Optimization of Energy Consumption in Delay-Tolerant Networks

Abstract

Delay tolerant network is a network architecture and protocol suite specifically designed to handle challenging communications environments, such as deep space communications, disaster response, and remote area communications. Although DTN [1]can provide efficient and reliable data transmission in environments with high latency, unstable connections, and high bit error rates, its energy consumption optimization problem is still a challenge, especially in scenarios with limited resources.To solve this problem, this study combines the Epidemic[2] and MaxProp[3] routing protocols with Machine Learning Models to optimize the energy consumption of DTNs. Hundreds of simulations were conducted in the ONE simulator, and an external real-world dataset from San Francisco taxi mobility traces [54] was imported. Random Forest[4] and Gradient Boosting Machine (GBM)[5] models were employed for data analysis. Through optimization involving Hyperparameter Tuning and Feature Selection, the Random Forest model achieved an R-squared value of 0.53, while the GBM model achieved an R-squared value of 0.65.