Loading...
Please wait, while we are loading the content...
Similar Documents
Hybrid Reactions Modeling for Top-down Design Framework
| Content Provider | Semantic Scholar |
|---|---|
| Author | Loc, Matthey |
| Copyright Year | 2008 |
| Abstract | This report presents the work accomplished by Loic Matthey during his Master project from Ecole Polytechnique Federale de Lausanne (EPFL). This project was a joint work between the Distributed Intelligent System and Algorithms (DISAL) Laboratory at EPFL, Switzerland and the General Robotics, Automation, Sensing and Perception (GRASP) Laboratory at University of Pennsylvania, United States of America. It took place in the Master springsummer semester 2008. We present a theoretical framework to design Top-down control scheme for arbitrary systems. Being able to control a complex system using high-level instructions only is a promising and attractive paradigm. Our approach is based on the use of a Chemical Reaction Network model, used as a proxy to derive the control schemes. To test the application of our method, we consider the Top-down control of a realistic multi-robots assembly platform, simulated using a 3D physics simulator, Webots. First we present the modeling of the robotic platform using a Chemical Reaction Network. The free parameters are precisely fitted. We simulate the system using an ODE approximation and an exact stochastic simulation. We find that the model can be made to fit quantitatively to the experimental data, especially when using a stochastic simulation approach. Second we define an optimization and control scheme for a class of Chemical Reaction Networks. We prove convergence results and write the optimization problem as a linear program of the time of convergence of the system under constraints on the equilibrium value. It allows us to design sets of reaction rates producing a specified converged behavior, in polynomial time. This optimization provides precise controls of the system using only highlevel goals. Finally, we map the optimized model down onto the realistic physical assembly platform. We find that the system can be controlled using the optimized parameters of the model level, but that small discrepancies can have disruptive effects. This page intentionally left blank. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.gatsby.ucl.ac.uk/~lmatthey/documents/loicmatthey_masterthesis.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
