Real-Time Control in Robotic Systems
چکیده
Robotic systems are beginning to interact with humans and are increasing in dimensionality, control complexity, task difficulty, and dynamic capability. It is therefore becoming increasingly imperative that control be performed in real-time (i.e. on or faster than the timescale of the system dynamics), intelligently making decisions on demand, and then executing them in a dependable fashion. In addition, high-dimensional bio-inspired systems are becoming more common (Butterfab & et al., 2001; Movellan et al., 2005; Simpkins et al., 2011; Todorov et al., 2010; Vandeweghe et al., 2004; Wilkinson et al., 2003), and are based upon a different strategy for control (and therefore control system design) than traditional engineered systems coordinated, compliant, and often coupled control rather than orthogonal, stiff, and independent control of joints. As a result, issues related to real-time capability, more critical now than ever, must be addressedwith a coherent, measured approach. In order to deal with this challenge appropriately, a number of sub-problems must be addressed and will be discussed in this chapter. These are timing, dimensionality, computational capabilities, system bandwidth, and safety. There are fields devoted to each problem independently, and the background of each will be given as each one is discussed. Rarely does a researcher or practitioner in one field work in any of the other fields, let alone all of them. However, ignoring any of these sub-problems will likely lead to a failure of a complex real-time system. Therefore, familiarization with these issues and methods will help the researcher and practitioner design, implement, and troubleshoot a real-time system most effectively. This chapter will present current solution methods for each component of the problem and show how they can be integrated together effectively.