Design and Development of a Hybrid Wheel-Climbing Warehouse Robot Based on Reinforcement Learning 

Abstract

The development of modern warehouse systems requires robots capable of autonomous navigation and obstacle traversal in operational environments. However, conventional wheeled robots still have limitations in overcoming small obstacles that can reduce their mobility. This study aims to design and develop a warehouse hybrid wheel-climbing robot based on Reinforcement Learning. The research employed the ADDIE (Analysis, Design, Development, Implementation, and Evaluation) method. The robot was developed using an Arduino Mega 2560, an HC-SR04 ultrasonic sensor, DC motors, and servo motors as the climbing mechanism. The experimental results showed a climbing success rate of 80%–90%, an average travel time of 18.80 s, an average speed of 13.87 cm/s, and an average climbing duration of 2.58 s. The implementation of Q-Learning  produced a simple navigation policy capable of selecting forward and turning actions based on environmental conditions. The results indicate that the integration of the climbing mechanism and Q-Learning  has the potential to improve the mobility of warehouse robots.