Active perception for person tracking

Active perception is the ability of an agent to take actions to reduce its uncertainty when it is uncertain about the world it is acting in. This thesis tackles the challenge of active perception for tracking people in multi-camera networks. Multi-camera systems are routinely used for security, surveillance and person tracking. A key challenge in the design of such networks is the efficient allocation of scarce resources such as the bandwidth required to communicate the collected data to a central server, the CPU cycles required to process that data, the energy costs of the entire system or the manpower required to manually monitor all the collected data. Maintaining surveillance is an example of an active perception task where an agent must select k out of the n available cameras to allocate the scarce resources to minimize its uncertainty about the state of the world. To this end, in this thesis, we propose multiple methods and results for resource allocation in multi-camera networks, that in principle, enable an agent to select k cameras out of n such that it reduces the agent's uncertainty about the position of each person in the scene.