Forbidding Fruit

….. industrial robots are ….. bolshiest workforce you can imagine. ….. slightest variation in their working conditions or duties, and ….. whistle will blow: they are off ….. job. That applies even to ….. task as simple (for humans) as fruit-picking. Last month ….. researchers from ….. University of Florida put ….. experimental robot picker into ….. orange groves of Catania in Sicily. It will be …. long while yet before ….. poor thing can cope.

….. two problems confront ….. fruit picker (be it hydraulic or human): recognizing ….. fruit, and harvesting it quickly and thoroughly without damaging ….. fruit or ….. tree. So far, ….. Florida team, led by Dr Roy Harrell, has solved only ….. first - and even that took some of ….. newest tricks in robotics. Their one-armed picker, which is mounted on ….. trailer, is fitted with ….. colour-television camera, ….. sonar equipment, and ….. computer to make sense of what it sees and hears.

Like ….. human eye, ….. television cameras split ….. world into ….. three primary colours: red, green and blue. It is not easy to define exactly which mixture of these colours should count as 'orange'. ….. definition must be wide enough to include most of ….. oranges on most of ….. trees, but not so wide as to encompass ….. unripe fruit or other objects in ….. vicinity. At ….. sunset, for example, when ….. light becomes redder, everything starts to appear more orange. At that stage, either harvesting has to stop or ….. definition must be recalculated.

Television only tells ….. robot in which direction to pursue its orange. To discover how far away it is, ….. machine has to rely on ….. bat-like sonar. ….. robot makes high-pitched squeaks and measures ….. time taken for its squeaks to bounce back to build up ….. orange-map. ….. map is overlaid on to ….. television image. Faced with more than one orange, ….. robot chooses ….. one nearest ….. centre of its field of view and makes ….. bee-line for it. At best, ….. Sicilian prototype picks ….. orange every three to four seconds. ….. employable machine would need to be able to sustain such ….. rate all day. It would probably have up to 12 arms, each with its own camera, sonar and computer. If such machines are to pay their way and harvest ….. profit, Dr Harrell thinks they will have to learn to pick at least 85% of ….. oranges on any given tree. At ….. moment, ….. prototype can only manage 75% of ….. fruit within its limited grasp.

The Economist (BrE)