Optimizing Q-Learning-Based Access Control Scheme Based on Q-Table Compression Method

This paper proposes a Q-table compression method to reduce in-access memory and resource usage. In the proposed method, the Q-table is sliced into smaller blocks called Q-state and indexed according to the state in the Q-table. Thereafter, each Q-state data are then encoded into a codeword of sequence. The performance of the proposed compression method is evaluated through a computer simulation. It is confirmed that using the proposed compression method, a smaller resource (i.e., overall processing time, and in-memory access) is used to update the Q-table.