Research On Key Technologies Of Robot Localization Based On 2D Map Match

Robot localization is always considered as one of difficult problems in a autonomous robot's navigation. In fact, only some local information in many applications can be observed by robots with exterior sensors, so the solution of localization based on local information has a broader significance in engineering fields. The paper systematically analyses the basic theory of localization algorithms based on 2D map match, which fast and practical approach not only realizes the robot localization of indoor environment, but also in outdoor environment, for the global map can be provided by GPS, so the application area is extended. The main results of my dissertation in the localization of robots are as follows:In classical probabilistic models, there is a representation for a straight edge in 2D using vector V_AF = (a,b)^T, parameters of the line equation: y=ax+b. This representation has a singularity for lines parallel to the Y axes. This forces to have an alternate representation for such lines, with x=ay+b, introducing more complexity in the estimation equations. In general, near singularities, covariances tend to infinity, and the precision of computations made with them decreases drastically. All the drawbacks leads us to define a new probabilistic model: the Symmetries and Perturbations Model(SPModel) that allows to represent the location of any type of geometric feature or sensorial observation and its uncertainty. Location information in SPModel was defined as four parts: location vector, an uncertain information of location, covariance matrix, bound matrix. An uncertain information of location and a bound matrix is defined as a random perturbation vector, and the product of the relative location vector and random perturbation vector is defined as fusion model. It is applied to the estimated ultrasonic feature paired with the sensor readings. Then the near singularities is solved , the fusion of the both is realized here.Assuming motion of the target will be considered as a constant acceleration motion, the state space equation of a stochastic system is derived, and the uncertain parameters of...