Cyber-roaches
赛博蟑螂
Robotised insects may search collapsed buildings for survivors
机械化昆虫可能成为废墟搜救奇兵
They can detect movement, body warmth and exhaled carbon dioxide
它们可以探测人体活动、体温和呼出的二氧化碳
One such is Sato Hirotaka of Nanyang Technological University, in Singapore. He has been working on cyber-insects (including flying versions, in the form of giant flower beetles) for 15 years. Now, he has added another twist to cyber-roaches. Instead of having their movements dictated by remote control, his are autonomous agents. They are run by algorithms that respond directly to sensors in their backpacks.
新加坡南洋理工大学的佐藤裕崇就是其中一位研究者。15年来，他一直在研究赛博昆虫（包括会飞的巨型花金龟)。现在，他又给赛博蟑螂增添了新花样。它们的行动不再受远程控制，而是成为自主行动的主体。算法根据仪器背包里传感器的信号直接驱使它们行动。
The insects thus fitted out by Dr Sato are Madagascar hissing cockroaches, which are about 6cm long. The backpacks contain a communications chip, a carbon-dioxide sensor, a motion sensor, an infrared camera and a tiny battery.
佐藤使用的是马达加斯加发声蟑螂，体长约六厘米。它们扛上的仪器背包里装有通信芯片、二氧化碳传感器、运动传感器、红外摄像机和微型电池。
- infrared （使用）红外线的
For search-and-rescue operations in collapsed buildings, fleets of these roaches would be released into the rubble, to crawl their way through while searching for signs of life such as movement, body heat and elevated CO2 levels from respiration. The artificial intelligence that decides whether a set of signals actually indicate the presence of a human being is programmed directly into the camera. If it thinks it has spotted someone, it alerts a rescuer.
在倒塌的建筑物里开展搜救时，这些蟑螂将被成批释放到废墟中，四处爬行寻找生命迹象，如人体移动、体温，以及由呼吸引起的二氧化碳浓度升高。摄像机中直接编入了人工智能程序，可以判断一组信号是否真的显示有人存在。如果程序认为自己发现了幸存者，就会向救援人员发出警报。
- respiration 【正式用语】呼吸
To test this arrangement, Dr Sato and his team ran trials in a simulated disaster zone. They laid out concrete blocks of various shapes and sizes in an area of 25 square metres. Interspersed among these were a number of people, and also some decoys, such as a heat lamp, a microwave oven and a laptop. They then released the cyber-roaches, having first programmed into them the search’s start and end points. The software proved able to recognise humans correctly 87% of the time, a success rate Dr Sato thinks could be improved still further by collecting multiple images from different angles.
为了验证效果，佐藤和他的团队在一个模拟灾区进行了试验。他们在25平方米的范围内布置了各种形状和大小的混凝土块，当中散布着一些人，也放置了一些干扰物，如保温灯、微波炉和笔记本电脑。他们将搜索的起点和终点输入程序，然后释放了这些赛博蟑螂。结果证明，该软件成功识别人类的准确率达到87%。佐藤认为如果从不同角度采集多个图像，成功率还可以进一步提高。
- intersperse 散布、散置、点缀
- decoy （诱捕鸟兽的）动物、假兽、假鸟；诱饵、诱惑物、用作诱饵的人、托儿
