Planning from Pixels in Environments with Combinatorially Hard Search Spaces

TL;DR

This paper introduces a novel approach that transforms raw visual input into a latent graph to enable efficient planning in environments with complex combinatorial structures, demonstrating strong generalization and robustness.

Contribution

It presents a method for environment representation as a latent graph with state reidentification, reducing planning complexity and introducing challenging environments for testing.

Findings

Achieves linear complexity in planning from visual input

Demonstrates strong generalization across environment variations

Performs well in one-shot planning and offline RL settings

Abstract

The ability to form complex plans based on raw visual input is a litmus test for current capabilities of artificial intelligence, as it requires a seamless combination of visual processing and abstract algorithmic execution, two traditionally separate areas of computer science. A recent surge of interest in this field brought advances that yield good performance in tasks ranging from arcade games to continuous control; these methods however do not come without significant issues, such as limited generalization capabilities and difficulties when dealing with combinatorially hard planning instances. Our contribution is two-fold: (i) we present a method that learns to represent its environment as a latent graph and leverages state reidentification to reduce the complexity of finding a good policy from exponential to linear (ii) we introduce a set of lightweight environments with an underlying discrete combinatorial structure in which planning is challenging even for humans. Moreover, we show that our methods achieves strong empirical generalization to variations in the environment, even across highly disadvantaged regimes, such as "one-shot" planning, or in an offline RL paradigm which only provides low-quality trajectories.