Intelligent Autonomous Systems

Dynamical systems such as robots, manipulators and vehicles rely on several building blocks and software layers that enable the them to operate autonomously in complex environments. The collection of building blocks called autonomy stack and consisting of perception, planning and coordination, localisation and mapping, and control, define the brain of autonomous systems allowing it to perform complex tasks without human intervention or in human machine teaming applications. This course introduces students to the methodology and algorithms involved in the autonomy stack. Perception is discussed on the example of various sensors and in particular LIDAR sensors used for autonomous driving. Localisation and mapping is covered through particle filters and graph based SLAM. Rapidly-exploring random tree plans, spline based Planning and raceline optimisation are discussed to give examples on planning and coordination. Control is covered through reference tracking and path following algorithms building on planning and coordination with explicit examples on 'wall following' and 'follow the gap' for obstacle avoidance. Students will be challenged with the combination of different concepts in the the autonomy stack based on group based projects. Simulations, student projects and examples are based on the virtual environment Webots (https://cyberbotics.com). Since modern controller designs and various components of the autonomy stack rely on big data, additionally 'optimization in control' with applications in reinforcement and machine learning, and model predictive control, will equip students with a complete set of tools necessary to work on or to supervise intelligent autonomous systems projects.