Memory Augmented Control Networks

TL;DR

Memory Augmented Control Networks (MACN) enhance planning in partially observable environments by integrating feature extraction, neural planning, and memory storage, enabling better generalization and reasoning in complex grid world tasks.

Contribution

Introduces MACN, a novel architecture combining convolutional features, neural planning, and memory control to improve reasoning and planning in partially observable environments.

Findings

MACN successfully learns to plan in grid world environments.

The network generalizes to new, unseen environments.

MACN outperforms baseline models in complex obstacle scenarios.

Abstract

Planning problems in partially observable environments cannot be solved directly with convolutional networks and require some form of memory. But, even memory networks with sophisticated addressing schemes are unable to learn intelligent reasoning satisfactorily due to the complexity of simultaneously learning to access memory and plan. To mitigate these challenges we introduce the Memory Augmented Control Network (MACN). The proposed network architecture consists of three main parts. The first part uses convolutions to extract features and the second part uses a neural network-based planning module to pre-plan in the environment. The third part uses a network controller that learns to store those specific instances of past information that are necessary for planning. The performance of the network is evaluated in discrete grid world environments for path planning in the presence of simple and complex obstacles. We show that our network learns to plan and can generalize to new environments.