When Should a Service Robot Switch Tasks ? A Learning-Based Approach for Interrupting Task Execution

While mobile robots reliably perform each service task by accurately localizing and safely navigating avoiding obstacles, they do not respond in any other way to the people and their environment. One aspect that increases a robot’s responsiveness to its environment is the ability to interrupt the task execution and switch to another task based on observations. We present a novel approach that enables an autonomous robot to learn from observations, when to interrupt task execution and what task to switch to next. We test the effectiveness of our approach in a human-robot interaction scenario and show that our approach significantly decreases the amount of sensory input processing during the execution of tasks.