Autonomous Robot Navigation using Flatness- based Control and Multi-Sensor Fusion

This research work contains results on flatness-based control and sensor fusion for mobile robot systems. Flatness-based control is applicable to differentially flat systems i.e. to systems the behavior of which is determined by the trajectory of a finite collection of quantities, consisting of the flat output and its derivatives [Mounier, H. & Roudolf, J. (2001)],[Rouchon, P. (2005)]. Flatness-based control is equivalent to the feedback linearization method where the error dynamics of the closed-loop system can be described by a linear ODE after state feedback and subsequently can be stabilized using methods from linear control theory [Fliess, M. & Mounier, H. (1999)], [Roudolf, H. (2003)]. An advantage of flatness based control is that it simplifies trajectory planning and enables open-loop controller design. For linear finite dimensional systems flatness coincides with controllability, while this property can be generalized in the case of infinite dimensional systems [Laroche, B.; Martin, P. & Petit, N. (2007)], [Martin, P. & Rouchon, P. (1999)], [Meurer, T. & Zeitz, M. (2004) ,[Lévine, J. & Nguyen, D.V. (2003)]. Motion planning and control of autonomous vehicles is an important research topic in robotics (results in [Rigatos, G.G. (2003)], [Rigatos, G.G.et al., (2001)], [Rigatos, G.G. (2008)]) and flatness based control has been proposed as a suitable methodology for this class of problems [Martin, P. & Rouchon, P, (1999)]. Using the concept of flatness-based control, motion control algorithms have been developed that permit steering of the robotic vehicle along any desirable path in the 2D plane [Oriolo, G.et al., (2002)]. It will be shown that the kinematic model of the robotic vehicle is a flat system and thus can be expressed using a flat output and its derivatives. Moreover, the case in which the mobile robot's state vector is estimated through fusion of measurements from distributed sensors will be examined [Caron, F. et al., (2007)], [Jetto, L. et al., (1999)], [Yang, N. et al., (2005)]. To this end, the state vector of the robotic vehicle will be reconstructed with the use of Gaussian or nonparametric state estimators (such as Extended Kalman Filtering or Particle Filtering) [Rigatos, G.G. (2007)], [Rigatos, G.G. & Tzafestas, S.G. (2007a)], [Rigatos, G.G. (2007b)]. Simulation experiments in the case of the completely measurable state vector can show that flatness-based control enables the mobile robot to follow any reference path. Additionally, simulation experiments can show that using the state vector which is estimated from sensor O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m