Built Environment Research Demonstrator

Objectives

Autonomous agents operating in built environments pose a range of research challenges in adapting to dynamic environments, in interacting with humans, and in understanding complex situations. Applications of such systems include aged care, security, entertainment, and office automation.

The built environments demonstrator will evaluate research in localisation and mapping, navigation, perception, learning and human-machine interaction. The demonstrator will be based around an aged-care application with facilities at UTS and UNSW, the latter consisting of a floor devoted to indoor robotics research and containing a Human-Computer Interaction laboratory for observing and testing interfaces to different kinds of devices.

Outcomes

The aim of the demonstrations is to show long-term and reliable deployment of autonomous agents in a real-world, dynamic environment. The demonstrations will build and maintain maps of the local environment, show continuous and accurate localisation of platforms, sense and recognise objects, humans and other agents, demonstrate meaningful interaction with humans and other autonomous systems, and learn how to interact with a changing environment.

The nursing home environment will consist of three main types of platform: mobile robots (with a variety of sensors, speech and visual interface), a semi-autonomous wheel chair (with a variety of sensors and input interfaces), and a robotic walker. These will be augmented with additional autonomous robots and intelligent building components. The key milestones for the built environment demonstrator are:

• Year 1: Definition of experimental scenarios in collaboration with industry partners. Analysis of problem domain, including evaluation of perception and navigation tasks and requirements for interaction with residents.
• Year 2: Demonstration of simultaneous localisation and mapping in a cluttered and dynamic environment. Single robot visual recognition of objects and integration with laser and sonar maps. Demonstration of natural language interaction with a single autonomous system.
• Year 3: Extension of demonstrations to multiple cooperating systems. Shared map making. Multi-agent learning and learning of patterns of behaviours of human residents. Natural language interaction with multiple devices.
• Year 4: Field-testing in real environments. Evaluation of research in planning, learning and adaptation with special emphasis on demonstrating robustness of methods and algorithms.
• Year 5: Demonstration of multiple autonomous systems (mobile robots, wheel chairs, walkers and fixed devices) in nursing human assisting residents in moving about, responding to requests and alarms, assisting carers in delivering services.

The built environment research demonstrator benefits from the partners leading research in mapping, machine learning, and control. The innovation in the research demonstrator is the focus on development and real-world demonstration of robust real-time methods of perception and learning.