Increasing autonomy in agricultural robots: unevenness estimation of the terrain ahead

Natural terrain traversing is playing an increasingly important role in service and field robotics to enable the achievement of several tasks in the most disparate fields of human activities, e.g. planetary explorations, nuclear site decommissioning, precision farming and so forth.In this research, a new method for terrain unevenness estimation is proposed in order to increase the ability of an agricultural bot to move autonomously in a countryside environment. By the use of exteroceptive sensors, such as stereoscopic or RGB-D cameras, soil surfaces can be acquired and then manipulated to obtain a stochastic description of their geometry by means of the Power Spectral Density (PSD)-based analysis, in order to infer soil irregularities. This automatic ground unevenness estimation is performed online during the motion, allowing safe management of robot activities.