Amsterdam Machine Learning Lab
The goal of our research group is to develop, analyse, and evaluate the algorithms needed to build intelligent autonomous systems. Creating autonomous systems that are effective in diverse settings is a key goal of artificial intelligence with enormous potential implications: robotic agents would be invaluable in homes, factories, and high-risk settings; software agents could revolutionize e-commerce, information retrieval, and traffic control.
There are, however, three main challenges in building such autonomous systems. These challenges involve perception, action, and interaction. Such systems observe their environment via cameras and other sensors. Perception is the process by which the system tries to understand these observations and form a meaningful picture, not only of the current state of the environment, but its uncertainty about it. Using the results of perception, the system must then reason about what actions to take to achieve its goals most efficiently. Since the environment often contains people and the system’s goals are typically to meet those people’s needs, it must also be able to interact with people effectively, by understanding their behaviour and responding in a useful way.
Currently, the group is developing tools for autonomous systems that enable robots to play football; helicopters to fly autonomously; cars to recognize pedestrians and brake in time to avoid them; and smart cameras to detect aggressive behaviour in a train station or determine whether an elderly person has fallen over at home.