SAPG: Split and Aggregate Policy Gradients
This addresses sample inefficiency in on-policy reinforcement learning for decision-making problems, though it appears incremental as it builds on existing methods to improve scalability.
The authors tackled the problem of policy gradient methods failing to benefit from highly parallelized environments, proposing SAPG, which splits environments into chunks and fuses them via importance sampling, achieving significantly higher performance than baselines like PPO in challenging tasks.
Despite extreme sample inefficiency, on-policy reinforcement learning, aka policy gradients, has become a fundamental tool in decision-making problems. With the recent advances in GPU-driven simulation, the ability to collect large amounts of data for RL training has scaled exponentially. However, we show that current RL methods, e.g. PPO, fail to ingest the benefit of parallelized environments beyond a certain point and their performance saturates. To address this, we propose a new on-policy RL algorithm that can effectively leverage large-scale environments by splitting them into chunks and fusing them back together via importance sampling. Our algorithm, termed SAPG, shows significantly higher performance across a variety of challenging environments where vanilla PPO and other strong baselines fail to achieve high performance. Website at https://sapg-rl.github.io/