1. Computer Vision Group Prof. Daniel Cremers Autonomous Navigation for Flying Robots Lecture 5.1: State Estimation Jürgen Sturm Technische Universität München

2. World State (or System State)  World State (real state of the real world) Jürgen Sturm Autonomous Navigation for Flying Robots 2  Belief State (our belief/estimate of the world state)

3. State Estimation What parts of the world state are (most) relevant for a flying robot?  Position  Velocity  Obstacles  Map  Positions and intentions of other robots/humans …… Jürgen Sturm Autonomous Navigation for Flying Robots 3

4. State Estimation  Cannot observe world state directly  Need to estimate the world state, but how?  Infer world state from sensor observations  Infer world state from executed motions/actions Jürgen Sturm Autonomous Navigation for Flying Robots 4

5. Sensor Model  Robot perceives the environment through its sensors  Goal: Infer the state of the world from sensor readings Jürgen Sturm Autonomous Navigation for Flying Robots 5 world statesensor reading sensor model (observation function)

6. Motion Model  Robot executes an action (or control) (e.g., move forward at 1m/s)  Update belief state according to motion model Jürgen Sturm Autonomous Navigation for Flying Robots 6 current state previous state motion model executed action

7. Probabilistic Robotics  Sensor observations are noisy, partial, potentially missing  All models are partially wrong and incomplete  Usually we have prior knowledge Jürgen Sturm Autonomous Navigation for Flying Robots 7

8. Probabilistic Robotics  Probabilistic sensor models  Probabilistic motion models  Fuse data between multiple sensors (multi-modal)  Fuse data over time (filtering) Jürgen Sturm Autonomous Navigation for Flying Robots 8

9. Lessons Learned  World state vs. (internal) belief state  Sensor and motion models  Model uncertainty using probability theory  Next: Recap on Probability Theory Jürgen Sturm Autonomous Navigation for Flying Robots 9