled on every UAV and will be activate in case of failure of the leader. The mission planner level translates a high level representation of the mission into the formation manage level task queue. The mission can be established as a sequence of actions to be executed, for instance: fly to a waypoint and hover there, fly to a waypoint at certain speed, keep the same velocity and heading for a certain period of time, etc. A mission may be completely specified before it is executed but may also be modified, re-planned, or expanded at run time. This feature enables the modification or extension of the mission at run time. Re-planning is particularly important for the future incorporation of obstacle and collision avoidance algorithms. There are several approaches to implement the mission planning level, such as Artificial Intelligence, Finite State Machines, Petri Net theory, etc. 2.2. Formation Manage Level According to the order from the mission planning level, the formation manage level generates a formation from the formation generation algorithm, and assigns roles for other UAVs as followers. Formation pattern can be various geometric types, such as platoon, V-type, triangle, etc. From time to time, a formation may rearrange itself in a minor way, perhaps to remove one member of the formation or to take in another UAV which was out of the formation; it may rearrange itself in a major way, perhaps for obstacle avoidance or predator avoidance, even to the point of splitting, or switching for one type to another type; and it may also merge with another formation[5]. Sensor Information Mission Information

Communication Information

Finite State Machine

Formation Configuration

Reference Trajectory

Leader Followers

Fig. 2. Structure of the formation manage level

The formation manage level can be captured by a finite state machine, which can be further studied through the Discrete Event Systems (DES) supervisory control theory initiated by Ramadge and Wonham[6]. Within the DES framework, we can design the discrete supervisors for formation generation, obstacle avoiding, formation switching, formation splitting, and merging of two or more formations, as shown in Figure 2. Consequently, the manage level will calculate the reference trajectory for all the

3

HuAuthor Zhi-weiname et al./ /Procedia ProcediaEngineering Engineering00 29(2011) (2012)000–000 3846 – 3851

4

UAVs in the formation, and pass the data to the followers through an ad hoc wireless network. 2.3. Formation Control Level The formation control level implements distributed controls such as Formation Reaching Algorithm, Formation Keeping Algorithm, Formation Approaching Algorithm, and Collision Avoidance Algorithm[7]. Starting from an initial state, the UAVs should achieve the desired formation within a finite time under Formation Reaching Algorithm. Then the Formation Keeping Algorithm is active, which negotiates any small disturbance in relative positions and maintains the UAVs in their respective slots. When an UAV is approaching a formation, the Formation Approaching Algorithm takes over the control and guides the UAV to join the formation. However, one thing must be noted is that the safety of UAVs should not be neglected during the formation flight. The collision between UAV will be prevented by Collision Avoidance Algorithm. The algorithms of formation control can be implemented by several ways, including consensus-based method[8], potential field function[9], model predictive control[10], etc. We have developed a formation flight algorithm based on nonlinear predictive control to solve the path following and trajectory tracking problems in [11]. Inspired by the George Vachtsevanos[12], an adaptive mode transition manager is used to coordinate mode selection, switch automatically based on the reference trajectory and the actual state of UAVs, then generate the trajectory for the UAV controller. Consequently, various algorithms can be implemented in the same architecture, and we can compare the performance and effectiveness of different algorithms conveniently as well. Reference Trajectory State Information

Formation Reaching Adaptive Mode Transition

Formation Keeping

Trajectory Generation

UAV Controller

Formation Approaching

Manager

Collision Avoidance

Fig. 3. Adaptive mode transition manager in the formation control level

2.4. UAV Control Level The UAV control level is responsible for the stabilization and navigation of each individual vehicle, steering the UAV to tracking the trajectory generated from the formation control level. In this level, we should take both the kinematics and dynamics of UAV into consideration. However, due to the complex dynamics features, nonlinearities and unstable nature, controller design is a challenge, and model simplifications and linearization may be possible under certain constraints. In principle, UAV controller design follows the well-known system decomposition in inner-loop and outer-loop[13]. The inner-loop enables stabilization of the unstable plant and partially decoupling of control inputs, and the outer-loop generates set points for the inner-loop controller. There are several methods to design the UAV controller in literature, such as PID, fuzzy controller, LQG and MPC.

3849

3850

Hu Zhi-wei al. / Procedia Engineering 29 (2012) 3846 – 3851 Author name / et Procedia Engineering 00 (2011) 000–000

3. Information Architecture The information architecture in the UAVs formation control system plays a vital role in the overall system[1]. The information flow includes communication data, control data and states data. Usually speaking, there are three kinds of information architecture, namely centralized, decentralized and hybrid. The centralized approach can reach global optimization with the cost of burdening the central node both in communication and processing. With decentralized architecture, each UAV node shares the information of its neighborhood nodes, and the overall system is more robust and scalable than that with the centralized architecture in case of node lost and information topology change. With hybrid architecture, the overall multi-agent systems consist of both decentralized and centralized architectures to achieve a trade off in the global performance as well as local burden in communication and computing. In the initial development of coordinate behaviors, the leader acts as the central node to assist the fulfillment of cooperative behavior of the whole formation while the coordination information exchange is realized in the decentralized approach. 4. Conclusion and Future work In this paper, we present a four-level hierarchical architecture for formation control of multi-UAV, including mission planning level, formation manage level, formation control level and UAV control level. With this scalable and flexible architecture, we can design and implement the whole task of the formation control of multi-UAV, from mission planning to the formation flight of unmanned aerial vehicles. The more realistic environment will be built by introduction of the X-Plane, which can simulate the dynamics of various UAVs and the air condition considering the wind effect. References [1]

X. Dong, B.M. Chen, et al. Development of a comprehensive software system for implementing cooperative control of

multiple unmanned aerial vehicles. IEEE International Conference on Control and Automation; 2009, 1629-1634. [2]

R.W. Beard, J. Lawton, F.Y. Hadaegh. A coordination architecture for spacecraft formation control. IEEE Transactions

on Control Systems Technology; 2001, 9(6), 777-790. [3]

W. Ren, N. Sorensen. Distributed coordination architecture for multi-robot formation control. Robotics and Autonomous

Systems; 2008,56, 324-333. [4]

A. Karimoddini, H. Lin, et al. Hybrid formation control of the Unmanned Aerial Vehicles. Mechatronics; 2011,21(5),

886-898. [5]

B.D.O. Anderson, C. Yu, F. Baris. Information Architecture and Control Design for Rigid Formations. Proceedings of

the 26th Chinese Control Conference; 2007, 2-10. [6]

P. Ramadge, W. Wonham. The control of discrete event systems. Proceedings of the IEEE; 1989,77(1) 8-98.

[7]

A. Verma, C.-N. Wu, V. Castelli. UAV formation command and control management. "Unmanned Unlimited" Systems,

Technologies, and Operations-Aerospac; San Diego, California, USA, 2003. [8]

R. Olfati-Saber, J.A. Fax, R.M. Murray. Consensus and cooperation in networked multi-agent systems. Proceedings of

The IEEE; 2007,95(1), 215-233. [9]

J. Wang, X. Wu, Z. Xu. Potential-based obstacle avoidance in formation control. Journal of Control Theory and

Applications; 2008,6(3), 311-316. [10]

A. Richards, J. How. Decentralized model predictive control of cooperating UAVs. The 43rd IEEE Conference on

Decision and Control, 2004, 4286-4291. [11]

Z. Hu, J. Liang, L. Chen, B. Wu. UAVs Formation Flight based on Nonlinear Model Predictive Control. 2011 3rd

International Conference on Computer Design and Applications, Xi'an, Shaanxi, China; 2011, 108-111.

5

HuAuthor Zhi-weiname et al./ /Procedia ProcediaEngineering Engineering00 29(2011) (2012)000–000 3846 – 3851

6 [12]

G. Vachtsevanos, L. Tang, G. Drozeski, L. Gutierrez. From mission planning to flight control of unmanned aerial

vehicles: Strategies and implementation tools. Annual Reviews in Control; 2005,29(1), 101-115. [13]

K.P. Valavanis. Advances in Unmanned Aerial Vehicles: State of the Art and the Road to Autonomy. Springer, 2007.

3851