How noisy sensor measurements can be interpreted in time? How can autonomous systems (in the virtual or in the real world) use their noisy measurements to learn the best course of actions (policy) to take? This topic teaches how to get robots to incorporate uncertainty into estimating and learning from a dynamic and changing world. Robust estimation techniques are employed to produce values that tend to be closer to the true values of the measurements and their associated calculated values and are an essential part of the development of advanced navigation systems and principles of machine learning (mainly reinforcement learning and deep reinforcement learning) will be applied to optimize the agent's behaviour in its environment. In other words, the syllabus includes probabilistic algorithms, Markov decision processes, Bayesian filtering and reinforcement learning.
This topic aims to introduce students to the fundamental principles of theory and practice of estimation and machine learning applied on the interpretation of noisy sensor data from any source. The performance of the sensors, including precision, accuracy, repeatability, sensitivity, linearity and dynamic performance, are analysed. Calibration and estimation techniques are also examined. Reinforcement learning and Deep Reinforcement Learning will be introduced as a means to take the students to the state-of-the-art of modern intelligent systems and sensor interpretation.