In the non-deterministic, large-scale navigation environment under the Mars exploration mission, there is a large space for action and many environmental states. Traditional reinforcement learning algorithms that can only obtain rewards at target points and obstacles will encounter the problems of reward sparsity and dimension explosion, making the training speed too slow or even impossible. This work proposes a deep layered learning algorithm based on the goal-driven layered deep Q-network (GDH-DQN), which is more suitable for mobile robots to explore, navigate, and avoid obstacles without a map. The algorithm model is designed in two layers. The lower layer provides behavioral strategies to achieve short-term goals, and the upper layer provides selection strategies for multiple short-term goals. Use known position nodes as short-term goals to guide the mobile robot forward and achieve long-term obstacle avoidance goals. Hierarchical execution not only simplifies tasks but also effectively solves the problems of reward sparsity and dimensionality explosion. In addition, each layer of the algorithm integrates a Hindsight Experience Replay mechanism to improve performance, make full use of the goal-driven function of the node, and effectively avoid the possibility of misleading the agent by complex processes and reward function design blind spots. The agent adjusts the number of model layers according to the number of short-term goals, further improving the efficiency and adaptability of the algorithm. Experimental results show that, compared with the hierarchical DQN method, the navigation success rate of the GDH-DQN algorithm is significantly improved, and it is more suitable for unknown scenarios such as Mars exploration.

Although broad reinforcement learning (BRL) provides a more intelligent autonomous decision-making method for the collision avoidance problem of unmanned surface vehicles (USVs), the algorithm still has the problem of over-estimation and has difficulty converging quickly due to the sparse reward problem in a large area of sea. To overcome the dilemma, we propose a double broad reinforcement learning based on hindsight experience replay (DBRL-HER) for the collision avoidance system of USVs to improve the efficiency and accuracy of decision-making. The algorithm decouples the two steps of target action selection and target Q value calculation to form the double broad reinforcement learning method and then adopts hindsight experience replay to allow the agent to learn from the experience of failure in order to greatly improve the sample utilization efficiency. Through training in a grid environment, the collision avoidance success rate of the proposed algorithm was found to be 31.9 percentage points higher than that in the deep Q network (DQN) and 24.4 percentage points higher than that in BRL. A Unity 3D simulation platform with high fidelity was also designed to simulate the movement of USVs. An experiment on the platform fully verified the effectiveness of the proposed algorithm.