Representation, Reasoning, and Learning for a Relational Influence Diagram Applied to a Real-Time Geological Domain

Mining companies typically process all the material extracted from a mine site using processes which are extremely consumptive of energy and chemicals. Sorting the rocks containing valuable minerals from ones that contain little to no valuable minerals would effectively reduce required resources by leaving behind the barren material and only transporting and processing the valuable material. This paper describes a controller, based in a relational influence diagram with an explicit utility model, for sorting rocks in unknown positions with unknown mineral compositions on a high-throughput rock-sorting and sensing machine. After receiving noisy sensor data, the system has 400 ms to decide whether to divert the rocks into either a keep or discard bin. We learn the parameters of the model offline and do probabilistic inference online.