The curriculum sequencing (CS) problem asks to find a personalized optimal path for a learner in an e‐learning environment. Solving this problem contributes to developing adaptive e‐learning systems which provide personalized learning paths compatible with the individual profiles of the learners. The solutions to large CS instances can be only approximated as this problem is NP‐hard. In this paper, we formulate the CS problem as a constraint satisfaction problem and investigate swarm intelligence methods to solve it, including a newly introduced method called SwarmRW and the widely used ant colony system (ACS). In addition, we introduce two variants of SwarmRW, called SwarmRW_rnd and SwarmRW_inc. The performance results obtained on real data show that SwarmRW_rnd achieved the best results in terms of maximizing the number of satisfied constraints. However, SwarmRW_inc demonstrated the best trade‐off between the solution quality and the running‐time to obtain it.