Evaluating Methods for End-User Creation of Robot Task Plans

Abstract

How can we enable users to create effective, perception-driven task plans for collaborative robots? We conducted a 35-person user study with the Behavior Tree-based CoSTAR system to determine which strategies for end user creation of generalizable robot task plans are most usable and effective. CoSTAR allows domain experts to author complex, perceptually grounded task plans for collaborative robots. As a part of CoSTAR's wide range of capabilities, it allows users to specify SmartMoves: abstract goals such as "pick up component A from the right side of the table." Users were asked to perform pick-and-place assembly tasks with either SmartMoves or one of three simpler baseline versions of CoSTAR. Overall, participants found CoSTAR to be highly usable, with an average System Usability Scale score of 73.4 out of 100. SmartMove also helped users perform tasks faster and more effectively; all SmartMove users completed the first two tasks, while not all users completed the tasks using the other strategies. SmartMove users showed better performance for incorporating perception across all three tasks.

Prospection: Interpretable plans from language by predicting the future

“…Communicating control and goals has traditionally been accomplished by specifying high level operations [5], [1], [6], via formal languages like the Problem Domain Description Language (PDDL) [7] or as a Hierarchical Task Network [8]. Such systems provide a straightforward way to compose black-box operations to solve problems.…”

Intuitive Task-Level Programming by Demonstration Through Semantic Skill Recognition

“…Skills are executable programs with logic for e. g. strategies and failurehandling designed for solving specific tasks (such as pegin-hole), and are crafted by an expert. This logic may be represented in various forms, such as Behavior Trees [14], state machines [3] and specialized controllers [11].…”

“…Task-Level Programming builds on the skills developed for a robot, by providing user interfaces (e. g. a GUI) to nonexperts to create tasks consisting of skills [1]- [3], [11], [14], [15]. These interfaces enable users to sequence the appropriate skills and specify relevant parameter values in order to tailor them to the task at hand.…”

“…The work described in this paper builds on RAZER [3], our framework for TLP. Another example is CoSTAR, which utilizes Behavior Trees [14]. CoSTAR lowers the complexity for the user by increasingly integrating the logic into the available building blocks.…”

The CoSTAR Block Stacking Dataset: Learning with Workspace Constraints

“…4. A dataset summary can be found in Table I. Data is collected utilizing our prior work on the collaborative manipulation system CoSTAR [3], [35]. CoSTAR is a system designed for end-user creation of robot task plans that offers a range of capabilities plus a rudimentary perception system based on ObjRecRANSAC.…”