Efficiently Manipulating Clutter via Learning and Search-Based Reasoning

TL;DR

This thesis introduces advanced algorithms combining deep learning, search-based reasoning, and parallel computation to improve robotic object rearrangement in cluttered environments, achieving high accuracy and efficiency.

Contribution

It presents novel algorithms integrating deep interaction prediction, Monte Carlo Tree Search, and parallelized planning for improved robotic manipulation in clutter.

Findings

Over 90% accuracy in push motion prediction

100% success rate in specific object retrieval scenarios

Significant speed-up in planning with maintained or improved solution quality

Abstract

This thesis presents novel algorithms to advance robotic object rearrangement, a critical task for autonomous systems in applications like warehouse automation and household assistance. Addressing challenges of high-dimensional planning, complex object interactions, and computational demands, our work integrates deep learning for interaction prediction, tree search for action sequencing, and parallelized computation for efficiency. Key contributions include the Deep Interaction Prediction Network (DIPN) for accurate push motion forecasting (over 90% accuracy), its synergistic integration with Monte Carlo Tree Search (MCTS) for effective non-prehensile object retrieval (100% completion in specific challenging scenarios), and the Parallel MCTS with Batched Simulations (PMBS) framework, which achieves substantial planning speed-up while maintaining or improving solution quality. The research further explores combining diverse manipulation primitives, validated extensively through simulated and real-world experiments.