A Framework with a Pedestrian Simulator for Deploying Robots into a Real Environment

摘要

We describe a simulation framework aimed to develop and test robots before deploying them in a real environment crowded with pedestrians. In order to use mobile robots in the real world, it is necessary to test whether they are able to navigate well, i.e. without causing safety risks to humans. This task is particular difficult due to the complex behavior pedestrians have towards each other and also towards the robot, that can be perceived either as an obstacle to avoid or as an object of interest to approach for curiosity. To overcome this difficulty, our framework involves a pedestrian simulator, based on a collision avoidance model developed to describe low density conditions as those occurring in shopping malls, to test the robot's navigation capability among pedestrians. Furthermore, we analyzed the behavior of pedestrians towards a robot in a shopping mall to build a human-to-robot interaction model that was introduced in the simulator. Our simulator works as a tool to test the level of safety of robot navigation before deploying it in a real environment. We demonstrate our approach showing how we used the simulator, and how the robot finally navigated in a real environment.