Automated trace analysis of test traces

Abstract

In current development of Parking Driver Assistant System, function developers get a lot of traces from various testing instances. Very often the issue is already known, and it is more or less a known standard error in the related SW-release but nevertheless the function developers have to do a time wasting trace analysis. This should be done automatically in future. So when function developers get a new trace there should be a filter which detects if it is a known error or if not. There are 3 main parts for this topic:

1. Trace analyser: The traces obtained from the bussystems like ‘Ethernet’ and ‘Flexray’ should be filtered to separate the known errors. One approach is to use CANOe tool with CAPL programming language. The other approach is pattern recognition. The known errors found using the pattern recognition approach are stored in a database. 2. User Interface: The user interface has been developed in Python. 3. System administration: The system administrator can train the system for new “standard errors” in an easy way. The intention is not to bring new errors by reprogramming the complete tool but to have an interface for learning new bugs.

This is a new filtering method which is not developed until now; not that we are aware of. Project is integrated in an overall project for testing strategy in driver assistant systems at Daimler AG in Sindelfingen together with Mercedes-Benz R&D India.