Continuous integration as mandatory puzzle piece for the success of autonomous vehicles 2020-01-0087
The transition to autonomous driving technology is is widely discussed topic today. In order to make autonomous vehicles work safely in the long run it will be a necessity to keep their software up to date at any time. Continuous integration methods need to get implemented into the automotive system development in order to keep up with the pace needed to make the new technology a success, and accepted by the users. With today’s traditional release methods vehicle updates are not deployed fast enough, a newly discovered corner case or glitch could restrict the usage of entire fleets for long time. In order to achieve turn around times measured in hours and not in weeks a sophisticated end to end continuous integration and validation process is needed on the highest system integration level. The development process has to contain smart branching strategies for fast turn around, it is mandatory to have a frozen and stable branch to release hotfixes in case of need, a validation branch with feature lock in order to stabilize, and a feature branch heavy development space that is supported by full system regression testing from the very beginning. The change content for validation per test execution has to be limited to minimum in order to support fast issue identification and root cause analysis.
This paper discusses possible solutions to provide full end to end continuous integration to the automotive system development.