Abstract:In recent years, machine learning technologies have gained immense popularity and are being used in a wide range of domains. However, due to the complexity associated with machine learning algorithms, it is a challenge to make it user-friendly, easy to understand and apply. Machine learning applications are especially challenging for users who do not have proficiency in this area. In this paper, we use model-driven engineering (MDE) methods and tools for developing a domain-specific modelling environment to contribute towards providing a solution for this problem. We targeted reinforcement learning from the machine learning domain, and evaluated the proposed language, reinforcement learning modelling language (RLML), with multiple applications. The tool supports syntax-directed editing, constraint checking, and automatic generation of code from RLML models. The environment also provides support for comparing results generated with multiple RL algorithms. With our proposed MDE approach, we were able to help in abstracting reinforcement learning technologies and improve the learning curve for RL users.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to simplify the application of Reinforcement Learning (RL) technology, making it more user - friendly, easier to understand and use, especially for users without expertise in this field. Specifically, the paper addresses this problem in the following ways: 1. **Develop a domain - specific modeling environment**: The paper proposes a method and tool based on Model - Driven Engineering (MDE) to develop a modeling environment specific to the reinforcement learning field. This environment aims to abstract complex RL algorithms and lower the threshold for their use. 2. **Introduce the Reinforcement Learning Modeling Language (RLML)**: The author designs a modeling language specific to reinforcement learning (Reinforcement Learning Modelling Language, RLML) to help users more conveniently describe and implement RL problems. RLML supports syntax - directed editing, constraint checking, and automatic code generation from models, thereby reducing the need for manual code writing. 3. **Support multi - algorithm comparison**: This environment also provides support for multiple RL algorithms, allowing users to compare different algorithms, analyze results, and select the most suitable algorithm. This helps users better understand the impact of different parameter settings on algorithm performance. 4. **Enhance accessibility**: In order to make RL algorithms more accessible to users without a Python programming background, the paper not only supports Python code generation but also implements Java code generation, ensuring availability for Java users. 5. **Simplify the input format**: Considering that the input formats of actions, rewards, and states in RL problems are not fixed, the paper proposes to simplify the processing of complex inputs through file import and adds a data validation function to ensure the validity of the input. Through these methods, the paper aims to provide a more intuitive and easy - to - use platform for non - professional users, enabling them to more easily apply reinforcement learning technology without having to deeply understand the complexity of the underlying algorithms. This not only lowers the threshold for using RL technology but also opens the door to applications in more fields.

Towards a Domain-Specific Modelling Environment for Reinforcement Learning

A survey on model-based reinforcement learning

DeLF: Designing Learning Environments with Foundation Models

Improving Large Language Models via Fine-grained Reinforcement Learning with Minimum Editing Constraint

Importance of Environment Design in Reinforcement Learning: A Study of a Robotic Environment

Eden: A Unified Environment Framework for Booming Reinforcement Learning Algorithms

Challenges of real-world reinforcement learning: definitions, benchmarks and analysis

LExCI: A framework for reinforcement learning with embedded systems

Model-based Reinforcement Learning for Semi-Markov Decision Processes with Neural ODEs

A Survey On Enhancing Reinforcement Learning in Complex Environments: Insights from Human and LLM Feedback

Jumanji: a Diverse Suite of Scalable Reinforcement Learning Environments in JAX

Staged Reinforcement Learning for Complex Tasks through Decomposed Environments

Hierarchical Framework for Interpretable and Probabilistic Model-Based Safe Reinforcement Learning

Environments for Lifelong Reinforcement Learning

REvolve: Reward Evolution with Large Language Models using Human Feedback

Mental Modeling of Reinforcement Learning Agents by Language Models

Using Large Language Models to Automate and Expedite Reinforcement Learning with Reward Machine

Simple Embodied Language Learning as a Byproduct of Meta-Reinforcement Learning

Large Language Models as Efficient Reward Function Searchers for Custom-Environment Multi-Objective Reinforcement Learning

Reinforcement Learning in Education: A Literature Review