自供电路侧单元能量-时延均衡分组调度策略

考虑了自供电路侧单元在分组传输过程中能量收集、车辆到达与车速的随机性, 基于受限马尔科夫决策模型建立分组调度系统模型, 研究了分组平均传输时延与能量消耗; 分析了在能量队列约束下最小分组平均传输时延的优化问题, 提出了自供电路侧单元能量-时延均衡分组调度策略, 通过仿真试验分析了最优分组调度策略性能, 并与贪婪中继方案和Q-learning算法进行对比。仿真结果表明: 该分组调度策略具有双门限结构, 系统通过自供电路侧单元的能量队列状态以及到达车辆的车速状态确定决策变量, 使系统可以在考虑能量利用效率的前提下降低监测数据分组的平均传输时延, 保证自供电路侧单元在能量存储不溢出不耗尽的同时, 最小化系统分组平均传输时延; 在单分组发送模型中, 提出的分组调度策略的平均传输时延相比贪婪中继方案降低了15.7%, 相比Q-learning算法降低了13.5%;在批量分组发送模型中, 其分组平均传输时延相比贪婪中继方案降低了20.4%, 相比Q-learning算法降低了11.5%。 

Packet scheduling scheme for energy-delay tradeoff in self-powered roadside units

Considering the randomness of energy harvesting, vehicle arrival and vehicle speed in the process of self-powered roadside units' packet transmission, the packet scheduling system was modeled as a constrained Markov decision model to analyze the average packet transmission delay and energy consumption. The optimization problem of the minimum packet average transmission delay under the constraint of energy queue was analyzed, and packets scheduling scheme for energy-delay tradeoff in self-powered roadside unitswas proposed. The performance of the optimal packet scheduling scheme was analyzed by simulation experiments, and compared with the greedy bundle relaying scheme and Q-learning method. Simulation result shows that the packet scheduling scheme has a dual-threshold structure. The decision variables were determined by the energy queue state of the self-powered roadside units and the speed state of the arriving vehicles, so that the system can reduce the average transmission delay of the monitoring data on the premise of considering the energy utilization efficiency, and ensure non-overflow of the self-powered roadside units' energy storage to minimize the average packet transmission delay. In the single packet transmission model, the average transmission delay of the proposed packet scheduling scheme is 15.7% lower than that of greedy bundle relaying scheme, and 13.5% lower than that of Q-learning method. In the batch packet transmission model, the average transmission delay of the proposed packet scheduling scheme is 20.4% lower than that of greedy forwarding scheme, and 11.5% lower than that of Q-learning method.