Navigation and control for assistive robotics

Publication Type:
Thesis
Issue Date:
2016
Full metadata record
Files in This Item:
Filename Description Size
01front.pdf147.31 kB
Adobe PDF
02whole.pdf10.43 MB
Adobe PDF
In this thesis, we address the problem of navigation and control for assistive robots. Autonomously creating a suitable representation of the environment (a map) and having the ability to localise a robot in that environment are considered to be the cornerstones of autonomous robot navigation. We present a distance function based framework to represent the occupancy of two-dimensional environments and a chamfer distance based sensor model to relate measurements captured from a sensor mounted on the robot to the environment representation. Employing the proposed representation and the sensor model, we propose two novel strategies to localise the robot on the map using an extended Kalman filter and an optimisation based method. These methods are computationally more efficient and are free of environment dependent tuning, which are necessities for assistive robots to operate in different environments ranging from small households to large shopping centres. We also demonstrate an adaptation of the popular particle filter based localisation algorithm using the distance function representation. A mapping algorithm that utilises the proposed distance function based framework, which can be used to create maps of considerably large scale crowded indoor environments with low error accumulation is also presented. Although we do not consider the effect of sensor uncertainties, we demonstrate that the algorithm can efficiently build high-quality maps that can be used in practical scenarios of importance associated with assistive robots. We present experiments conducted using simulations, public domain datasets, and experimental datasets we collected in real environments to evaluate and compare these algorithms. The control strategy used in an assistive robot needs to be specifically designed to suit the task that the robot is expected to perform. Using standard user centred design methods often result in complicated, unintuitive control interfaces for assistive robots, which are difficult to be integrated into the daily activities of the end users. We demonstrate that design approaches based on the principles of cooperative design can be used to alleviate the complexities in the design process. We propose and develop a control system based on admittance control for a robotic hoist, and evaluate it using user studies to experimentally illustrate that this design framework could be used for developing controllers for assistive robots in general. The analysis of electromyographic measurements and forces exerted by the end users while using the robotic hoist confirm that the robot has the potential to reduce musculoskeletal injuries amongst care workers in the aged and disabled care sector, by providing assistance during the patient transfer process. As a result of the cooperative design process, the control interface became simple, intuitive, and easy to use, which made the robot readily incorporable to the work-flow of care facility.
Please use this identifier to cite or link to this item: