Autonomous systems are becoming more and more important when it comes to tasks that are either too dangerous, too difficult or too monotonous for humans. The systems react to events in their environment and make an appropriate decision: They perceive, learn, think and act confidently and react intelligently to unforeseen changes in the environment.
Autonomous systems also offer great potential for collaborative robotics. For example, they can predict from trained experience what the worker will do next, and control their movements accordingly. This motion prediction makes collaboration safer and more efficient.
If an autonomous robot knows the desired end state, it can use the corresponding artificial intelligence (AI) algorithms to derive all the work steps that are necessary to solve the task. In addition to reinforcement learning, we also use bio-inspired AI: this efficient and specialized AI is inspired by the intelligence of ants, bees and other insects.
In order to enable these and other intelligent robot skills such as packing shipping crates or estimating weight, the AI can be taught using virtually generated data sets. Sim2Real Learning – learning in simulations or hybrid systems and executing what has been learned on the real robot – is also used.