Robot navigation based on visual feature perception and Monte Carlo sampling

The mobile robot navigation algorithm with visual feature perception and Monte Carlo sampling is explored in the paper. To recover the 3D physical environment from the 2D visual measurement, the mapping between image plane and the robot-centered frame is investigated firstly. The artificial markers, employed as system observation, are recognized by means of the grey value variance-based method. To achieve the system consistency, the robot pose and visual feature state are inferred simultaneously by wrapping them in a single joint Posterior. Furthermore, the Monte Carlo sampling technique is introduced to our data fusion framework to relieve the linearized errors which is not sound tackled by the parameterized filtering. Finally, the experiments demonstrate the outperformance of the proposed platform over the traditional method.