Development of Safe and Secure Control Software for Autonomous Mobile Robots

Abstract

It is often said that autonomous robots should be trustworthy or dependable, meaning by this safe, secure, reliable, etc This terms are too general to be useful, so we prefer to limit the scope of this chapter to two of their inter-related components safety and security. They must be built into a system from the start; it is difficult, if not impossible, to add them in an adequate and cost-effective manner later on. We view autonomous robots as situated, real-time embedded systems endowed with enough intelligence to adapt to changing environment and learn from their experience. They may operate unattended and through an unsafe operation may cause significant human, economic or mission losses. The focus of this chapter is on safety and security of robot control software. This software allows unprecedented complexity of robotic systems, which goes beyond the ability of current engineering techniques for assuring acceptable risk. Most of the publications on safety has a form of recommendations on providing safe environment for robot operators, like the Occupational Safety and Health Administration regulations or the more recent NASA recommendations for space robots. This approach is effective when accidents are primarily caused by hardware components failures. As software becomes more and more important in robot control, we have to consider ways to prevent accidents caused by software. Robot control software consists of many interacting components. Accidents arise in the interactions among the components rather than the failure of individual components. The need for safety is obvious, but how to ensure it is less obvious. Autonomous robots may operate unattended and through an unsafe operation may cause significant human, economic, or mission losses. Similar problems were encountered early on in manufacturing automation; but autonomous mobile robots may change their behaviour and operate in much less controlled environments. We will review at first the principal concepts of system safety like risk and hazard and some traditional approaches to dealing with them. We consider security as a subset of safety and we will present our point of view on this issue. The present trend to make the robots more autonomous requires new approaches to deal with much more complex problems of their safety. After review of several robot control architectures from the viewpoint of their safety we present an approach based on systems theory. While the theory was developed long time ago it turns out very useful to ensure