Developing a navigation model for optimizing social comfort for side-by-side robotic wheelchairs

Abstract

Side-by-side robotic wheelchairs have significant direct benefits for users and lessening the burden on caregivers. The autonomous navigation function for side-byside robotic wheelchairs has attracted the attention of many researchers recently. The challenge is to enable side-by-side robotic wheelchairs not only to continuously maintain the side-by-side formation with caregivers, but also to find suitable methods to avoid collision with obstacles in the environments, and bring comfort to those sitting in the wheelchairs. This problem is more complex in crowded environments where robots have not only to navigate to destinations alongside caregivers, but also to avoid any moving obstacles and pedestrians. Moreover, the robots need the ability to move in harmony in crowded environments, respecting the comfort of the caregivers and surrounding people. Those capabilities are necessary for sideby- side robotic wheelchairs to become accepted in the daily activities of humans.

The main objective of this research is to develop a novel navigation model for side-by-side robotic wheelchairs in order to help them navigate as humans normally do in human environments.

First, the thesis presents the research background information, including human walking habits, prediction methods for indoor robot navigation, and recent related research results. To design a mobile robot capable of navigating in crowded environments, it is essential to understand how humans move, how they avoid collisions and maintain social relationship, how they interact with each other, and how they react in each circumstance. Based on that knowledge, the robot is able to predict the next movements and intentions of humans and objects in each situation, since then it can adjust moving plans to avoid collision, bringing safety and comfort to the wheelchair users, and maintaining a harmonious atmosphere with its partners and surrounding people.

In particular, my investigation shows that human walking plans depend not only on physical constraints such as architectures of the environments but also on their walking habits. When people can maintain their walking habits, they normally feel more comfortable. Therefore, to effectively anticipate human intentions and adjust accordingly, side-by-side robotic wheelchairs not only have to measure the physical factors, e.g. walking velocity, acceleration, etc., of partners and surrounding people but also have to consider factors related to human walking habits as well.

This thesis proposes the novel robot navigation methods that take into account the social interactions between people in various situations, especially in the scenarios of walking sessions with a side-by-side robotic wheelchair alongside a caregiver. In order to acquire more intelligence for the navigation algorithm, I propose to integrate seamlessly the factors related to human habits information into a new robot planning approach. In this study, two factors - Vision and Friendly Link factors - were discovered and modeled into the novel navigation models. Active mode was also discovered as a main walking mode of pairs. In addition, a method to determine the Preferred Walking velocity of pairs was proposed. All of these elements have been modeled and integrated into the new navigation model. I believe that, by applying the developed models in this study, side-by-side robotic wheelchairs can bring more comfort to wheelchair users, caregivers, and surrounding people, and therefore they can be more easily accepted in daily activities.

Performance evaluation is essential for the validation of the navigation models. In this thesis, I have implemented all the navigation algorithms in a simulation to measure the performance of our models based on real recorded data. The solutions proposed by the new models were compared with the previous models and the real decision made by humans, and the results showed that the new models can bring significantly better outcomes than previous models.

© Vinh The Nguyen, 2018. All rights reserved.