2026 ASEE Annual Conference & Exposition

Reinforcement learning (RL) has emerged as a powerful tool for solving complex decision-making problems in engineering systems, from production scheduling to energy management. However, RL is often introduced to students through OpenAI Gym, the most widely used open-source RL training and testing environment, which primarily features game-based tasks such as CartPole, MountainCar, and LunarLander. While these environments are effective for teaching algorithm mechanics, they lack the context of real-world industrial systems, making it difficult for students, especially those in systems engineering (SE), to recognize the practical relevance of RL techniques.
This work presents the development of an open-source simulation environment designed to support RL education using SE context. The environment includes system models such as serial and assembly lines, and it supports the exploration of problems like resource allocation and dynamic scheduling under uncertainty. The simulation environment is designed using Python to seamlessly integrate with existing RL toolkits. The environment provides the necessary interfaces and data structures to support both student-coded RL algorithms and general-purpose libraries such as Stable-Baselines3 (SB3), enabling flexible demonstration and experimentation in instructional settings. The instructional approach encourages students to engage with key RL concepts in a realistic SE setting, such as trying new, uncertain options to find potential future benefits (exploration) versus using existing, known-to-be-good options to maximize immediate rewards (exploitation). The environment is designed to be integrated as a project-based learning component to help students better understand how RL can support decision-making in complex, uncertain systems. This paper focuses on the design and instructional integration of the environment; classroom deployment and empirical assessment are planned as future work.