VizRob: Effective Visualizations to Debug Robotic Behaviors

Abstract

Building and debugging robotic programs is known to be difficult. The robotic community has produced numerous tools, APIs and middlewares to help debug and trace the construction and execution of robotic behaviors. However, most of available debugging tools are text and log-oriented, leading to a tedious and ad-hoc debugging activity. In this paper we fully describe VizRob, a tool to debug robotic behaviors using logs and execution time. VizRob produces interactive visualizations built from log traces within a state machine model, that is, the visual representation of the behavior. VizRob is founded on deficiencies we empirically found from semi-structured interviews and a revision of tutorial materials. A small case study received an initial feedback of VizRob in a robotic software engineering team. Our case study shows: (i) VizRob helps engineers solve intricate debugging scenarios and (ii) engineers perceive VizRob as filling a relevant gap within the current tools for building robotic behavior.