A Near-Optimal Sensor Scheduling Strategy for an on–off Controller With an Expensive Sensor

This paper describes an efficient method for scheduling an energy-consuming sensor sparingly in combination with an on-off controller, specifically for a finite horizon control problem in which only end states are critical. In certain low-power applications, such as autonomous microrobotics, on-off controllers can be very efficient in operating piezoelectric actuators (and other capacitive actuation schemes) compared to traditional analog and pulsewidth modulation controllers. However, with existing sensing circuitry, sensing at the same frequency as control can be prohibitively expensive, because energy consumption in the sensing circuitry may be comparable or even much higher than energy consumption for actuation. Instead, a method is presented for best scheduling a limited number of sensor measurements and updates to control inputs during a finite horizon on-off control problem, in response to Gaussian disturbances and measurement noise. To simplify the problem, a lower bound for the expected value of a quadratic error function of the end states is found, which permits rapid evaluation of candidate sensor times. When actuator energy consumption is incorporated in the optimization, this produces a numerically efficient near-optimal strategy for determining best measurement times and updates to the control input sequence.