A Novel Reinforcement Learning Architecture for Continuous State and Action Spaces

Figure 9

Comparison of the reliability of the policies found with the SARSA Actor-Actor-Critic algorithm and the 𝑄(𝜆)-learning algorithm.