I think he has to sample his position deltas from a gaussian with mean 0 and variance equal to the time since last update for it to be a valid brownian motion.

See https://en.wikipedia.org/wiki/Brownian_motion#Mathematics