Recent advancements in Large Language Models (LLMs) and Visual Language Models (VLMs) have significantly impacted robotics, enabling high-level semantic motion planning applications. Reinforcement Learning (RL), a complementary paradigm, enables agents to autonomously optimize complex behaviors through interaction and reward signals. However, designing effective reward functions for RL remains challenging, especially in real-world tasks where sparse rewards are insufficient and dense rewards require elaborate design. In this work, we propose Autonomous Reinforcement learning for Complex Human-Informed Environments (ARCHIE), an unsupervised pipeline leveraging GPT-4, a pre-trained LLM, to generate reward functions directly from natural language task descriptions. The rewards are used to train RL agents in simulated environments, where we formalize the reward generation process to enhance feasibility. Additionally, GPT-4 automates the coding of task success criteria, creating a fully automated, one-shot procedure for translating human-readable text into deployable robot skills. Our approach is validated through extensive simulated experiments on single-arm and bi-manual manipulation tasks using an ABB YuMi collaborative robot, highlighting its practicality and effectiveness. Tasks are demonstrated on the real robot setup.
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.
29 August 2025 in IEEE Transactions on Automation Science and Engineering
Deep Reinforcement Learning (DRL) has emerged as a powerful paradigm for learning complex policies directly from high-dimensional input spaces, enabling advances across a variety of domains. Modern DRL algorithms often rely on dual-network Q-learning architectures to approximate optimal policies to overcome overestimation bias. Recent research has introduced approaches leveraging multiple Q-functions to further mitigate overestimation effects and enhance policy reliability. However, there is a growing emphasis on deploying DRL in edge scenarios, where privacy concerns and stringent hardware constraints necessitate highly efficient algorithms. In such environments, the computational and memory efficiency of learning methods is of critical importance. In this context, we propose Edge Delayed Deep Deterministic Policy Gradient (EdgeD3), a novel reinforcement learning algorithm specifically designed for edge computing settings. EdgeD3 offers significant reductions in GPU time (by 25%) and computational and memory usage (by 30%), while consistently achieving or surpassing the performance of state-of-the-art algorithms across multiple benchmarks and in real-world tasks.