Although deep reinforcement learning (DRL) has recently emerged as a promising technique for optimal trade execution, two problems still remain unsolved: (1) the lack of a generalized model for a large collection of stocks and execution time horizons; and (2) the inability to accurately train algorithms due to the discrepancy between the simulation environment and real market. In this article, we address the two issues by utilizing a widely used reinforcement learning (RL) algorithm called proximal policy optimization (PPO) with a long short-term memory (LSTM) network and by building our proprietary order execution simulation environment based on historical level 3 market data of the Korea Stock Exchange (KRX). This paper, to the best of our knowledge, is the first to achieve generalization across 50 stocks and across an execution time horizon ranging from 165 to 380 min along with dynamic target volume. The experimental results demonstrate that the proposed algorithm outperforms the popular benchmark, the volume-weighted average price (VWAP), highlighting the potential use of DRL for optimal trade execution in real-world financial markets. Furthermore, our algorithm is the first commercialized DRL-based optimal trade execution algorithm in the South Korea stock market.
15 December 2025 in IEEE Robotics & Automation Magazine
In robotics, many different approaches ranging from classical planning over optimal control to reinforcement learning (RL) are developed and borrowed from other fields to achieve reliable control in diverse tasks. To get a clear understanding of their individual strengths and weaknesses and their applicability in real-world robotic scenarios, it is important to benchmark and compare their performances not only in a simulation but also on real hardware. The second AI Olympics with RealAIGym competition was held at the 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024) to contribute to this cause and evaluate different controllers according to their ability to solve a dynamic control problem on an underactuated double-pendulum system (Figure 1) with chaotic dynamics. This article describes the four different RL methods submitted by the participating teams; presents their performance in the swing-up task on a real double pendulum, measured against various criteria; and discusses their transferability from simulation to real (sim-to-real) hardware and their robustness to external disturbances.