MULTIOBJECTIVE CONTROL AND THE SELECTION OF SENSORS, ACTUATORS IN ROBOTICS. Conference Paper uri icon

abstract

  • Summary form only given, as follows. The problem of regulating linear stochastic systems which have some form of amplitude constraints on either the system inputs or outputs has a wide range of practical engineering applications. The author is motivated specifically by robot structures where the inputs represent noisy actuator signals, and the outputs represent positioning errors. These structures have mission requirements naturally stated in terms of variance constraints on the outputs, subject to variance constraints on the actuator inputs (such as power limitations). The performance requirements of robot tasks are usually specified in terms of root-mean-squared (rms) values on both the input and output variables. The rms contribution of each input or output in the overall system performance metric is called the 'input' or 'output' cost. A technique is derived that allows a linear controller to assign each of the multi-input or output costs. By assigning only the output costs rather than the input costs, the procedure serves to determine the power required of the actuators (required input costs) to achieve the requirements (specified output costs). This determination of actuator sizing (and location) for the control of flexible structures is an important feature of the method.

author list (cited authors)

  • Skelton, R. E.

complete list of authors

  • Skelton, RE

publication date

  • January 1987