Improving Machine Learning-Based Robot Self-Collision Checking with Input Positional Encoding

2025-09-09

AI Summary

To address the challenges of modeling high-frequency boundary variations and insufficient accuracy under low-dimensional feature representations in robot self-collision detection, this paper proposes a lightweight multi-layer perceptron (MLP) method integrated with learnable positional encoding. Unlike conventional geometric approaches—such as triangle-mesh intersection or bounding volume hierarchy (BVH)—this work introduces, for the first time, trainable positional encoding into low-dimensional joint-angle or pose inputs to explicitly capture spatial ordering and local high-frequency boundary variations, thereby enhancing discriminative capability for complex collision boundaries. The method achieves millisecond-level inference latency while significantly improving classification accuracy across multiple robot configuration datasets (average +3.2%). It establishes a novel, purely learning-based collision detection framework that balances high precision with low computational overhead, enabling real-time motion planning.

Technology Category

• Machine Learning: Learning with Manifolds
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This manuscript investigates the integration of positional encoding -- a technique widely used in computer graphics -- into the input vector of a binary classification model for self-collision detection. The results demonstrate the benefits of incorporating positional encoding, which enhances classification accuracy by enabling the model to better capture high-frequency variations, leading to a more detailed and precise representation of complex collision patterns. The manuscript shows that machine learning-based techniques, such as lightweight multilayer perceptrons (MLPs) operating in a low-dimensional feature space, offer a faster alternative for collision checking than traditional methods that rely on geometric approaches, such as triangle-to-triangle intersection tests and Bounding Volume Hierarchies (BVH) for mesh-based models.

Problem

• Enhancing robot self-collision detection accuracy
• Improving classification of complex collision patterns
• Accelerating collision checking with machine learning

Innovation

• Input positional encoding enhances self-collision detection
• Lightweight MLPs in low-dimensional space accelerate checking
• Captures high-frequency variations for precise collision patterns