Route searching based on neural networks and heuristic reinforcement learning

Area of Science:

• Artificial Intelligence
• Machine Learning
• Robotics

Background:

• Reinforcement learning (RL) faces challenges with increased state spaces and limited environmental information.
• Efficient route searching is critical in complex environments.

Purpose of the Study:

• To propose an improved RNH-QL method for enhanced route searching in large state spaces.
• To address the inefficiency of traditional RL in complex environments.

Main Methods:

• Utilized a Radial Basis Function (RBF) network for weight updating.
• Implemented heuristic Q-learning with reward shaping for agent guidance.
• Incorporated a greedy exploitation strategy to train the neural network.

Main Results:

• The RNH-QL method demonstrated improved efficiency in large state spaces.
• Reward shaping effectively guided agents towards goal states.
• Experimental results validated the enhanced learning efficiency.

Conclusions:

• The RNH-QL method offers a robust solution for route searching in complex, large-scale environments.
• The integration of RBF networks and Q-learning enhances learning efficiency and agent guidance.