OnDeFog: Online Decision Transformer under Frame Dropping

The study introduces OnDeFog, an online reinforcement learning method that enhances the Decision Transformer under Random Frame Dropping (DeFog) by integrating its mechanisms with the Online Decision Transformer (ODT). OnDeFog demonstrates improved performance in environments with high frame dropping rates and effectively handles datasets with significant low-reward data, outperforming both ODT and DeFog. This advancement is crucial for practitioners as it enables more robust decision-making in real-world applications where data loss is common.