Here, we are always sampling direct illumination and indirect illumination. How do we prove that this approach is unbiased?
Because we are using Russian Roulette with an unbiased Monte Carlo estimator, we know this approach is unbiased.
For the recursive case, why are we not using EstRadiance (i.e. the function that includes light emitted by the next object in path) in place of AtLeastOneBounceRadiance? I though the emitted light was included in the rendering equation?
P.S. Thanks @rishiu for answering my previous question!