Mission Planning for Autonomous Systems

Path Planning: Recap from differential geometry and numerical analysis, including properties of parametric 2D/3D curves, curvature and torsion, interpolation/approximation. Overview of vehicle kinematic and dynamic constraints and their connection to path planning. Basic path generation for connecting waypoints: Dubins paths, Reeds-Shepp car, splines, Pythagorean Hodographs, spirals. Obstacle representation. Roadmap methods for generating waypoints: Voronoi diagrams, probabilistic roadmaps, RRTs. Graph-search algorithms for dynamically computing the optimal sequence of waypoints. Optimal control and optimization approaches. Potential fields.

AI planning: Introduction to Markov Decision Processes (MDPs) and solutions based on dynamic programming and reinforcement learning. Deliberative vs reactive behaviour. Fundamental automated planning methods and algorithms (state machines, STRIPS). Hierarchical task networks (HTNs). Temporal models.

The course will also include one lecture with an introduction to the Robot Operating System (ROS).

Knowledge: Detailed knowledge about path planning and AI planning. Be able to read and understand methods published in the literature and evaluate and compare these with methods used in practical systems.

Skills: Design and implement path- and action planning systems for ships, underwater vehicles, and aerial vehicles. Be able to simulate fundamental path- and action planning approaches, and their main variations, on such systems, including implementations on a real-world small-scale mobile robot and an autonomous passenger ferry. Independent management of small R&D projects and contribute actively in larger projects.

General competence: Communicate work related problems with specialists and nonspecialists.

Lectures and five mandatory computer assignments, supported by guidance sessions. The objectives of the assignments are to simulate and test self-developed path- and action-planning methods for marine and ground vehicles.

Computer assignments 1-3 (CAs 1-3) are evaluated as pass/fail and include simulations. Computer assignments 4 and 5 involve designing and implementing missions on two robots in the real world: A small scale mobile robot, and the full-scale passenger ferry milliAmpere1. CAs 4-5 each count for 20% of the total grade.