(Deep) Reinforcement Learning allows us to handle a real-world environment and have an agent interact with that environment. The agent sees the current state of the environment and is then able to make a decision as to what to do next. The process is then repeated as the environment changes. This process is behind some of the most powerful AI breakthroughs in recent years including beating the world champion at Go. It is also the baisis behind the idea for self-driving cars.

In the Lab we are using (Deep) Reinforcement learning to:

Tracking

We are using Deep Reinforcement Learning to track moving animals. Understanding the way dolphins live and interact is constrained by the fact that we can often only observe them from fixed locations (boats or the shore) our work here is to develop an autonomous underwater vehicle which can track and follow a dolphin.

Searching

Drones have been used very successfully in recent years for searching through their ability to move around. However, in order to fully exploit the benefits of drones they need to fly autonomously. Here we are developing autonomous drones which can search within an environement.

Modifying a system

We can also use Deep Reinforcement Learning to modify how a system operates. For example how the energy collected from solar pannels is used within the home energy network. Is it used directly, stored in battery or sold back to the power company?