Handling heterogeneous information sources for multi-robot sensor fusion

In real world scenarios, measurements from a robot's environment and their respective models are rarely homogeneous in terms of their uncertainty. Instead it is likely to have classes of objects that greatly differ in this respect, such as static and dynamic, unique and ambiguous or previously known and previously unknown objects. This paper extends the concept of FastSLAM to exploit this fact in order to more efficiently localize an autonomous mobile robot and simultaneously map features and track dynamic objects in its environment in a cooperative multi-robot scenario.