Machine learning has become a key driver of efficiency and innovation in the field of financial mathematics and transport economics. The implementation of algorithms offers a deep understanding of the microstructure of the transport market. The subject of the research is the optimal strategy for placing exchange orders, taking into account the market influence in their execution. The goal is to derive a solution based on reinforcement learning (RL) algorithms that takes into account a more complex structure of market influence, in particular, through the use of a propagator model. This paper proposes a model of market influence on the execution of limit and market orders, including a propagator and a quadratic penalty based on the placed volume. These elements are used in the reward function and form the learning environment for the RL agent. The resulting solution concerns the optimal execution of exchange orders. The results show that RL algorithms can approximate existing analytical solutions of optimal trading strategies where the parameters of the pricing model are known, demonstrating the applicability of the approach to solving such problems.