Encoding The Sensor Allocation Problem For Reinforcement Learning