Learning and Planning for Temporally Extended Tasks in Unknown Environments

TL;DR

This paper introduces a planning method for complex, temporally extended tasks in unknown environments, leveraging learned models to estimate action success and cost, enabling efficient planning without retraining across different scenarios.

Contribution

It presents a novel approach combining neural network-based estimates with high-level planning for temporally extended tasks in partially known environments, generalizing across environments and tasks.

Findings

Improved total cost in simulated experiments

Effective planning in real-world scenarios

Generalization across environments and tasks

Abstract

We propose a novel planning technique for satisfying tasks specified in temporal logic in partially revealed environments. We define high-level actions derived from the environment and the given task itself, and estimate how each action contributes to progress towards completing the task. As the map is revealed, we estimate the cost and probability of success of each action from images and an encoding of that action using a trained neural network. These estimates guide search for the minimum-expected-cost plan within our model. Our learned model is structured to generalize across environments and task specifications without requiring retraining. We demonstrate an improvement in total cost in both simulated and real-world experiments compared to a heuristic-driven baseline.