Dynamics aware 3D occupancy grid map with semantic information

In many modern robotics applications the agents are expected to produce accurate representations of dynamic environments. Most mapping algorithms, however, assume the world to be static. Furthermore, these applications may also require semantic information about the environment. Most popular approaches add extra information as additional layers on top of the map used for navigation, leading to complex, decoupled systems. In this paper we present a probabilistic grid-based approach for modeling changing environments which can also incorporate augmented information right in the core of the map. Our method represents both, the dynamic of the cells and the objects contained in each of them by the state transition probabilities of a Hidden Markov Model. We evaluated the map in different use cases, proving its usefulness for modern robotics applications. The framework is available as a new, open source, mapping software.