Jefferson Ridgeway
NASA Ames Intelligent Robotics Group (IRG)

Deployment of a Regression Testing System for Analysis and Visualization of Astrobee Localization
Mentor: Brian Coltin, IRG Group, brian.j.coltin@nasa.gov

The autonomous free-flying robot Astrobee, designed for the ISS, localizes on a sparse visual map in six degrees of freedom using an Extended Kalman Filter. Currently, the Astrobee team conducts numerous tests to ensure Astrobee's localization will succeed on the International Space Station (ISS). However, in testing different mapping and localization algorithms for Astrobee, the method of manual trial and error has been used against individual test cases to determine whether the change was a success. Hence, we are creating a regression testing system to utilize in lieu of trial and error on specific tests, which has proven to be ineffective. The regression testing system will automate the testing process and show the Astrobee engineers how their changes affect the localization system, as the EKF is highly dependent on small changes and provides a complex output challenging for humans to understand. Furthermore, this research seeks to analyze the results of the testing system and then visualize the results in easily understandable ways. The regression testing system is extensible and prepared to be used on new tests created on the ISS when Astrobee launches next spring.