Dynamic Robot Navigation in Confined Indoor Environment: Unleashing the Perceptron-Q Learning Fusion.

Area of Science:

• Robotics
• Artificial Intelligence
• Machine Learning

Background:

• Robot navigation in confined spaces is challenging due to dynamic obstacles and reliance on pre-defined maps.
• Existing methods struggle with real-time path-planning in unpredictable environments.
• Minimizing moving costs and detours is crucial for practical robot navigation.

Purpose of the Study:

• To propose a novel perceptron-Q learning fusion (PQLF) model for efficient robot navigation.
• To enhance robot navigation in dynamic, confined indoor environments.
• To address limitations of existing methods in handling dynamic obstacles and reducing navigation costs.

Main Methods:

• Developed a perceptron-Q learning fusion (PQLF) model combining perceptron learning and Q-learning.
• Utilized robot sensors for dynamic obstacle distance determination during local path-planning.
• Modeled dynamic robot navigation as a Markov Decision Process (MDP) with the PQLF controller as the agent.

Main Results:

• The PQLF model achieved a reduced moving cost of 1.1.
• The proposed model resulted in a detour percentage of 7.8%.
• Simulation results demonstrated superior performance over existing robot navigation methods.

Conclusions:

• The PQLF model effectively enhances robot navigation in dynamic, confined indoor environments.
• The fusion of perceptron learning and Q-learning provides a robust solution for real-time path-planning.
• The model significantly improves efficiency by reducing moving costs and detours.