基于反馈的合作强化学习水下路由算法

Underwater Routing Algorithm based on Feedback Collaborative Reinforcement Learning

针对水声传感器网络(UASNs)中存在节点能量有限、传播延时大、时变性等问题,提出了一种基于反馈的合作强化学习水下路由算法。将路由建模成一个离散Markov决策过程,使用合作强化学习算法寻找最优路由。基于反馈的水下路由算法中,节点不断从所处环境中学习反馈,获得的回报函数中考虑节点深度、剩余能量以及邻居节点等信息,在保证可靠性的情况下,达到传输路径和能量消耗的折中,增强了路由算法对网络动态变化的适应性。

For the limited energy, large delay and time variability of UASNs（underwater acoustic sensor networks）, CMAQ（collaborative reinforcement learning based on feedback underwater routing） protocol for UASNs is proposed. The routing is modeled as a discrete Markov decision process, and the cooperative reinforcement learning algorithm is used to find the optimal route. In the underwater routing algorithm based on feedback, the nodes continuously learn feedback from the environment, and consider the node depth, residual energy and neighbor nodes from the acqiured return function. In the case of reliability, the tradeoff between transmission path and energy consumption is achieved, and the adaptability of routing algorithm to the network dynamics enhanced.