A major research project into autonomous systems and machine learning that promises to “re-invent our relationship with computers” has shown off a number of applications at an event in London.
The £10 million Orchid project, which is led by the University of Southampton and includes BAE Systems as an industrial partner, was launched in January 2011 and concludes at the end of this year.
Although autonomy is often seen in extremes, Professor Nick Jennings, head of the electronics and computer science department at the University of Southampton and project lead, said that Orchid had successfully established ways in which humans and machines can act in partnership at a “cognitive level” to complete tasks.
Orchid has developed several “fundamental algorithms” which are freely available as software in the public domain, and can be used across a broad range of sectors. They enable features such as “flexible autonomy”, which allows engineers to give computer-controlled “agents”, such as UAVs (Unmanned Aerial Vehicles) or robots, varying degrees of autonomy depending on the situation.
Jennings said: “Neither the human or the machine are always in control. The decision-making algorithm that lets you do flexible autonomy is challenging because you are in a complex and changing environment, collaborating between a complicated mix of agents and agents or humans and agents.”
An Orchid application has been developed with international charity Disaster Response to improve the handling of resources and emergency services after a disaster, such as an earthquake or flooding. The flexible-autonomy algorithm enables one operator to control up to six UAVs (Unmanned Aerial Vehicles). Meanwhile other Orchid algorithms are used to establish the accuracy of data, prioritise and plan, enabling faster and more effective decision-making.
Another application combines Orchid algorithms with a low-cost temperature sensor to provide homeowners bespoke guidance on how to save money on their energy bills. Joulo was spun out from the University of Southampton 2013 and was purchased by Dutch “smart thermostat” company Quby earlier this year.
Jennings said: “There have been eight patents generated so far and the public domain software is available for anyone to use. We're looking for commercial exploitation for the research. The algorithms are generic, you could take them out and apply them almost anywhere.”
Orchid also has applications for command and control in the defence sector. Dave Nicholson, knowledge transfer partnership officer for the Orchid project, said: “Human analysts have become burdened with the complexity and amount of data - they need automatic analysis. But the human being brings years of experience at looking at data. Orchid is about finding the right balance between the human and the machine and creating a partnership.
“In the military domain there is a mix of assets and capabilities. Orchid's agile teaming algorithm uses machine-based learning to create teams that get jobs done as quickly and efficiently as possible, but it doesn't take the human out of the loop. Strategic and tactical decisions on a high level are best taken by humans, but lower level decision-making can be autonomous.”