Aceves López, AlejandroAguilar Aldecoa, Aldo Iván2022-09-262022-09-262021-05-27Aguilar Aldecoa, A. (2021). Design of a proprietary self driving car platform and development of autonomous driving algorithms based on computational vision and deep neural networks. Instituto Tecnológico y de Estudios Superiores de Monterrey. Recuperado de: https://hdl.handle.net/11285/649735https://hdl.handle.net/11285/649735This research project presents a detailed description of the design and development of the first small-sized self driving development platform at the ITESM Campus Estado de México. The implemented hardware and software is presented based on the state of the art research platforms. Additionally, the required sensor and instrumentation implementation is described as well as the platform's mechanic and electric design. Moreover, the dynamic identification of the vehicle actuators is presented for the linear velocity control of the platform however no clear evidence of a linear dynamic behavior could be identified, leading to the implementation of a herustically tuned PI velocity control system. Specific Computer Vision (grayscale color thresholding) and Deep Neural Network (U-Net semantic segmentation) based road lane segmentation mechanisms were developed, tested and validated. These segmentation mechanisms served as the main input of the final autonomous driving system proposal based on a road lane following strategy. Finally, a well-defined autonomous driving performance evaluation methodology is described and implemented to compare the proposed systems response, identifying comparable performances between CV and DNN segmentation systems.TextoengopenAccesshttp://creativecommons.org/licenses/by-nc-nd/4.0INGENIERÍA Y TECNOLOGÍA::CIENCIAS TECNOLÓGICAS::TECNOLOGÍA E INGENIERÍA MECÁNICAS::OPERACIONES MECANIZADASTechnologyTesis de maestríaroad lane followersmall-sized autonomous vehiclecomputer visionconvolutional neural networksroad segmentationproprietary platform