A robot should be able to follow or approach a selected person in an environment - indoor or outdoor - with static and dynamic obstacles. It would be fine if the robot could find its way from A2B by itself, but this (SLAM) problem is not an option here and we select a certain person to be approached. Identifying a person can be performed in many ways, but here we focus on an audio solution. Identifying based on human speech is possible, but to begin with, we let the human identify him self by sending out a specific audio signal already known by the robot - a beacon from Mr. B. Doing this, we can get around reflections in a simple way and also identify Mr. B. To get experience in the area, we will equip Mr. B with a ‘stick' sending out this beacon in regular time intervals. With two synchronous audio input channels and some computing, we are able to calculate the direction from the robot to Mr. B.
When the robot is moving towards Mr. B, some obstacles may occur such as people, dogs and other robots. The robot must avoid collision and eventually stop and wait for a free path towards the destination. The robot must obey social distance and therefore it is possible to highjack the robot. To prevent this, the robot must stop if it can't find a free path not increasing the distance to Mr. B.
|Effective start/end date||29/05/2008 → 19/09/2008|
- iRobot, Sound Source Location