Research On Sound Source Localization Model For Robot Under Uncertain Conditions

About the sound source localization of mobile robot based on microphone array, the factor that affects the accuracy of orientation is not only the environment noise, but also the robot pose. Besides, the unpredictability of robot pose, that is microphone array pose, directly causes the uncertainty of time difference of arrival, thus the results of orientation. Considering the obtain of perceptual information limited by accuracy in practical application, sole static detection cannot get the source target distance information theoretically, combined the robot multiple active probes, thereby establishing the distance estimation models. Not only the orient model and perceptual information affect the accuracy of localization, but also the uncertainty of robot motion influences its distance. The thesis targets on localizing the source target in order to achieve the mobile robot’s precise localization under the uncertain conditions from the above, encircling the study followed by three aspects:First of all, the inverse modeling was built by analyzing sound propagation mechanism, setting the given sound location as the output, the time difference of arrival calculated by geometry and the tetrahedron array model as the input, and utilizing RBF neural networks applied to multiple input and output and non-linear systems. Sound source orientation model based on RBF neural network and tetrahedron array was proposed, then the theoretical simulation and experiments verified the applicability and reliability of the modeling.Secondly, the uncertainty of motion pose was studied by building the robot motion model and introducing particle filer algorithms of Monte Carlo probabilistic model. Utilizing the established RBF neural network model to study the impact of array pose uncertainty on the error of directional source was proposed.Finally, building the localization model under uncertain circumstances, the distance away from the target was acquired through robot twice directional motion. Thereafter, it respectively analyzed the impact on the distance error of localization source due to the error of robot motion and the directional source.