Group for Research in Decision Analysis

Adaptive self-learning mechanisms for updating short-term production scheduling in an industrial mining complex

Ashish Kumar McGill University, Canada

A new method based on a fast and smart reinforcement learning framework is proposed to adapt and update the short-term production scheduling of an industrial mining complex with new information. The latter is first used to update the stochastic models of mineral deposits using ensemble Kalman filter. Subsequently, a combination of neural networks and policy gradient reinforcement learning is used to update the short-term production scheduling decisions. Results from a large copper-gold mining complex show that the proposed adaptive framework better meet production targets and consistently generates higher cash flows, when compared to the conventional approach.

Coffee and biscuits will be offered at the beginning of the seminar.
Welcome to everyone!