Goal-Aware Prediction: Learning to Model What Matters
This addresses a fundamental challenge in using learned dynamics models for vision-based control tasks in diverse real-world environments, offering an incremental improvement by aligning model objectives with downstream tasks.
The paper tackles the mismatch between learned dynamics models' objective of future state reconstruction and downstream tasks' goal of task completion, proposing a goal-aware prediction method that directs modeling towards task-relevant information in a self-supervised manner. It results in more effective modeling of relevant scene parts conditioned on goals, outperforming standard task-agnostic dynamics models and model-free reinforcement learning.
Learned dynamics models combined with both planning and policy learning algorithms have shown promise in enabling artificial agents to learn to perform many diverse tasks with limited supervision. However, one of the fundamental challenges in using a learned forward dynamics model is the mismatch between the objective of the learned model (future state reconstruction), and that of the downstream planner or policy (completing a specified task). This issue is exacerbated by vision-based control tasks in diverse real-world environments, where the complexity of the real world dwarfs model capacity. In this paper, we propose to direct prediction towards task relevant information, enabling the model to be aware of the current task and encouraging it to only model relevant quantities of the state space, resulting in a learning objective that more closely matches the downstream task. Further, we do so in an entirely self-supervised manner, without the need for a reward function or image labels. We find that our method more effectively models the relevant parts of the scene conditioned on the goal, and as a result outperforms standard task-agnostic dynamics models and model-free reinforcement learning.