NoRML: No-Reward Meta Learning
This addresses the challenge of real-world adaptation for agents in scenarios where rewards are unavailable, though it appears incremental as it builds directly on MAML.
The paper tackles the problem of adapting reinforcement learning agents to new environments without external reward signals by introducing No-Reward Meta Learning (NoRML), which extends MAML to use observable dynamics instead of rewards, and shows that it outperforms MAML when dynamics change between tasks.
Efficiently adapting to new environments and changes in dynamics is critical for agents to successfully operate in the real world. Reinforcement learning (RL) based approaches typically rely on external reward feedback for adaptation. However, in many scenarios this reward signal might not be readily available for the target task, or the difference between the environments can be implicit and only observable from the dynamics. To this end, we introduce a method that allows for self-adaptation of learned policies: No-Reward Meta Learning (NoRML). NoRML extends Model Agnostic Meta Learning (MAML) for RL and uses observable dynamics of the environment instead of an explicit reward function in MAML's finetune step. Our method has a more expressive update step than MAML, while maintaining MAML's gradient based foundation. Additionally, in order to allow more targeted exploration, we implement an extension to MAML that effectively disconnects the meta-policy parameters from the fine-tuned policies' parameters. We first study our method on a number of synthetic control problems and then validate our method on common benchmark environments, showing that NoRML outperforms MAML when the dynamics change between tasks.