This thesis explores how a low-cost robotic platform can be used to prototype and evaluate the core functionalities of a robotic guide dog. The project investigates whether affordable robots can support independent mobility for visually impaired people and how such platforms can serve as accessible research tools in assistive robotics.
An analysis of guide-dog behaviour identified four essential functionalities: obstacle avoidance, speech recognition, navigation towards recognised objects, and communication with the handler. Due to the small scale of the chosen robot, the latter was not implemented. The Dogzilla S2, a low-cost quadruped, was selected for its affordability, sensing capabilities, and suitability for safe human interaction. Using a modular ROS2 architecture, separate subsystems for perception, navigation, and speech were integrated into a unified framework.
Testing in controlled scenarios showed that the robot could execute basic guiding behaviours using lightweight recognition models. Although limited in sensing accuracy, the system demonstrated that meaningful functionalities can be achieved on inexpensive hardware.
The results highlight that low-cost platforms enable rapid, affordable experimentation and can guide the development of future high-end robotic guide dogs. This approach lowers the threshold for research, supports collaboration, and helps balance innovation with accessibility, bringing the vision of intelligent, reliable, and inclusive mobility assistance one step closer to reality.