Principles, architectures and trends in autonomous control

Feedback systems introduced some degree of autonomy of machines throughout the history of engineering. Feedback has been introduced to the operations of engines, servos, production processes, aircraft, car and ship control systems. These systems are limited to particular tasks under limited variety of disturbances. They execute human commands very well but it is not within the scope of these feedback systems to make plans and execute them without human supervision. A next level of autonomy is when feedback is still present but is augmented with small time-scale planning and autonomous execution by the machine on its own. At this level human command is an instruction to execute a short time scale mission. For instance a robot is sent to take pictures in a dangerous place for humans to enter, and is asked to return after the job is completed. In this case there is no remote control used, the robot plans itself and executes to whole job on its own. Another example is an unmanned aeronautical vehicle that surveys an area and returns, without any remote control. Yet another example is the autonomous control of a part of a manufacturing process. A higher level of autonomy is when a machine is capable to function at the previous two levels and in addition it also has some long term objectives. Long term objectives are formulated as operating conditions to be satisfied and some goals to be achieved. For instance during a long term space mission an autonomous system has to maintain the functioning of its communications, power sources and functioning of its mission relevant instrumentation despite adverse circumstances that may emerge by micrometeorite hits or malfunction of some components. It has to maintain its course that is actually a vital part of its mission. After landing on a planet not every task can be immediately executed, circumstances have to be created and measurements can only be taken under certain conditions. So planning and execution is becoming much more involved. At this level of autonomy a machine has to organize itself to achieve a list of final objectives, in whatever order it can manage to do these. The conclusion is focus on highlighting the areas where there is a gap in knowledge and more effort should be made to quicken progress.