Learning robot tasks with loops from experiences to enhance robot adaptability

Learning the Scope of Applicability for Task Planning Knowledge in Experience-Based Planning Domains

“…table) is an abstract operator for the concrete operator (pick ?hoist ?block ?table ?location). More in [Mokhtari et al, 2017b].…”

“…From Listing 1, the constants are replaced with variables (Generalization), some actions are excluded from the abstract plan (Operator abstraction), abstract actions are associated with features (Feature extraction), and repetitive abstract actions with same features form loops (Loop detection). See [Mokhtari et al, 2017b].…”

“…See [Mokhtari et al 2017a;2017b] for the algorithms of learning activity schemata and task planning in EBPDs.…”

Special issue on user profiling and behavior adaptation for human-robot interaction

“…When dealing with modeling and learning of the user preferences with respect to trajectories and tasks, the typical approaches rely on the so-called learning by demonstration methods, where the user teaches the robot by providing examples. For example, Mokhtari et al [6] analyze the problem of learning robot task models in the case of loops. Such loops may correspond, for example, to a sequence of actions that is successively applied to different objects.…”

“…This work presents a model to detect and represent loops of actions in a task demonstration and to adapt and exploit loops to different instances of the same class of tasks. While in [6] , in order to conceptualize the task, the robot is provided with step-by-step operating instructions, indirect teaching can also be achieved by letting the robot relying only on user feedback. In this issue, two contributions deal with learning processes that rely on the user feedback.…”