Autonomous driving is a rapidly growing field and can bring significant transition in mobility and transportation. In order to cater a safe and reliable autonomous driving operation, all the systems concerning with perception, planning and control has to be highly efficient. MPC is a control technique used to control vehicle motion by controlling actuators based on vehicle model and its constraints. The uniqueness of MPC compared to other controllers is its ability to predict future states of the vehicle using the derived vehicle model. Due to the technological development & increase in computational capacity of processors and optimization algorithms MPC is adopted for real-time application in dynamic environments. This research focuses on using Model predictive Control (MPC) to control the trajectory of an autonomous vehicle controlling the vehicle actuators for high dynamic maneuvers. Vehicle Models considering kinematics and vehicle dynamics is developed. These models are used for MPC as prediction models and the performance of MPC is evaluated. MPC trajectory control is performed with the minimization of cost function and limiting constraints. MATLAB/Simulink is used for designing trajectory control system and interfaced with CarMaker for evaluating controller performance in a realistic simulation environment. Performance of MPC with kinematic and dynamic vehicle models for high dynamic maneuvers is evaluated with different speed profiles.

[pt] Veículos autônomos são um problema recente, com aplicação em carros e motocicletas ainda nos estágios iniciais. Além das dificuldades inerentes de fazer um veículo mover-se independentemente, a motocicleta autônoma deve permanecer estável em qualquer velocidade e trajetória. O objetivo principal deste trabalho é desenvolver uma motocicleta elétrica autônoma com sistema de instrumentação de baixo custo. Para tanto, foi analisado um modelo dinâmico de motocicleta, capaz de reproduzir o comportamento real e permitindo a implementação de estratégias de controle linear em tempo real. O controlador tem dois objetivos diferentes: manter a motocicleta estável e seguir uma trajetória desejada, de forma autônoma. Experimentos foram realizados com a motocicleta de escala reduzida com o objetivo de caracterizar seus elementos; as estratégias de controle propostas foram simuladas com o modelo dinâmico ajustado. Por fim, os algoritmos de controle são aplicados ao sistema real através de uma plataforma atuada capaz de reproduzir a dinâmica de veículos de duas rodas. O presente trabalho é uma ferramenta para o ensino de engenharia, envolvendo estudantes de diferentes níveis em torno de um problema complexo. O sistema permite uma aprendizagem contínua com dificuldade crescente, envolvendo temas como dinâmica de multicorpos; análise de resultados através de simulações de software; eletrônica e filtros na instrumentação embutida e técnicas de controle para manter o sistema estável em todos os caminhos desejados, culminando na aplicação experimental dos conceitos citados. / [en] Autonomous vehicles are an interesting and recent problem, with its application in cars and motorcycles still in its early stages. In addition to the inherent difficulties in making a vehicle move independently, the autonomous motorcycle has to be able to remain stable at any speed and trajectory. The vehicle s stability can be achieved by different solutions and control techniques. The main objective of this work is to develop an autonomous electric motorcycle with low cost sensing system. For this, a dynamic model of two-wheeled vehicles is analyzed, capable of describing the dynamic behavior while being simple enough to allow the implementation of real-time linear control strategies. The controller has two different objectives: to maintain the motorcycle stable and to follow a desired trajectory, in an autonomous way. Experiments were carried out with the small scale motorcycle aiming to characterize its elements for the theoretical model; then the proposed control strategies were simulated with the adjusted dynamic model. Finally, the control algorithms are applied to the real system through an actuated platform capable of reproducing the dynamic behavior of single-track vehicles. At last, the present work is a tool for teaching engineering, involving multilevel students around a complex, but familiar, problem. The system allows for continuous learning with increasing difficulty, involving multibody dynamics, experimental results analysis via software simulations, electronics and filters present in the embedded instrumentation and many control techniques to keep the system stable in every desired path, culminating in the experimental application of cited concepts.