Environment modeling for maintenance automation-a next-best-view approach for combining space exploration and object recognition tasks

In order to enable robot systems to solve complex tasks autonomously, it is essential to model the environment in a sufficient way. This work describes a new approach to create a consistent environment model, using next-best-view planning which combines the optimization of the space exploration and the object recognition task. The approach is designed for the automation of maintenance tasks, which is a challenging task for autonomous robotic systems. After a general overview of the developed system, we explain the theoretic foundations for building a consistent environment model which can be used for manipulation planning. The concept is validated by means of application-oriented examples.