Continual process adaptation and task learning

Even if a robot is working robustly within an operating range, that operating range may change because of new products, variants or tasks. In such cases, the robotic system must be changed over to the new situation, which, at present, may require as much time as the initial setup phase. Within each working situation, there must be a continuous adaptation of parameter sets and ranges in order to optimise production.

This must be based on proper underlying models, such that learned or adapted parameters can be efficiently transferred to new settings, including cases in which a task is handed over to a different type of robot. The desire for continuos process optimization gives rise to the objective of continual, (semi-)automatic system adaptation. The same applies to learning, which can be iterative, structural or interactive, to automate seamless integration of new tools or requirements.


Learning based error recovery