A taste of future robotics
14 Dec 2017 - 09:01:12 in ResearchDid you know that robots can now ‘taste’ as humans do? We do now, thanks to Victoria Honours student Michael Pearson. Michael’s ENGR 489 project gave robots the ability to mimic a human’s inquisitive nature—a nature which normally enables us to recognise objects much better than artificial systems can. Robots are an increasingly prevalent part of our society, but they struggle to achieve some tasks which are trivial to humans, says Michael. “Anecdotally, humans don’t just use vision to recognise objects, so why should robots?” says Michael. “My project explored how we can add more senses to a robot to improve its ability to recognise those objects too.” In the project, which included machine learning, networking and electronics, Michael created a multi-sensory robot using a low-cost spectrometer to allow basic recognition of objects, such as a cricket ball compared to a nectarine. Mimicking a human’s sense of taste in this way could be described as using a crude approximation of an artificial ‘mouth’, says Michael. “When humans taste food we immediately get a sensation of how sweet, sour or perhaps bitter it is. The sensor used in this project is able to crudely detect molecules just like a human's tongue. From this information the sensor can then make a prediction as to what has been scanned.” This classification system has varied uses, including in the self-checkout aisle of supermarkets if a customer were to weigh and scan other items to receive a cheaper price than their actual product. Michael’s robot can tell a carrot from a cucumber, for example.
The end goal of the project was to improve the accuracy of existing classification systems, says Michael. He used a lot of the knowledge he gained studying at Victoria, especially his 400-level Artificial Intelligence papers, which gave him the understanding of the algorithms necessary for the project. However, it turned out to be a challenge. “Often there was a lot of learning required before progress could be made.”
The facilities provided by the University were also invaluable, from the software Michael used while studying, to the hardware to run his experiments. He also has some advice for future students: “Make sure you don’t forget to document all of the small decisions that seem obvious to you. Every aspect of your project is important—and the more you can communicate what you did, the happier you’ll be with your final report.” Being in Wellington also means Michael is close to several high-profile technology companies, including TradeMe and Xero, which could now feature in his future. “I’d love to work on embedded systems, with some aspect of machine learning,” he says. “This project has given me so many skills that I hope to use in my future, both personally and professionally.”