Expertness based cooperative Q-learning
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
| Hi-index | 0.00 |
Four knowledge sharing techniques based on fuzzy-Q learning are investigated in this paper. These knowledge sharing techniques are ‘Shared Memory', ‘Adaptive Weighted Strategy Sharing', ‘Exploration Guided Method', and ‘Greatest Mass Method'. Different robot expertness measures are applied to these knowledge sharing techniques in order to improve learning performance. We proposed a new robot expertness measure based on regret evaluation. The regret takes uncertainty bounds of two best actions, i.e. greedy action and the second best action, into account. Simulations were performed to compare the effectiveness of the three expertness measures i.e. expertness based on accumulated rewards, on average move and on regret measure, when applied to different sharing techniques. Our proposed measure resulted in better performance than the other expertness measures. Analysis and comparison of different knowledge sharing techniques are also provided herein.