SLAM Foundations: How Robots Map the World and Find Themselves In It

Robots and autonomous vehicles need to know where they are and what is around them, often without any prior map of the environment. Solving that chicken-and-egg problem is the heart of SLAM, Simultaneous Localization and Mapping. This course gives you a calm, structured introduction to the field so you can read SLAM papers, evaluate stacks, and reason about their behavior. You will learn what SLAM does, what it cannot do, and how different sensors and algorithms make different tradeoffs. The course stays grounded in widely used concepts and points to the modern advances reshaping the field. What you'll learn: - Understand the SLAM problem and why it is fundamentally a joint estimation challenge - Recognize the main sensor modalities used in SLAM including LiDAR, cameras, depth sensors, and inertial units - Explore the major algorithm families including filtering approaches and modern graph-based optimization - Read how loop closure detection turns drifting estimates into globally consistent maps - Identify the limits of SLAM including environment changes, sensor failures, and ambiguous scenes - Understand the integration points between SLAM, motion planning, and higher-level robot behavior The course begins with the SLAM problem itself, moves through sensors and algorithms, and closes with the operational realities of running SLAM on real robots. Written exercises help you connect each concept to a specific robot, environment, or use case. This course is designed for absolute beginners with no robotics or SLAM background, including computer science students, software developers exploring robotics, and engineers entering autonomous systems work. No prerequisites are needed beyond general comfort with mathematics. The course explains every concept as it appears and stays focused on understanding rather than implementation.