和跟驰理论相比元胞自动机模型的优势在哪里？

tiany

感觉两种模型的思想差不多啊，都是由前后车的速度和距离决定后车下一时刻的速度，而元胞自动机模型考虑的又不如跟车模型全面，也不灵活，那么它的优势在什么地方呢？

阿辛

Cellular Automata (CA) model is both space and time discritized,  and the updating rule is very simple requring only interger operations. Because of this, it is computationally VERY efficient for large scale transportation network simulation and analysis. For eaxmple, the TRANSIMS employs CA models.
CA models are widely used by physicsts. However, most of CA models don';t explicitly model drivers behaviors and are not applicable of behavior-intensive  
modeling and evaluation.
Car following models, particularly, the Pipes-Gazis-Herman-Rothery family models,  are the so-called stimulus-response car following models, which assume the followers response ( accelration ) is dependent on the relative speed, distance headway and momentary speed. But there are other car-following models that are not based on this logic, for example, the Gipps behavioral model, which assumes the following driver always keep a safety distance from preceding vehicel; or the so called Aciton Point Model, which is solely derivead from Psycho-physical princples. By the way, Gipps model is used in AIMSUN simulator, Action Point Model is the underlying model that Paramics and VISSIM use. Cellular Automata model is what TRANSIMS use.

itslu

谢谢阿辛的帖子！那么CA模型的优势就是计算效率高吗？在描述交通状况方面有没有优势啊？现在有人提用多智能体multi-agent来模拟，multi-agent的优势在哪呢，作为一个新概念的提出，它的特有含义是什么啊？多谢回答！

阿辛

[这个贴子最后由阿辛在 2005/07/17 10:13am 第 1 次编辑]

我只是略知一二。另外上面说的Cellular Automata Models 主要是指 Nagel-S model. Cellular Automata Model 主要是强调　与宏观数据的吻合，比较适合大型路网的规划仿真。
CA model 的更新规则　是对　drivers behavior 的 minimal description。现在　交通仿真似乎出现一个新的方向，i.e.,更加注重对　drivers behaviors 的描述，通过在微观上model driver behaviors 来研究宏观交通流。所以有人　把这种 behavioral intensive 的model 叫做 nano-scopic model, 以区别于已有的 macroscopic, mesoscopic, and microscopic.
我个人的看法，　exisintg CA 和agent based models 比较适合非常大的网络的　planning, 比如TRANSIMS 的应用。但是对于　日新越异的　ITS technologies 的evaluation ( 比如　traffic calming stragety, Accident Avoidance System), 依赖于对　driver behaviors 的　深入理解，　所以，　除非CA 的agent-based models incorporate realistic, complex behaviroal rules, 它们目前并不适合这类 ITS 技术的evaluation and simulation.
不过，　就我所知，　我已经见到　有　研究者　使用 Cellular Automata Model 来　研究　traffic accidents 的发生。　不过，除非使用　更复杂　和　更　实际的behavioral rules, models 在ITS　其它领域的应用，恐怕最多只是漂亮的exercise而已啵。

阿辛

By the way , the fact that CA models can successfully replicate complex traffic patterns (e.g., spontaneous clustering, scychronized flow, large moving jams, capacity drop, traffic hysteries), should not be be the solely motivation promoting such models. At least I don';t agree that such models realistically or correctly describes the microscopic mechnism that generates the macroscopic flow dynamics (I don';t drive the way CA model describes, at least). Personally I don';t favor CA models, even though may physics-oriened researchers are really doing beautifual academic exercises out of such models.However, I don';t think such models really advance the state-of-the-art of the understanding of traffic dynamics.

阿辛

[这个贴子最后由阿辛在 2005/07/17 10:15am 第 2 次编辑]

对了，实际上我没有回答你的问题呢。　其实我对multi-agent simulation 没什么理解。就不敢乱放了。

阿辛

Just my 2-cents

itslu

多谢

lglabc

元胞自动机模型，我个人感觉更像一种数据结构
其他的各种跟驰模型和换道模型完全可以应用在其上面的！