Enhancing Danger and Hazard Perception in Robotic Guide Dogs

About the Project

For many years, guide dogs have been incredibly helpful friends for people with vision impairments such as Blind or Low Vision (BLV), giving them the freedom and mobility they need to go about their everyday lives. However, there is a shortage of trained dogs, and the process of training requires a lot time and money. Moreover, maintaining guide dogs requires a lot of resources. Furthermore, not all people with vision impairments are able to provide the maintenance and care that a living animal needs. A viable substitute are robotic guide dogs, which combine the advantages of conventional guide dogs with cutting-edge technology to deliver dependable, constant aid without requiring a lot of upkeep or training.

Recent years have witnessed enormous advancements in robotic guide dogs, especially in the areas of obstacle avoidance and autonomous navigation. More progress has been made in the creation of quadruped robots, which can guide human through confined and crowded space. However, despite the advancements for developing a robot guide dog, there are few research gaps which requires extensive research such as the lack of danger and hazard perception capability of robot dog. Furthermore, there is a need for better integration of user feedback system for the hazard response between the human and robot guide dog.

Aims and Objectives

This PhD project aims to address these gaps by:

• Developing advanced sensor fusion techniques to enhance real-time hazard detection.
• Integrating machine learning algorithms to improve the ability to predict and react to dynamic hazards.
• Utilizing simulated environments for training the robot, leveraging generative AI and physics simulators to create diverse and realistic virtual training scenarios.
• Developing user feedback system between the robot and human for hazard response

Impact of the research:

The successful completion of this project will significantly advance the field of assistive robotics by:

• Providing visually impaired individuals with safer and more reliable mobility solutions.
• Methods for developing hazard perception for robot with machine learning and simulated environment
• Contributing to the broader field of autonomous systems and human-robot interaction.

For further information and any queries, please contact Dr. Miftahur Rahman - m.rahman@kingston.ac.uk