The Q(λ) algorithm based on heuristic reward function

For reinforcement learning often show slow convergence speed problem in continuous and complex tasks, this paper proposes a Q(λ) algorithm based on heuristic reward function-Q(λ)-HRF algorithm. This algorithm can extract features from the environment and get the heuristic information, which can be applied to the study by Agent in the form of reward function, which can accelerate the convergence speed significantly. We also proved the convergence of the algorithm by mathematical way, and applied the algorithm to the Maze platform, the experimental results show that: the Q(λ)-HRF algorithm has better convergence speed than Q(λ) algorithm.