Testing of Robotic Systems

Challenges and Motivation

  1. There is no systematic way of generating test scenarios for robot applications.
  2. Robot applications are generally more complex in the number of scenarios compared to autonomous vehicles.
  3. Robot vulnerability identification is also a key issue. Nobody worked on this before.

Goals

  1. Non-determinism and Entropy in Robotic Applications/Software
  2. How effective are traditional coverage-based testing?
  3. A good metrics to evaluate mutation?
  4. A good strategy to test robotic applications and software?

Ideas

  1. Generating test cases based on tasks and requirements.
  2. Key insights:
    1. Coverage testing in robots means both line/function coverage and scenario coverage. Taking exploration task as an example: a good testing scenario should achieve best code coverage rate, while allow the robot to cover all the corner cases in the map.
  3. Use Turtlebot4 as an example to study.

Reading Materials