Interpersonal coordination of movements often involves precise synchronization of action timing, particularly in expert domains such as ensemble music performance. According to the adaptation and anticipation model (ADAM) of sensorimotor synchronization, precise yet flexible interpersonal coordination is supported by reactive error correction mechanisms and anticipatory mechanisms that exploit systematic patterns in stimulus timing to plan future actions. Here, we provide a tutorial introduction to the computational architecture of ADAM and present a series of single- and dual-virtual agent simulations that examine the model parameters that produce ideal synchronization performance in different tempo conditions. In the single-agent simulations, a virtual agent synchronized responses to steady tempo sequence or a sequence containing gradual tempo changes. Parameters controlling basic reactive error (phase) correction were sufficient for producing ideal synchronization performance at the steady tempo, whereas parameters controlling anticipatory mechanisms were necessary for ideal performance with a tempo-changing sequence. In the dual-agent simulations, two interacting virtual agents produced temporal sequences from either congruent or incongruent internal performance templates specifying a steady tempo or tempo changes. Ideal performance was achieved with reactive error correction alone when both agents implemented the same performance template (either steady tempo or tempo change). In contrast, anticipatory mechanisms played a key role when one agent implemented a steady tempo template and the other agent implemented a tempo change template. These findings have implications for understanding the interplay between reactive and anticipatory mechanisms when agents possess compatible versus incompatible representations of task goals during human-human and human-machine interaction.

中文翻译：

---

ADAM的教程和仿真：感觉运动同步的适应和预期模型。

人与人之间的动作协调通常涉及动作时间的精确同步，特别是在诸如合奏音乐表演之类的专家领域中。根据感觉运动同步的适应和预期模型（ADAM），精确而灵活的人际协调得到反应性错误纠正机制和预期机制的支持，后者利用刺激性时机中的系统模式来计划未来的行动。在这里，我们为ADAM的计算体系结构提供了一个教程介绍，并提出了一系列单虚拟和双虚拟代理仿真，这些仿真检查了在不同速度条件下产生理想同步性能的模型参数。在单代理程序模拟中，虚拟代理将对稳定速度序列或包含逐渐速度变化的序列的响应同步。控制基本电抗误差（相位）校正的参数足以在稳定速度下产生理想的同步性能，而控制预期机制的参数对于具有速度变化序列的理想性能是必需的。在双主体仿真中，两个交互的虚拟主体从一致或不一致的内部性能模板生成了时间序列，这些模板指定了稳定的速度或速度变化。当两个代理实施相同的性能模板（稳定速度或速度变化）时，仅通过反应式错误校正即可实现理想性能。相反，当一个代理实施稳定的速度模板而另一种代理实施速度更改模板时，预期机制起着关键作用。这些发现对于理解代理在人与人和人机交互过程中具有任务目标的兼容或不兼容表示时理解反应性机制和预期机制之间的相互作用具有影响。