Topological characterization of mobile robot behavior

We propose to classify the behaviors of a mobile robot thanks to topological methods as an alternative to metric ones. To do so, we adapt an analysis scheme from physics of nonlinear systems in chaotic regime, assuming a dissipative dynamics that relaxes on a low-dimensional manifold. Sensor data recorded from a mobile robot during a wall-following experiment allows to compute topological invariants that give a standardized representation of the structure of the set of trajectories, and enable us to discriminate among similar behaviors in a systematic and quantitative way