Control, Autonomy and Intelligence
What makes the difference between robots and mechanisms is the ability of robots to adapt to changes of their tasks or their subjects of operation or of their operating environment. A robot should be able to
- interprete the directives that describe its task,
- understand the operating environment from data provided by its perception sensors,
- reason about its state and the state of other robots/human ("agents") present in the same environment
- perform motion planning and activity planning based on task description, on the environment and own/agents states,
- control the execution of the actions, while allocating attention to task-related events,
- and anticipate outcome of actions,
When all this is performed without human guidance the robot can be called Autonomous.
Degrees of Autonomy
There are several degrees of "smartness" of how Robot Autonomy can be implemented. Examples of possible degrees of Robot Autonomy for a mobile robot (with respect to the abilities mentioned above):
- the robot could understand directives in the form of
- low level program sequences (e.g. drive to x,y; move robot arm to x,y,z)
- natural language (e.g. analyse any strange stone closeby)
- the understanding of the environment could be
- limited to determination of the configuration of a standard environment (e.g. the positions of doors in a corridor)
- reconstruction of the 3D model of an outdoor environment and association of entities to it (e.g. boulder, tree, pond)
- reasoning on own/co-agent states could be
- simple tracking of own resources (e.g. level of energy in batteries)
- determining how tasks could be performed in co-operation with other agents
- planning could be
- geometrical/temporal planning of motion (e.g. interpolating a trajectory)
- break down high level task into elementary actions including resources and contingency actions (e.g. survey area; stop on strange looking stone; grasp stone; deposit in analysis instrument; run analysis procedure)
- control could be implemented as
- feedback execution of a command (e.g. a proportional, integrative and derivative control to follow a trajectory in space)
- a set of behaviours triggered by events (e.g. when bump in obstacle backtrack)
The implementation of even the simplest Autonomy requires a computer with suitable interfacing means to the robot sensors and actuators. Such computer is called Robot Controller.
Robot Controller
Robot Autonomy is implemented by means of a computer system, dedicated electronics and software making the so-called robot controller. Due to the absence of suitable space-rated Robot Controllers, the A&R section has developed one in the course of several R&D projects. The first one was the “SPAce Robot COntroller” (SPARCO) in 1994, followed by “Common European Space A&R COntroller” (CESAR) in 1996, by “Servo and Power Electronics for A&R” in 1997, by “Compact Integrated Robot Controller Unit and Servo Amplifier” (CIRCUS) in 1999 and “space A&R CONTroller Extensions” (CONTEXT) in 2002. All robot control software has been designed to support the “Interactive Autonomy” control mode.
