Following Assembly Plans in Cooperative, Task-Based Human-Robot Dialogue

Abstract

The JAST dialogue system allows a human and a robot to jointly assemble construction toys on a common work area.Supporting this type of dialogue requires that the system have a representation of assembly plans that permits it both to discuss the details of the plan and to monitor its execution.We present a conceptual representation of assembly plans based on AND/OR graphs, and then describe how the dialogue manager uses these plans as the basis for a range of strategies for jointly carrying out the plans with the user.