Rapid Prototyping of Smart Structure Controllers

George Kantor and Prof. W.P. Dayawansa


We have combined Integrated System Inc.'s (ISI) AC100 system with existing resources in the Intelligent Servosystems Lab to create a test bed for real time control applications. Using this system, we can quickly obtain a model for a given plant, design a controller, simulate the controller, and implement the controller on the actual plant. The plant considered here consists of a flexible aluminum cantilever beam with piezoceramic (PZT) sensor and actuator. The control objective is to damp vibrations in the beam as quickly as possible.


The first step in the design process is to obtain a model for the beam. This is the purpose of automatic system identification. Here the AC100 is used to generate a random disturbance, which is sent to the input of the plant (the PZT actuator) via one of the AC100's digital to analog converters (DAC). The plant's response (the PZT sensor output) is recorded using one of the AC100's analog to digital converters (ADC). This input/output data is then processed to generate a real rational transfer function model of the plant. This can be achieved online using a recursive least squares (RLS) algorithm or it can be accomplished using the wavelet based matching pursuits (MP) algorithm. The model generated by automatic system identification is next used to design a controller for the plant. ISI's SystemBuild provides the computational tools necessary to implement most modern controller design algorithms. SystemBuild is also used to verify the controller design in simulation. Once we find a controller that works well in simulation, AC100 provides the resources to implement the controller on the actual plant. The Autocode Generator generates C code for the controller developed in SystemBuild. Using the Hardware Connection Editor we can specify which input and output devices will be connected to the inputs and outputs of the controller. The AC100 then compiles this information and downloads it to a digital signal processing chip which then runs the controller.


This test bed was used to successfully design and implement the saturation respecting control law of P. Gutman and P. Hagander [1985]. An accurate 10th order linear model of the beam was obtained using online RLS. SystemBuild was then used to execute the Gutman-Hagander control algorithm, and the resulting controller was implemented with the AC100. The controller successfully reduced the settling time due to a step disturbance from 6.42 seconds to 2.45 seconds.


The steps outlined in the above process apply not only to the flexible beam, but to a wide variety of other plants as well. In fact, a controller for any stable plant can be designed in exactly the same way. Controllers for unstable plants can be designed with only a slight modification in the system identification step. Hence, our test bed greatly reduces the amount of work involved in implementing a new controller for any system.


Our rapid prototyping test bed is complete and we are beginning to expore the potential of this new tool. The system has been succesfully employed in several smart structure experiments and is now being used to investigate various practical applications of the Modular Dextrous Hand (movie). As for the future, this rapid controller design ability will undoubtably be used for a vide variety of projects covering all aspects of research in the Intelligent Servosystems Lab, allowing us to back our theoretical research with strong experimental results.