Search

Autonomous microsystems often operate under both strict power and energy constraints and substantial environmental variation. To improve sensing and control performance, parameter identification techniques can be useful if they may be implemented within an appropriate power budget. While it is well known that identification algorithm performance depends on sampling rate, for energy minimization sensor power models and rates of parameter adaptation can also significantly influence optimal sensor usage. In this paper, an empirical, simulation-generated model is found for parameter error of recursive least square identification of a prototypical second-order linear continuous system as a function of sampling rate, number of samples, and sensor noise density. This model is coupled with representative power models of certain common sensing circuits used in microelectromechanical systems (MEMS) to recommend optimal sensing schemes for low-power parameter identification. A case study of a walking micro-robot is presented.

In micro-scale electromechanical systems, power to perform accurate position sensing often greatly exceed the power needed to generate motion. This paper explores the implications of sampling rate and amplifier noise density selection on performance of a system identification algorithm using a capacitive sensing circuit. Specific performance objectives are to minimize or limit convergence rate and power consumption to identify dynamics of a rotary micro-stage. A rearrangement of the conventional recursive least-squares identification algorithm is performed to make operating cost an explicit function of sensor design parameters. It is observed that there is a strong dependence of convergence rate and error on sampling rate, while energy dependence is driven by error that may be tolerated in final identified parameters.

An on-off iterative adaptive controller has been developed that is applicable to servo systems performing repeated motions under extremely strict power constraints. The motivation for this approach is the control of piezoelectric actuators in autonomous micro-robots, where power consumption in analog circuitry and/or for position sensing may be much larger than that of the actuators themselves. The control algorithm optimizes the switching instances between ‘on’ and ‘off’ inputs to the actuator using a stochastic approximation of the gradient of an objective function, namely that the system reach a specified output value at a specified time. This allows rapid convergence of system output to the desired value using just a single sensor measurement per iteration and discrete voltage inputs.