If AURORA rolls out a new perception algorithm, how would you statistically prove that it reduces the rate of near-miss incidents?
