双语阅读｜人工智能的智能节油之道

FROM to to , , the , it is clear that (AI) will be an part of the cars of the . But it is not only the of them that will . AI will also such cars to use more .

避开乱穿马路的行人，紧急制动，或许最终实现自动驾驶，毫无疑问，人工智能（AI）将会成为未来汽车的重要部件。不过，这不仅有利于驾驶本身。另外，人工智能还能使这样的车更加节能。

Cars have long had – that on the fly to in . The of power has, , . , which have both a and an motor run by a that is by as the slows or , need more than does a alone. get even with plug-in , which can be from the mains and have a -only range.

引擎管理的计算机化早已应用于汽车，用以快速对行驶过程中路面状况的变化做出反应。然而，电力的加入让事情变得复杂 。混合动力车——同时具有汽油发动机与电动发动机——在减速或刹车时通过动能转化而充电，但这需要投入较普通车更多的引擎管理。带插头接点的混合动力车虽然能通过电源充电，并拥有更长的电动行驶时间，但引擎的智能化管理更加繁琐。

This is where AI could help, Qi, Barth and their at the of , . They are a of which uses a piece of AI that can learn from past .

美国加州里弗赛德的加州大学的齐学伟（ Qi）、 Barth以及他们同事认为，这是人工智能的用武之地。他们正在研发一套能源管理系统，能利用人工智能学习以往的经验。

Their works by the trip down into small , each of which might be less than a long, as the . In each the to see if the has the same , using data from to the ’s speed, , time of day, the of the road, the ’s state of and the ’s rate of fuel . If the is , it the same – that it used for the next of the . For that it has not , the what the best power might be and adds the to its for . , the idea is that the will also learn from the of its in other cars, by for all such to share their data .

该程序的工作机制是在行驶的过程中，将整个行程拆分成独立小路段，每个路段的时长可能还不到一分钟。在每个路段中，系统通过校验以确认该车是否曾遇到过相同的行驶状态，参考数据包括交通信息，行驶速度，行驶位置，在一天中的时间，路面坡度，电池的电量状态以及发动机的燃油消耗率。如果条件相似，人工智能系统就会在剩下的行程中启动与前一次相同的行驶策略。如果是未遇到过的情况，系统会评估出最佳省能方案并将结果添加到数据库中以便日后的参考。最终，通过数据在线分享的方式，可以获得其他车辆所获得的数据信息，这样就可以学习其他车辆上的经验。

, such a would be fed its route and in , to make to . But Dr Qi and Dr Barth are , and know that is to . If a were , it would be able to pass on to the . But use only to get them to . Hence the ’ to a that does not rely on where it is going.

在理想条件下，该系统会事先预设路线与目的地，让计算变得更为容易。但齐学伟与Barth均为现实主义者，知道那并不太可能发生。如果使用卫星导航，它确实可以给该系统传送相关信息。但司机只有在去陌生地点时才会使用到导航。因此，研究人员决定研发一款不依赖于目的地位置的系统。

It seems to work—at least, in . Using live to mimic in , Dr Qi and Dr Barth their with a basic – for plug-in that to power once the is . As they in a paper to be in IEEE on , their was 10.7% more than the one. If the is aware in that a will be used as part of the trip (which might be by one via the ’s ) it can the use of power the , to , when the will be up. In such the fuel was 31.5%.

至少在模拟状态看起来可行。利用现实交通实况模拟南加州道路，刘学伟与Barth用他们的程序与普通燃油管理系统进行比较，后者用于传统带插头接电的混合动力车，当电池电量耗尽时，它能直接转换成汽油驱动的程序。正如他们在一篇即将发表于IEEE智能交通系统的论文上所述的那样关于人工智能的问题，他们的系统比传统车燃油利用率高出10.7%。如果系统提前知道路上可以充电(这能通过在车上的信息屏幕进行预定，然后提前安排)，它能够在旅途中均衡使用电能关于人工智能的问题，在知道电池在何时能够被充满的前提下，达到最大利用率。

Dr Qi and his now hope to work with a to test the on real roads. If all goes well, and their able to cope with the of in , they will not be left on the hard .

目前，齐学伟与他的同事希望能与汽车制造商合作，在实际道路上测试算法。如果一切顺利，他们的系统将被证实能够解决南加州上下班路段的难题，那么也将不会出现车辆拥堵的情况了。

编译：佚名

本文到此结束，希望对大家有所帮助。