Solving periodic timetabling problems with SAT and machine learning

In this research work we address periodic timetabling, namely the optimisation of public transport timetables with respect to travel time using Boolean Satisfiability Problem (SAT) and reinforcement learning approaches. Some works already done in the field of railway timetabling propose solutions to the optimisation problem using Mixed Integer Linear Programming (MILP), genetic algorithms, SAT, the modulo network simplex, among other techniques. In this work, we propose a novel approach based on reinforcement learning and multiagents which also uses a SAT solver to get optimised solutions with respect to the travel time, a combination of techniques which (to our knowledge) has never been applied in this field. Finally, we present promising results which show that our approach applied to real world data performs better than existing SAT approaches and even outperforms the current state-of-the-art algorithms (based on the modulo network simplex, mixed integer programming and heuristics) on some problems.