Individual localization using landmarks in the presence of interacting robots
Mesbah , Anousha
MetadataShow full item record
In order to path plan and navigate smoothly toward a goal, a mobile robot must know where it is in the environment. Thus robust state estimation is a requirement especially in indoor environments where GPS is not available. Particle filtering is a key technique for addressing this challenge. However, it has been mostly used for navigation of a single robot or robots that interact with other agents such as humans, modeling them as random noise. In this thesis, we generalize particle filtering in order to localize in the presence of other robots. Our approach is to maintain a nested set of particles for tracking the other robots and modeling the behavior of other robots that may or may not be cooperative. In our specific environment, another robot interacts with the subject robot by changing the location of landmarks previously known to a subject robot in order to confuse and mislead the subject robot. Our approach performs good localization of the subject robot, tracks the other agent by estimating its pose, predicts its behavior, and serves to make localization more robust and practically applicable.