ICARSC'2017 - Vahid Mokhtari

Abstract: We extend our previous work on experience-based planning domains in robotics to detect and represent loops in actions sequences that achieve certain tasks; and to generate action sequences from loop descriptions. The approach includes methods for conceptualizing robot experiences containing loops; and generating plans based on learned task knowledge with loops. This facilitates the reuse of existing task knowledge for different instances of the same task with varying numbers of objects. A demonstration of this system in a real environment illustrates how a robotic arm can learn and carry out a task.