Self-Paced Curriculum Reinforcement Learning for Autonomous Superbike Racing in Simulation
Researchers have developed a new framework for training autonomous agents to race superbikes in a simulated environment. This approach combines Soft Actor-Critic (SAC) with Self-Paced curriculum Deep Reinforcement Learning (SPDL), which automatically creates increasingly difficult training tasks. The system aims to address the unique challenges of motorbike control, such as balance and lean angle management, which are more complex than those in four-wheeled vehicles. Initial results indicate that SPDL is more efficient and leads to better performance in terms of lap times and stability compared to standard SAC. AI
IMPACT Introduces a novel RL approach for complex robotic control, potentially advancing autonomous systems in challenging dynamic environments.