System and method for learning sequences in robotic tasks for generalization to new tasks

A robotic controller, comprising:

• a memory circuitry storing instructions;
• and
• at least one control processor configured to:
• acquire demonstration data of a plurality of demonstration tasks;
• segment the demonstration data of the plurality of demonstration tasks into multiple segments by selecting features from the demonstration data based on a feature selection method and using a segmentation metric, wherein each segment of the multiple segments represents a subtask associated with a corresponding demonstration task of the plurality of demonstration tasks, and wherein the segmentation metric represents a variance between different demonstrations and within an identical trajectory;
• generate dynamic movement primitives (DMPs) based on the multiple segments, wherein each dynamic movement primitive (DMP) of the DMPs corresponds to a corresponding segment of the multiple segments;
• generate a planning graph by searching a feasible path of at least one object for a desired task, wherein the desired task is represented by an initial state and a goal state with respect to the at least one object;
• parameterize the feasible path based on at least one DMP of the DMPs using the initial state and the goal state;
• and
• implement the parameterized feasible path as a trajectory of a robot by tracking and following the parameterized feasible path for the desired task.