User-friendly security robots

Robotic systems are today capable of performing patrolling and surveillance tasks in indoor structured environments. However, they need to be designed by taking into account the operational environment and the specific task to be accomplished. This dependency from the specific features of task and environment (contextual information, according to Turner [11]), severely restricts their practical deployment. In this paper, we address the creation of semantic maps for the effective deployment of surveillance robot in an a priori unknown indoor scenario. More specifically, we propose an architecture based on multimodal human-robot interaction that allows the user to specify in a natural way, knowledge about the environment that is represented in a semantic map. Given the acquired knowledge, the robot is then capable to plan and execute an effective patrolling of the scenario.