Programming Behaviour of Personal Service Robots with Application to Healthcare

Abstract

The objective of this research work is to advance the process of programming personal service robots. As robots enter the open and complex world of humans, they are expected to perform tasks that require close interactions with humans and the environment. Industrial robotics researchers have challenged the process of constructing novel robotic mechanisms. However, the fluid integration of a matured personal service robot platform that can be easily used and easily re-programmed is yet to be seen. Personal service robots are examples of systems that are expected to work in an autonomous or semiautonomous way to deliver useful services for the well-being of humans. An example of such a personal service robot is the Healthbot developed at the University of Auckland. The Healthbot is a multi-purpose reprogrammable robot which can be customised to various scenarios of usage. The Healthbot assists older people with cognitive tasks, such as remembering medication schedules. This thesis provides a workflow for converting the service design knowledge into working software on a robot in a multi-disciplinary team. The service requirements have been developed for a unique class of service applications for the Healthbots project. The analysis provided in this thesis can also serve as a blueprint and thus provide insight for developing services in similar scenarios of service robot design. The thesis presents a case study of design of the Healthbot medication management service to motivate the complexities involved in designing services in multidisciplinary teams. This case study shows that the design process should be supported by programming languages and development environments. It presents the requirements for a software architecture that unifies the design process and fulfils the requirements for an application design framework. The service design was not purely technology driven. Rather, it looked at the design challenges at a system-of-systems level and integrated a closed-loop medication workflow involving multiple user roles in the healthcare system. The proposed design approach also helped domain experts in the Healthbots design team to conduct field trials to identify opportunities for effective use of personal service robots with older people. This thesis develops a formal language specification of a domain-specific programming language for representing the robot’s behaviour and presents a graphical metamodel for the robot’s behaviour definition. The behaviour is defined as what the robot does in response to an event and is the most basic element of a service. The behaviour is encoded as a finite state machine model. The robot’s multi-modal dialogue and behaviour is specified in a set of states that are organised in the domain-specific programming language. A collection of states constitutes a service. The research has also developed a visual programming environment for authoring this domain-specific programming language. This visual programming environment is named as RoboStudio. This programming environment abstracts the textual domain specific language and provides a familiar visual environment using common design elements present in most Integrated Development Environments. An evaluation through two studies provides sufficient evidence to validate and confirm the hypotheses related to the successful attainment of RoboStudio’s design goals. In the first study, using the Cognitive Dimensions Framework, RoboStudio was found to have better Closeness of mapping, causing lesser Viscosity, better Role Expressiveness, fewer Hidden Dependencies, fewer Hard mental operations, better Visibility, and being less Error prone compared to a structured Code Editor. Qualitative data analysis revealed that RoboStudio led to higher user experience compared to a Code Editor for creating, editing, visualising and diagnosing the robot behaviour in both studies with programmers and non-programmers. Feedback from the evaluation study with nonprogrammers showed that RoboStudio could also enable clinicians to create simple healthcare applications. For example, a therapist should be able to easily develop a self-guided exercise/training program. Although the thesis concepts have been developed with an application in context, the research is applicable within a broader context of service robots and would be useful for the larger robotics community. The thesis contributions pave a way for a unified software platform that supports the design process for personal service robots. While programming languages and tools are important, the robotics community will find it useful to know that the service design task can best be supported by providing specialized tools for the different levels of abstraction. Clearly understanding the requirements of those layers of abstraction, especially how the interaction design influences the holistic integration of hardware and software for practical use is a greater system-of-systems integration task. This is currently an active area of research pursued by multiple schools of thought. The progress made in this research supports the application of robotics in healthcare. In the future, there will be multiple cases of robotics technology influencing emerging health technologies. Smart technologies for aging, disability, and independence working in confluence with devices for self-care and activities of daily living are likely to be highly prevalent among the older people for improving self-efficacy and quality of life.