Linguistics could help future driverless cars cooperate better
Fri 26 Feb 2016
A team of swarm robotics researchers have applied a linguistics technique typically used in manufacturing to automatically program and control a 600-strong robot fleet.
The scientists, based at Sheffield University’s Automatic Control and Systems Engineering department, trialled the method to help the bots complete a specified set of tasks cooperatively. They found that human error was significantly reduced, making the solution safer and more reliable than previous ‘trial and error’ approaches, which often result in undesirable behaviours.
For this reason, the researchers suggested that the technology could be beneficial in areas where predictability and safety are paramount, such as in the case of driverless vehicles.
The automated programming method was applied to 600 of the university’s 900 robots – one of the largest collections worldwide. Detailed in the latest issue of Swarm Intelligence, the new theory aims to reduce the amount of human hours spent inputting data and maintaining source code – and therefore, the margin of error.
The tasks in the experiments were defined by a graphical tool, which a machine automatically translated to the bots. The supervisory technique uses a linguistics system through which the robots construct their own ‘words’, related to what they can ‘see’ and which moves they choose to action next. Robots will only perform actions from valid ‘words’, which means they are guaranteed to carry out the required tasks.
The tests involved the 600 robots making independent decisions to achieve the desired tasks, such as grouping together, moving objects, and arranging themselves into different patterns [see right].
Research lead Roderich Gross explained: “Our research poses an interesting question about how to engineer technologies we can trust – are machines more reliable programmers than humans after all? We, as humans set the boundaries of what the robots can do so we can control their behaviour, but the programming can be done by the machine, which reduces human error.”
The team underlined the importance of reducing human error in programming, quoting the estimated $312bn spent on debugging software every year.
Future research is expected to explore how humans could cooperate with robot swarms to facilitate a two-way communication and learning channel.