A comparative study of nature-inspired metaheuristic algorithms on sustainable road network planning

Abstract: Global warming is a serious threat to the existence of human life on earth. Greenhouse gas emission is the major cause of global warming. Carbon dioxide is the major greenhouse gas emitted due to human activity. Road transport accounts for 15% of CO2 emissions worldwide. There are many initiatives adopted worldwide to minimize the emission of CO2 due to road transportation. Vehicle engines are upgraded to make it environment friendly, electric vehicles are promoted, public transport is promoted, etc. Apart from these, proper road network planning could also reduce emissions. It is not practical to completely replace the road transport system due to its importance in the transport of passengers and goods. This thesis is focused on finding ideal road conditions, that produce minimum CO2 emissions. The road network parameters that are studied in this thesis are speed limit, the number of vehicle lanes, and junction design. Due to time constraints, it is not feasible to do this study on a real road network. Hence, a simple artificial road network is created using a traffic simulator named SUMO for analysis. Both traffic congestion and the higher speed of the vehicles cause higher emissions. Roads are heavily interconnected in cities and road parameters of the adjacent roads should be adjusted together with the road being studied, to have an impact on the overall traffic. As adjacent road networks are too many numbers in cities, it is not feasible to validate every possible option. There is no algorithm invented so far to analyze such problems having too many possible states. However, there are optimization algorithms that can determine approximate solutions for such problems. In this thesis, I compared the performance of five nature-inspired metaheuristic algorithms on the sustainable road network problem. The five algorithms studied in this study are Particle Swarm Optimization, Genetic Algorithm, Artificial Bee Colony, Differential Evolution, and Harmony Search. The differential evolution algorithm generated the best result and was able to reduce the emissions by 6% and it is followed by genetic algorithms. Statistical tests are performed to evaluate whether the differences are significant or not.