It’s not practical. Crashing cars is expensive and you need consistent data. You cannot use a different dummy on each test and expect to make sense of the data.

It’s not practical. Crashing cars is expensive and you need consistent data. You cannot use a different dummy on each test and expect to make sense of the data. The best we could do is come up with an acceptable model and use that for all/most tests. If we could accurately model crashes in software then there would be no limitation on doing thousands of crashes with all different dummy types.