Reinforcement Learning for Liquidity-Weighted Optimal Trade Execution strategies

Abstract

Optimal execution of large institutional orders is a critical problem in electronic markets. Traditional algorithms such as TWAP (Time-Weighted Average Price) and VWAP (Volume-Weighted Average Price) fail to adapt to dynamic liquidity fluctuations in modern high-frequency markets. This paper proposes a novel Liquidity-Weighted Reinforcement Learning Execution (LW-RLX) framework that learns optimal order slicing policies based on real-time liquidity signals, order book imbalance, and market impact estimates.

The proposed model formulates trade execution as a Markov Decision Process (MDP) and integrates deep reinforcement learning with liquidity-aware state representations. The agent dynamically adjusts order sizes to minimize market impact, slippage, and execution risk. Simulation results using realistic limit-order-book environments demonstrate that the proposed method outperforms classical execution benchmarks such as TWAP, VWAP, and Almgren-Chriss optimal execution models.