Program Synthesis by Examples for Object Repositioning Tasks

We deal with the problem of synthesizing human- readable computer programs for robotic object repositioning tasks based on human demonstrations. A domain specific language (DSL) is introduced for generic robotic object reposi- tioning tasks. A learning algorithm is designed and implemented to synthesize a generalized program in this DSL based on human demonstrations. Experiments show that a great variety of object repositioning tasks can be programmed through human demonstrations using our system. Synthesized programs can be executed in simulation for verification and refinement and then on a robot to accomplish the corresponding learned tasks in the physical world.