Scott Bortoff's Research Interests
Return to S. A. Bortoff's home page.
My areas of research include...
- Design of control systems
for nonlinear plants, especially robots;
- Design of observers for nonlinear systems. (An observer is a
dynamic systems that can estimate the internal state
of a nonlinear system, given real-time measurements
of input and output signals;
- Control of complex mechanical and robotic systems, especially
underactuated mechanical systems;
- Issues in real-time computer control; and
- Industrial control applications.
The area of nonlinear systems and their control is a very active
area of research. I
concentrate my efforts on new methods to design
practical nonlinear feedback controllers and observers, and
also in experimental verification and industrial implementation.
I have made extensive use of spline functions to represent
nonlinearity in plant models and nonlinear controllers. This representation
makes existing algorithms more computationally efficient,
and leads to entirely new methods. These contributions include:
- Pseudo-linearization using splines;
- Approximate feedback linearization using splines;
- Adaptive feedback linearization using multivariable splines;
- Adaptive nonlinear control of a variable reluctance motor;
- Synthesis of nonlinear observers.
For a list of publications, including on-line
abstracts, click here.
I have also constructed a number of experimental test beds. Their
purpose is to demonstrate the computational efficiency of an
algorithm, to illustrate the synthesis of a particular type of
nonlinear controller, and to compare performance of several competing
controllers on a real hardware. These test beds include:
- An aerial robot. This is a "60-size"
radio-controlled (RC) helicopter,
powered by a 3.2kW electric motor, and controlled by three
ground-based networked PCs and an airborne microcontroller.
The robot is equiped with an airborne camera and is capable of
remote inspection. The aerial robot is presently under construction.
- The Acrobot, an under-actuated
double pendulum, which is open-loop unstable, non-minimum phase,
highly nonlinear, and not feedback linearizable;
- The Rotating Inverted Pendulum, a
variation of the traditional linear inverted pendulum, which is
also open-loop unstable, non-minimum phase, and not feedback
linearizable;
- The Rotating Inverted Double
Pendulum, a novel type of double pendulum; and
- A Variable Reluctance Motor Test Bed,
consisting of a small, three-phase,
eight-pole VRM and current amplifiers for the windings, which we
use to test adaptive nonlinear control.
- A magnetic suspension system, which is an excellent demonstration
of feedback stabilization, root-locus design, and zero dynamics.
Several of these systems are controlled by a TI 320C30 Digital Signal
Processor (DSP), which is a commercially available chip that excels
at floating-point arithmetic. The DSP is hosted by a 486-class PC.
Available software includes an ANSI-compatible C compiler for the DSP,
Matlab and Mathematica. The set-up allows a nonlinear controller to
be designed for a given experiment using Mathematica, which can
directly generate C-code for the controller as output. Once the C is
compiled, experimental demonstration of a new control algorithm can be
accomplished in a matter of minutes.
bortoff@control.toronto.edu