Systems, Control and Signal Processing

Cooperative Control of Multi-Agent Systems 

Analysis and Control of Communication Networks

Robust Adaptive Nonlinear Control System Designs

Development of Hardware/Software Architecture for Autonomous Unmanned Vehicles

Nonlinear Control for Electric Motors

Development of an Ultra-Accurate High-Speed Six DOF Manipulator and Other Robotic Systems

Decentralized Control of Nonlinear Large-Scale Interconnected Systems

Cooperative Control of Multi-Agent Systems

Participating Faculty: Z. P. Jiang (zjiang@control.poly.edu)
Research supported by the NSF

Recently, the rapid advances in communication, computation and miniaturization technologies spur the keen interests of building more and more sophisticated man-made multi-agent systems in numerous industrial and military fields, including multi-vehicle search, rescue and monitoring, traffic control and management, information systems, network security systems, object recognition systems, cooperative decision making mechanism, path planning, task assignment and production scheduling. Compared with conventional single agent systems, multi-agent systems can be made to have more efficiency and robustness and less cost. In the coordination and control of multi-agent systems, decentralized strategies are desirable since the agents often can only get local information from the neighboring agents they can sense and communicate with.

For a multi-agent system to achieve common group objectives or collectively react to unexpected external changes, some information state (e.g. moving direction) of all the agents sometimes need to reach a common value, or consensus. We have contributed some novel solutions leading to relaxation of some existing conditions for the consensus by the well-known averaging consensus protocols. Also our research has derived sufficient conditions for the consensus by a class of non-averaging protocols. In addition, we have presented some novel protocols for the heading consensus of multi-vehicle systems. Our results on the analysis and design of consensus protocols have been integrated into a control strategy that realizes pattern preserving path following of a multiple nonholonomic vehicle team.

[1] Q. Li and Z. P. Jiang, “Relaxed conditions for consensus in multi-agent coordination,” Journal of System Science and Complexity, vol. 21, no. 3, pp. 347-361, 2008.

[2] Q. Li and Z. P. Jiang, “Two decentralized heading consensus algorithms for nonlinear multi-agent systems,” Asian Journal of Control, special issue on “Collective Behavior and Control of Multi-Agent Systems”, vol. 10, no. 2, pp. 187-200, 2008.

[3] Q. Li and Z. P. Jiang, “Global analysis of multi-agent systems based on vicsek's model,” IEEE Trans. on Automatic Control, accepted.

[4] Q. Li and Z. P. Jiang, “Pattern preserving path following of unicycle teams with communication delays,” submitted for publication.

Analysis and Control of Communication Networks

Participating Faculty: Z. P. Jiang (zjiang@control.poly.edu) and S. Panwar (panwar@catt.poly.edu)
Web Sites: http://eeweb.poly.edu/faculty/jiang, http://catt.poly.edu/CATT/
Research supported in part by the CATT, NSF and AFOSR.

In the past decade, the analysis and control of communication networks attracts great research interests. As an example, optimization tools have been successfully applied to flow control problems in Internet. The design objective of optimization-based flow control is to maximize the overall network utility function, subject to the constraints of link capacities. In [5], we solve this problem using a modified Aitken Extrapolation algorithm. Furthermore, we perform convergence speed analysis for optimization-based flow control algorithms.

Except for the utility and cost optimization problem, recently, great attention is paid to Internet congestion control since congestion causes packet loss and results in network under-utilization (Figure 1). We notice that, compared with the large body of work on stability analysis for existing congestion control schemes, the synthesis of nonlinear controllers with improved performance has not received enough attention. We focus on designing new AQM schemes to stabilize the nonlinear network model, with particular interest in output feedback design, owing to the advantage of only measuring limited output information--buffer queue length. Previously, the output feedback design for AQM schemes is based on linearized model with known network parameters. With the help of Lyapunov design technique, adaptive feedback linearization and filtering techniques, we design a new, nonlinear controller to achieve both asymptotic stabilization and adaptation to unknown parameters [2]. The output feedback solution represents a nontrivial application of modern nonlinear control theory. We also believe that the design in [2] should provide benefits to future networking technical developments and serve as guidance for distributed protocol designs.

There have been continued interests in applying modern control theory to systematically address new challenges in large-scale networks. Currently, many existing work on controlling communication networks are based upon linear control techniques. Hard nonlinearities, such as saturation caused by capacity constraints, have not been thoroughly addressed, especially when designing control schemes for large-scale networks. In this project, we apply nonlinear control theory to cope with saturation constraints and nonlinear disturbances. In [1], [3] and [4], the constrained regulation of a class of network systems is studied. Explicit conditions are identified under which the problem of asymptotic regulation against unknown traffic interferences is solvable, with control and state saturations. We achieve either asymptotic or practical regulation for a single-node system in [1]. We also propose decentralized, discontinuous control laws to achieve asymptotic regulation of cascaded nodes and large-scale networks in [3], [4]. Our research demonstrates that tools from nonlinear system theory can play an important role in tackling “hard nonlinearities” and “unknown disturbances” for controlling communication networks. The control architecture for a single-node system is shown in Figure 2.

Another line of our research is the analysis and design of wireless ad hoc broadcasting protocols. Ad hoc network is composed of a set of self-organized users that agree to relay packets for each other. Different from conventional cellular wireless systems, ad hoc networks have no fixed infrastructure and central administration. Furthermore, each user can move randomly and the topology changes occur frequently. In such distributed and dynamic networks, broadcasting is widely used to distribute small control packets such as route request packets and warning packets. Numerous broadcasting protocols are proposed in literature to minimize overhead and maximize reception ratio. Different from conventional simulation based analysis methods, we theoretically analyze these protocols. Our results in [6], [7], [8] reveal the relation between broadcasting efficiency and network parameters. Furthermore, we have proposed a mobility sensitive mechanism to improve protocol performance in highly-mobile environment [9]. Broadcasting protocols based on the proposed mechanism are adaptive to nodal movement and hence reduce packet loss rate due to mobility.

[1] Y. Fan, Z.P.Jiang and H. Zhang, Network flow control under capacity constraints: a case study, Systems & Control Letters, Vol. 55, No. 8, pp. 681-688, 2006.

[2] Y. Fan and Z. P. Jiang and S. Panwar, An adaptive control scheme for stabilizing TCP, Proc. 5th World Congress on Intelligent Control and Automation, Hangzhou, China, 2004.

[3] Y. Fan and Z.P.Jiang, A nonlinear flow control scheme under capacity constraints, Acta Automatica Sinica, vol. 31, no. 1, pp. 64-74, Jan. 2005.

[4] Y. Fan, Z.P. Jiang and X. Wu, A control-theoretic approach to stabilizing queues in large-scale networks, IEEE Communications Letters, Vol. 9, No. 10, pp. 951-953, 2005.

[5] H. Zhang, Z.P. Jiang, Y. Fan and S. Panwar, Optimization-based flow control with improved performance, Communications in Information and Systems, vol. 4, No. 3, pp. 235-252,  2004.

[6] H. Zhang and Z.P.Jiang, Analysis of two ad hoc broadcasting protocols, IEEE wireless Communications and Networking Conference (WCNC), Atlanta, GA, March, 2004.

[7] H. Zhang and Z.P. Jiang, Performance analysis of broadcasting schemes in mobile ad hoc networks, IEEE Communications Letters, vol. 8, no. 12, pp. 718-720, Dec. 2004.

[8] H. Zhang and Z.P. Jiang, Modeling and performance analysis of ad hoc broadcasting schemes, Performance Evaluation, Vol. 63, pp. 1196—1215, 2006.

[9] H. Zhang and Z.P.Jiang, Mobility sensitive broadcast algorithms in highly mobile ad hoc networks, Ad Hoc & Sensor Wireless Networks, Vol. 3, Issues 2-3, pp. 171-196, 2007.

Robust Adaptive Nonlinear Control System Designs

Participating Faculty: Farshad Khorrami
Web Site:  http://crrl.poly.edu/pub1.html
Funding Sources: NSF and ARO

We have developed a powerful and flexible paradigm for dynamic high-gain based design of controllers and observers for various classes of nonlinear systems.  The design technique is applicable in both the state-feedback and the output-feedback cases. The technique utilizes a state scaling generated through an appropriately designed dynamics driven by the measured outputs of the system. The resulting controller and observer are algebraically simple requiring no recursive computations and the associated Lyapunov functions have a simple scaled quadratic form. The stability analysis is based on the solution of a pair of coupled matrix Lyapunov equations. Necessary and sufficient conditions for the solvability of the coupled Lyapunov equations have been obtained in our recent results.  

The approach provides a unified design procedure applicable to both lower triangular (feedback) and upper triangular (feedforward) systems and also to some classes of polynomially bounded systems without requiring any triangularity in the system structure. The controller provides strong robustness properties and allows coupling with a dynamic high-gain observer whose design is dual to that of the controller to obtain output-feedback stabilization and tracking results. This represents the first output-feedback result for feedforward systems.  The controller and observer designs in the case of feedforward systems are strongly parallel to the designs in the case of strict-feedback systems suggesting that the proposed technique could allow further extensions for feedforward systems along various lines that have been hitherto investigated only for strict-feedback systems. In both, the strict feedback and the feedforward cases, a greater generality and complexity of bounds on uncertain functions in the system does not increase the complexity of the control law, the observer, and the Lyapunov function, but is instead handled through the dynamics of the scaling parameter. Furthermore, the scaling parameter dynamics can be designed to provide robustness to various perturbations including unknown parameters, additive disturbances, inverse dynamics, and appended Input-to-State Stable (ISS) dynamics. A generalized scaling technique utilizing arbitrary powers of the scaling parameter has been developed to weaken the assumptions on the system and to extend the results to non-triangular systems. Current research effort on this topic is focussed on extending the results in the following directions: 1) systems with ISS appended dynamics considering general interconnections (dependent on all states) 2) adaptive control without a priori bounds on unknown parameters 3) decentralized control for large-scale systems with each subsystem control input designed using the dynamic high-gain scaling based technique 4) disturbance attenuation 5) relaxation of assumptions by considering more general scaling patterns/ multiple scalings and more general scaling dynamics.

Development of Hardware/Software Architecture for Autonomous Unmanned Vehicles

Faculty: Farshad Khorrami
Web Site:  http://crrl.poly.edu/pub1.html
Funding Sources: ARO and Industrial Funding

There are many civilian (e.g., weather forecasting, traffic monitoring, police operation, fire fighting, port security) and military applications (e.g., surveillance, target identification, ordinance delivery, communication relay) for the use of unmanned autonomous vehicles (UAV). For aerial vehicles, current FAA requirements analyses identify two possible means by which UAVs may be accepted into civilian airspace: 1) requiring all UAVs to have the necessary transponder hardware to identify themselves to the control tower or 2) requiring UAVs to have autonomous obstacle avoidance systems (OAS). This latter approach of requiring OAS on UAVs is favored by the industry.  An autonomous unmanned aircraft equipped with an OAS will be able to carry out a multitude of missions requiring flight to a designated target without benefit of remote piloting or any prior knowledge of the local geography. Our on-going effort is in the direction of development of a low resource OAS and its implementation on a small autonomous helicopter.

There has been much interest in UAVs in the past few decades; however, miniaturization of sensors, electronics, and fast microprocessors in the past decade have improved feasibility of small autonomous vehicles with on board OAS. Nevertheless, this is a challenging problem due to limitation of space/volume, payload capacity, power, and computational resources on board a small UAV. We have been considering several classes of UAVs, namely unmanned aerial vehicles, unmanned underwater vehicles, unmanned surface vehicles, unmanned rotary wing aircrafts. The picture of our unmanned rotary wing aircraft and a hardware-in-the-loop simulation for this test bed are given below. We have on-going efforts in regard to fixed wing aircrafts, unmanned surface vehicles (USVs), and unmanned underwater vechicles ( UUVs).

Nonlinear Control for Electric Motors

Participating Faculty: Farshad Khorrami
Web Site: http://crrl.poly.edu/pub1.html
Funding Sources: NSF, ARO, and Industrial Funding

Electromechanical actuators have been utilized in many applications from home appliances to sophisticated guidance and control systems. Various electromechanical actuators such as electric motors, hydraulic and pnuematic actuators, smart materials (e.g., piezoceramics, magneto-restrictive materials, shape-memory alloys, electrorehological fluids, etc.) have been considered. Modeling and control design for such actuators have been and are pursued to achieve a higher level of performance. As a part of our on-going efforts, we consider modeling and control design of electric motors, namely step motors, brushless DC motors, and induction motors. These electrical motors are used in many applications; some requiring a high level of accuracy and performance such as machines used in electronics industry for assembly or semiconductor wafer probing and inspection.

To achieve high precision and bandwidth for the motors, we explore modeling of these devices to the extent needed to provide a high performance controller but at the same time amenable to model-based nonlinear designs. To  this extent, we consider nonlinear and adaptive controllers to derive robust and high performance feedback controllers which is essential for applications that require high performance and accuracies. The recent nonlinear and adaptive design tools have shown to be effective in designing robust controllers achieving robust performance. We have utilized existing nonlinear tools and their extensions to design robust adaptive controllers for various motors under full state or partial state measurement (sensors less control). We have experimental test beds for such motors at the Contol/Robotics Research Laboratory at Polytechnic Institute of NYU and one such test bed is a dual axis linear stepper motor used in electronics industry. A picture of this set up is shown below.

Book:
F. Khorrami, P. Krishnamurthy, H. Melkote, Modeling and Adaptive Nonlinear Control of Electric Motors, Spring Verlag, Heidelberg, 2003.

Development of an Ultra-Accurate High-Speed Six DOF Manipulator and Other Robotic Systems

Participating Faculty: Farshad Khorrami
Web Site: http://crrl.poly.edu/pub1.html
Funding Sources: NSF and Industrial Funding

Dual-axis linear motors (i.e., Sawyer motors) are utilized in various manufacturing applications such as electronics industry for assembly, packaging, lead bonding, and wafer probing. In these applications, additional degrees of freedom are attained by using other types of motors and mechanisms (e.g., geared motors or lead screws). Inherently, these additional mechanisms degrade the overall performance of the system and various motor technologies need to be employed to design a four to six degree of freedom (DOF) manipulation system (four being common for many packaging and assembly operations).
We have considered the use of Sawyer motors to design a direct drive six DOF manipulator. Many 6 DOF industrial manipulators have been built by various groups and companies. The unique features of the proposed 6 DOF manipulator are that it is direct drive, uses a single motor technology, has a large work space, is high speed, and is capable of achieving high positional resolution. We have performed detailed kinematic and dynamic modeling for this type of manipulators. Furthermore, we carried out a kinematic optimization to maximize the manipulator workspace. The advocated six degree-of-freedom positioning system is a tripod structure with inextensible limbs actuated at the base by two dimensional linear stepper motors (although other types of actuators may be utilized). The proposed manipulator achieves large range of motion in all the six degrees of freedom. Furthermore, high resolution and high speed motion may be achieved in all axes. A picture of our tripod manipulator is given below.

Decentralized Control of Nonlinear Large-Scale Interconnected Systems

Participating Faculty: Farshad Khorrami
Web Site: http://crrl.poly.edu/pub1.html
Funding Sources: NSF and ARO

Several real-world large-scale systems can be viewed as interconnections of linear/nonlinear subsystems with constraints on information flow between the subsystems. We have addressed the decentralized control problem for large-scale systems under various sets of assumptions on the subsystem structures and interconnection topologies. We have also applied these results to a variety of large-scale systems including power networks, smart structures, and satellite formations. Our results on nonlinear control techniques have enabled us to weaken the required assumptions on the structures of the individual subsystems and also on the interconnection (or coupling) among the subsystems. We have also extended the results to include adaptations to compensate for unknown system parameters and also to provide robustness to uncertain terms and appended nonlinear dynamics. Furthermore, we have investigated techniques to achieve decentralized attenuation of disturbance inputs and provided explicit guaranteed bounds on the disturbance attenuation along with tuning strategies to achieve desired disturbance attenuation properties through the proper choice of controller parameters. Decentralization of the control may be achieved both through a centralized or a decentralized design of the decentralized controllers. In our research, both strategies have been utilized.