Feedback Control

Fuzzy-Based Feedback Control of a Tip-Mounted Module

In this work we seek to use on-board proprioceptive sensing data to close a position loop through selective activation (via Joule heating) and cooling (via forced water convection) of a distally-mounted SMA antagonistic actuation scheme. While the classical PID control paradigm is fairly versatile and effective for many position control problems, complications arise when dealing with highly nonlinear plants with parameters that vary in time. Such is the case when dealing with closed-loop control of SMA actuators due to their highly-nonlinear, hysteretic, and time-variant thermomechanical properties (coupled with the discontinuous nature of forced convection control). For this reason, it is advantageous to adapt the PID gains in real time to deal with these nonlinearities. Many attempts to improve the closed-loop performance of SMA actuated systems have been presented in literature, including adaptive sliding-mode controllers, inverse hysteresis controllers, nonlinear controllers and neural networks. Another popular approach to tune a PID-based SMA control system is to adapt the gains in real time based on a fuzzy logic controller. Due to the ability to intuitively and deterministically encode expert intuition into the control of the SMA actuators, without relying on black-box approaches (neural networks) or model-based controllers, we developed a fuzzy-tuned PID/PWM (FPID/PWM) controller, which consists of a fuzzy inference engine that outputs the fuzzy-tuned PID gains which are fed into a high-level PID position controller, which subsequently generates a current command sent to a low-level PID-PWM current controller for controlling the current in each SMA.

To control the forced convection cooling valve states, we use a logic-based controller based on the error and its derivative.Therefore, the master control architecture consists of continuous FPID/PWM SMA current controller (which switches between on/off states for each actuator based on the sign of the error), combined with a discontinuous hysteresis valve controller.

Broader Topics Explored: Control theory, fuzzy logic

Fuzzy-based PID/PWM control: (left) control architecture: (a) overall block diagram, (b) fuzzy inference engine, (c) high-level PID position controller, (d) low-level current PID/PWM controller

Fuzzy control surfaces for the PID gains for both the agonist and antagonist actuators

(Top) sinusoidal position tracking for an actuator-only module, (bottom) automated and teleoperated control of a fully-integrated distal module

Controller data of a fully integrated closed-loop module executing (left) automated triangular (constant velocity) profile, (middle) automated sinusoidal profile, (right) rate-based teleoperation.

Real-Time Controller for Validation Platform

To support my research in the development of custom shape memory alloy actuators and force/displacement sensors for modular robotics applications, I developed an evaluation and characterization platform that implements numerous closed-loop controllers for quasi-static and dynamic evaluation purposes. It is essentially a desktop instron/DMA, with the ability to control for SMA temperature and current simultaneously. Control software was developed in MATLAB/Simulink's real-time control environment. The various controllers are listed below:

Position Controller: PID with velocity/acceleration feed-forward and gravity compensation
Force Controller: PD with gravity feed-forward
Velocity Controller: PD controller on velocity (discrete derivative of position, low-passed)
SMA Current Controller: PD current controller to analog-to-PWM converter
SMA Temperature controller: PID controller based on feedback from infrared temperature sensor
Virtual Bias Spring: PID position controller where a static equilibrium equation is solved in real time to make the stage behave like a pre-stretched linear-elastic spring

Broader Topics Explored: Precision machine design, control theory, electromechanical design, embedded systems design

Block diagram of custom characterization/evaluation platform developed to support my research