This paper describes an agent-oriented approach that aims to create learning situations by solving problems. The proposed system is designed as a multi-agent that organizes interfaces, coordinators, sources of information, and mobiles. The objective of this approach is to get learners to solve a problem that leads them to get engaged in several learning activities, chosen according to their level of knowledge and preferences in order to ensure adaptive learning and reduce the rate of learner abundance in an e-learning system. The search for learning activities procedure is based on evolutionary algorithms typically a genetic algorithm, to offer learners the optimal solution adapted to their profiles and ensure a resolution of the proposed learning problem. In terms of results, we have adopted “immigration strategies” to improve the performance of the genetic algorithm. To show the effectiveness of the proposed approach we have made a comparative study with other artificial intelligence optimization methods. We conducted a real experiment with primary school learners in order to test the effectiveness of the proposed approach and to set up its functioning. The experiment results showed a high rate of success and engagement among the learners who followed the proposed adaptive learning scenario.