Title: GUARDIANS: Group of Unmanned Assistant Robots Deployed In Aggregative Navigation supported by Scent detection
Funding Organization: European Commission under the 6th Framework Programme - Priority 2 “Information Society Technologies,” Specific Targeted Research Project (STREP)
Duration: 01/12/2006 – 31/01/2010
Abstract: The GUARDIANS are a swarm of autonomous robots applied to navigate and search an urban ground. The project's central example is an industrial warehouse in smoke, as proposed by the Fire and Rescue Service. The job is time consuming and dangerous; toxics may be released and humans senses can be severely impaired. They get disoriented and may get lost. The robots warn for toxic chemicals, provide and maintain mobile communication links, infer localization information and assist in searching. They enhance operational safety and speed and thus indirectly save lives. The robots navigate autonomously and accompany a human squad-leader. They connect to a wireless ad-hoc network and forward data to the squad-leader and the control station. The network is self-organising, adapts to connection failures by modifying its connections from local up to central connections. The autonomous swarm operates in communicative and non-communicative mode. In communicative mode automatic service discovery is applied: the robots find peers to help them. The wireless network also enables the robots to support a human squad-leader operating within close range. The aim is for flexible and seamless switching between these modes in order to compensate for loss of network signals and to support and safeguard the squad-leader. Several robot platforms are used, off-the-shelf mini-robots as well as middle sized robots. The emphasis in data collection is on toxic plume detection, to enable olfactory-based navigation, allow safe progress for the human squad leader and to detect plume sources. The major aim of the project is to develop a swarm of autonomous robots that is able to adequately assist and safeguard a human squad leader. The project organises workshops with end-users (rescue workers and fire-fighters) and the advisory board, to assess the demonstrations and to disseminate research results. The workshops, moreover, aim at exploring additional exploitation of results.
Title: Robot Swarms: Coordination and Formation Control of Groups
The Scientific and Technological Research Council of
Duration: 01/08/2005 – 01/08/2008
Abstract: Swarming behavior (the behavior of being in a group and acting in coordination) can be seen in many creatures in nature starting from simple bacteria swarms to flocks of birds and schools of fish. Research has shown that evolution of such behavior is due to evolutionary advantages of such behavior. For example it has been observed that swarming individuals are more effective in finding food and avoiding predators. From the perspective of systems theory swarm (or multi-agent) systems are more robust, more flexible, more affective, and can be produced at relatively low cost. For instance consider an ant colony: The loss of one or several ants does not prevent the system from functioning (robustness), the ants can organize in one form to perform certain task and in another form to perform a different task (flexibility). Moreover, few ants cooperatively can perform tasks which are much above their individual capacity (effectiveness). Furthermore, the cost (of breeding etc) of an ant is much cheaper (compared to some other animals, e.g., mammals), therefore it is possible for the ants to reproduce in large numbers in short periods of time (lower cost). Technological developments allow implementation of artificial swarms or simply swarms of robots. However, in order for such distributed systems to become operational there is a need for development of new effective coordination mechanisms.
The main scientific objective of this project is to mathematically model and perform rigorous analysis of group coordination and dynamics of leaderless swarms such as schools of fish and to implement the theoretical developments on a swarm of mobile robots. With this perspective the coordination mechanisms of a group or a swarm, the behavior of the group under outside disturbance (such as a predator attack), the aggregation and dissolution dynamics of the group will be investigated. Moreover, new formation control and formation reconfiguration mechanisms will be developed and investigated. Such modeling and analysis will not only be helpful in understanding swarm behavior, but will also allow development of new decentralized coordination and control methods for swarms of robots. The research findings will help in successful development and implementation of unmanned air, underwater or land vehicles/robots. Our main approach to the problem will be from the systems theoretic perspective and we will use well known and respected rigorous analysis methods such as Lyapunov stability theory and if needed (in case these are insufficient for the problem) new methods/approaches will be developed.
The mobile robots, which will be obtained from a Turkish company, will not only be used in this project, but will also be helpful in other experimental research, teaching, student class or senior design projects, or even graduate studies and research. They will help in improving the robotics infrastructure of our university and will provide important contribution to our research in the area. Moreover, it will help our institution in participating more actively in international programs such as the 6th Framework Programme of the European Union.
Title: Particle Swarm Optimization: Parallel and Asynchronous Implementation, Rigorous Convergence Analysis, and Implementation on a Multi-Robot System
Funding Organization: The Scientific and Technological Research Council of Turkey (TÜBİTAK), Career Project/Award
Duration: 01/02/2007 – 01/02/2010
Abstract: Particle Swarm Optimization (PSO) has become popular optimization (function minimization) method in the recent years. Since its introduction in the nineties there have been about 300-400 conference and journals papers published on the topic. However, most of these publications are empirical studies. The method has been applied to may benchmark and other problems and effective results have been obtained; however, no rigorous analysis of the convergence properties of the algorithm has been performed. In this project the convergence properties of the method will be analyzed using rigorous techniques. Moreover, a different asynchronous modeling and implementation of the algorithm will be performed. Furthermore, the neighborhood of the particles will be allowed to be time-varying (or dynamic) and the affect of this dynamic topology on the performance of the algorithm will be studied. Finally, the algorithm will be implemented on a multi-robot system in a way not seen in previous applications. In this implementation multiple robots will act as a particle in the PSO system or basically will determine their motions/positions based on the dynamics of the particles in the PSO algorithm and based on that will perform search operation.
Title: Adaptive Swarm Control
Funding Organization: The Scientific and Technological Research Council of Turkey (TÜBİTAK), Research Project
Duration: 01/03/2010 – 15/07/2014
Abstract: The number of works in the literature on swarm control using adaptive methods has been insufficient. In this project we will fill in the gap in this field and will develop adaptive methods for swarm control. The problem will be approached from several different viewpoints/frameworks. The first approach will be from the point of output regulation of nonlinear systems. Using the results on adaptive internal model, developed by Isidorı and Serrani, we will develop adaptive formation acquisition and formation reconfiguration (such as rotation, expansion, and shape change) and trajectory tracking methods. Using this method we will also work on achieving behavior such as aggregation, social foraging, orientation alignment, and distributed agreement. The second approach will be from the point of adaptive intelligent methods (adaptive neural networks and/or adaptive fuzzy systems). With this objective using the direct and indirect adaptive control approaches and adapting the parameters of the intelligent controller we will again work on achieving formation acquisition, formation reconfiguration, aggregation, social foraging, orientation alignment, and distributed agreement. In this approach for adapting the parameters of the intelligent controller we will use mathematical methods which guarantee stability and the controller will be augmented with boundary and sliding mode terms. As a third approach we will use optimization algorithms such as Particle Swarm Optimization and/or the Genetic Algorithm for adapting the parameters of classic or intelligent controllers and achieving the above mentioned swarm behavior.
RCS is a methodology for development of complex, hierarchical, and distributed control systems. It is based on the Reference Model Architecture (RMA) for intelligent systems design, developed by Intelligent Systems Division (ISD) of the Manufacturing Engineering Laboratory (MEL) at the National Institute of Standards and Technology (NIST).
I worked on this project at the Ohio State University in order to support my graduate studies (May 1997-December 2000).
· Implemented RCS controllers using the RCS Software Library to several experiments in the OSU Controls Laboratory,
· Participated in developing and teaching a university laboratory course on RCS,
· Developed a web based introductory documentation on the RCS Library (see the page below),
· Developed a catalog of generic control and estimation routines for the RCS Library (see the page below),
· Co-authored a book on the RCS Library (see the list of publications).
Visit Catalog of Generic Control and Estimation Routines for more information on the RCS Software Library.
This is another
project that I worked on (between January 2001 and September 2001) during my
graduate studies at the
I worked on this project between October 2001 and August 2002. It was funded by the Defense Advanced Research Project Agency (DARPA) ``Mixed Initiative Control of Automa-teams (MICA)'' program. It was a large project with several universities and companies involved. My primary job in the project was to develop a kinematic model for swarms and perform a rigorous stability analysis (that can serve as a "proof of concept" for collective aggregating behavior etc.)