Thank you for your comment.

One should estimate the elasticity and demand function from historical data.

But since it is not the scope of the post, in our example we assume that the elasticity and demand function are known.

In particular, we worked under linear assumption: d(Volume) = Elasticity * d(Price) + error.

As for the multiple opposing objectives, they could be modeled in both the state and reward mechanism for representation/learning.

Hope this helps.