车载算法有望快速解决堵车问题

译者 猪的米

In recent years, various mathematical models and experimental measurements of traffic patterns have led to a consensus about the general kinds of traffic flows that can occur. There are three types.

近年来，多种关于交通模式的数学模型和实验测量都得出了一个相同的结论：一般来说发生拥堵都经历三个过程。

First is free flow in which the density of traffic is low enough to allow vehicles to travel at the maximum speed allowed. Then there is synchronised flow when a higher traffic density forces cars to travel at similar slow speeds but without stop-start motion. Finally, there is the jam in which the speed drops to zero when the traffic density rises above some threshold.

首先是畅通无阻，交通密度足够低，使得车辆能以最高时速行驶。不久之后由于交通密度的提高，车辆被迫以相对较低但仍相同的速度行驶，但不至于走走停停。最终，某一区域的汽车速度跌至零，造成交通密度骤然上升从而引起交通拥堵。

The way the flow transitions from one regime to another is hugely complex but a number of models, in particular those using cellular automaton, have become useful in studying how it occurs.

车辆流量的变化从一个区域传导到另一个区域的方式极其复杂，但一些模型如元胞自动机算法，对研究这个过程如何发生十分有用。

One interesting question is how best to dissolve jams once they form. Most traffic experts agree that the basic idea is to ensure that cars leave the jam more quickly than they arrive, so that the jam dissolves.

一个有趣的问题是，你如何在拥堵刚形成时就解决它。大多数交通专家都认为如果能够保证车辆逃离拥堵的速度比车辆聚集的速度更快，那么拥堵就能自行消失。

Now Hyun Keun Lee and Beom Jun Kim at the University of Seoul in South Korea have a come up with a simple idea to automate and improve this dissolving process. They define two types of drivers: optimistic and defensive. Defensive drivers leave more room to the vehicle ahead than required by safety. Optimistic drivers leave too little.

现在来自韩国首尔大学的Hyun Keun Lee和Jun Kim提出了一个简单的想法，能够自动并改善缓解过程。他们把司机分为两类：乐观型和防御型。防御型司机与前车留的距离会比安全距离大很多，而乐观型司机空的太少。

They then use a cellular automaton to model traffic flow in a way that reproduces most of the usual driving behaviours such as exceeding the speed limit, overreacting to road conditions by accelerating and braking to hard and so on.

接着他们用元胞自动机算法来模拟交通车流，这种算法能够重现普通驾驶员的行为诸如超速、对道路情况过度反应，比如加速或刹车过猛等等。

But they also add an extra ingredient. All the vehicles in this model share their speed and position with their neighbours and this information filters downstream. That means downstream vehicles immediately become aware that the traffic ahead has come to a standstill.

不过他们在这种情况中附加了一个条件，那就是所有的车辆都与附近的车辆共享它的速度和位置，并且这个信息会向“下游”传播。这意味着“下游”的车辆立即就能知道前面的车子停住了。

When that happens, Lee and Kim's algorithm immediately switches all the downstream driving behaviour to defensive, so that vehicles exceed the safe distance between them. This slows the rate at which vehicles join the jam.

当这种情况发生时，Lee和Kim的算法就能立即将“下游”车辆的驾驶模式调为“防御型”，因而车辆会将与增大与前车的安全间距，这样就能减缓后续车辆聚积的速度。

At the same time, vehicles leaving the jams are made to accelerate away quickly using automated cruise control. This increases the rate at which vehicles leave the jam.

与此同时，逃离拥堵的车辆将自动提高巡航速度，这样就加快了车辆离开拥堵的速度。

The result is that the jam quickly dissolves.

这种情况下拥堵将得到快速缓解。

That's an interesting and simple approach that could be implemented relatively easily in the next generation of cars. It's greatest value is that it requires no central control, only an on-board algorithm on most cars. It also requires a little more automated on-board control than cars currently have but not an unrealistic amount.

这是一个有趣而又简单的解决方案，能够相对轻松地应用于下一代的车辆上。它最大的价值在于它无需中央控制，只需在大多数车辆上加载一套算法。这只需要在现有的车辆控制上再装一个自动算法就行了，并不是一个不切实际的幻想。

But it will require a little more modelling to ensure that this kind of group control doesn't lead to other emergent behaviour that could be detrimental. Neither is it clear what percentage of cars must be fitted with this ability for the effect to work. But it certainly looks worthy of further investigation.

然而这套系统还需要一点建模来保证这种群体控制不会导致其他我们不想看到的情况。我们不太清楚需要多少车辆装备这个系统才能让整个系统有效运作，但这值得我们去进一步研究。

It's a little premature to say the traffic jams could be banished from the roads of the future but we may bot have to spend as long sitting in them.

现在说我们从此解决了拥堵问题还为时过早，但至少我们不用花费如此多的时间坐在车里干等。