Multisensory Data Fusion for Autonomous Vineyard Sprayer Robot Navigation

This thesis is a research for localization problem of an autonomous vineyard spraying robot. The project is motivated by the lack of manpower to perform the vineyard spraying task and the exposure to hazardous pesticides during the spraying. Localization is a fundamental problem in the field of autonomous mobile robotics. In order to allow basic autonomous functionality of a robot, Position and Path-Planning control loops are needed. Both loops require knowledge of the accurate state (position orientation and velocity) of the robot This research focus in modeling the kinematic model of the robot. The model will be later use for controller design and model based state estimation (filtering). Second, we research how to filter a noisy raw sensor data and the integration of data from all sensors (DGPS, IMU and Vision). The research focus on statistical and probability theory such as Bayesian estimator and Kalman filter for filtering and data fusion. The simulations of Data Fusion presented in this thesis show major improvement for the state estimation errors. The main contribution of this thesis ere the set of tools and methodologies to design low cost navigation system with high precision of localization and a novel algorithm for data fusion using likelihood ratio test as decision making technique for choosing the most probable solution. Each sensor is pre-filtered according to its noise distribution, not like other methods for data fusion i.e. Kalman filter. Also, this algorithm can also be used for many other applications where decision making is needed This thesis published in report for the Chief Scientist, Ministry of Agriculture [1] and it is a basis for further research for examination of the data fusion algorithm, its robustness and real-time implementation.