We investigate the computational power of the deterministic single-agent model where the agent and each node are equipped with a limited amount of persistent memory. Tasks are formalized as decision problems on properties of input graphs, i.e., the task is defined as a subset $\mathcal{T}$ of all possible input graphs, and the agent must decide if the network belongs to $\mathcal{T}$ or not. We focus on the class of the decision problems which are solvable in a polynomial number of movements, and polynomial-time local computation. The contribution of this paper is the computational power of the very weak system with one-bit agent memory and $O(1)$-bit storage (i.e. node memory) is equivalent to the one with $O(n)$-bit agent memory and $O(1)$-bit storage. We also show that the one-bit agent memory is crucial to lead this equivalence: There exists a decision task which can be solved by the one-bit memory agent but cannot be solved by the zero-bit memory (i.e., oblivious) agent. Our result is deduced by the algorithm of simulating the $O(n)$-bit memory agent by the one-bit memory agent with polynomial-time overhead, which is developed by two novel technical tools. The first one is a dynamic $s$-$t$ path maintenance mechanism which uses only $O(1)$-bit storage per node. The second one is a new lexicographically-ordered DFS algorithm for the mobile agent system with $O(1)$-bit memory and $O(1)$-bit storage per node. These tools are of independent interest.
翻译:我们调查确定型单一试剂模型的计算能力, 代理器和每个节点在其中安装了有限的耐久记忆量。 任务被正式化, 原因是输入图的特性决定问题, 即, 任务被定义为所有可能的输入图的子数$\ mathcal{T}$, 代理器必须决定网络是否属于$\mathcal{T} 美元, 代理器必须决定网络是否属于$\mathcal{T} 美元。 我们还显示一比特的记忆对于这一等值来说至关重要: 存在一个决定任务, 由一比特的移动数和多比特的本地计算来解决, 但无法通过一比特的存储( i. bitrde) 路径的计算能力, 即一比特的存储和 $- 美元的存储( NOdemedemeal deal) 等值。 我们的一比特的存储工具是一比特的存储工具, 由一比特的存储机为一比特的存储工具 。