Enabling End-Users to Create Real-World Robot Applications through Visual Programming Interfaces and Automation

Thesis (Ph.D.)--University of Washington, 2020Interactive robots in human environments have the potential to improve our daily lives by providing education, eldercare, entertainment services, and more. However, creating new applications and behaviors for such robots is complex for many reasons, including the challenges of prototyping and testing. This thesis first presents findings such as user requirements, constraints, and design implications from designing and evaluating two robot applications; information-gathering application for a mobile robot in a university building and customer feedback collection application for hotels. Our goal is to enable non-experts, such as salespeople and designers, to create and explore new robot applications without requiring additional technical expertise. This thesis presents two consecutive field studies for evaluating a non-expert friendly robot programming system for creating socially interactive robots in commercial spaces. Based on these studies, expressing concurrency and modifying low-level behaviors without scarifying ease of use are identified as the two important requirements for interactive robot programming systems. This thesis addresses the identified requirements by presenting (1) concurrency interfaces for block-based visual robot programming and (2) an iterative, corrective feedback-driven program repair approach for adjusting low-level robot behaviors. ConCodeIt! is a framework for identifying programming constructs required to express concurrency. Three representative concurrency interfaces were implemented investigated for a block-based visual robot programming system to assess ease-of-use and expressiveness via a systematic evaluation and a user study. Iterative Bayesian Repair enables programmers or the robot's users to change details that govern a robot's low-level behaviors through interaction feedback. Simulation experiments and a user study demonstrated the effectiveness of improving the fluency of interactive robot behaviors