The Internet of Things (IoT) paradigm envisions billions of interconnected things. This high density of things generates a huge amount of data, which results in communication delays and increases the amount of contention. Thus, it is necessary to have efficient Medium Access Control (MAC) protocols to coordinate channel access. In this paper, we consider a network throughput maximization problem in which a set of selfish nodes compete for transmission opportunities in the IoT scenario. To enhance cooperation among the nodes and to reduce the collision rate, we formulate a memory-one channel access game in which the nodes maximize their payoffs by optimizing their channel access probabilities, conditioned on their previous transmission state. A distributed learning algorithm is used by the nodes to solve the problem efficiently in order to overcome any coordination overhead. We investigate the impact of the network topology on the solution to the problem. Our simulation results show that the throughput achieved by the memory-one game outperforms the throughput achieved by other methods including IEEE 802.11 DCF protocol.

物联网（IoT）范例设想了数十亿个互连的事物。如此高的密度会产生大量数据，这会导致通信延迟并增加争用量。因此，必须具有有效的媒体访问控制（MAC）协议来协调信道访问。在本文中，我们考虑了一个网络吞吐量最大化问题，其中一组自私节点在物联网场景中竞争传输机会。为了增强节点之间的协作并降低冲突率，我们制定了一种内存单一的通道访问游戏，其中，节点根据之前的传输状态，通过优化其通道访问概率来最大化其收益。节点使用分布式学习算法来有效解决问题，以克服任何协调开销。我们调查了网络拓扑对问题解决方案的影响。我们的仿真结果表明，记忆一游戏所实现的吞吐量要优于其他方法（包括IEEE 802.11 DCF协议）所实现的吞吐量。