Semantic Segmentation for Low-Obstacle Cleaning Robot Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous cleaning devices face challenges in detecting low-height obstacles effectively, as depth sensor data is noisy and unreliable, making it difficult to avoid debris accumulation and other low-height obstacles during cleaning operations.
Innovation Solution
Implementing a system and method using machine learning and visual data to teach a deep neural network to distinguish between ground and non-ground spaces, enabling the detection of low obstacles like cardboard boxes and forklift tines, and integrating this information into the cleaning device's path planning to enhance obstacle avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If depth sensors are used to detect obstacles, then the cleaning device can detect obstacles and height, but the detection of low-height obstacles becomes unreliable and noisy
Solution Approach 1:
The patent segments the obstacle detection task into two parts: using depth sensors for general obstacle and height detection, and using RGB camera with semantic segmentation for low-height obstacle detection. This division allows each sensor system to optimize for its specific function, resolving the contradiction between general detection capability and low-height detection precision.
Solution Approach 2:
The patent introduces semantic segmentation as an intermediary processing layer that bridges the gap between raw RGB camera data and obstacle detection. By training a deep neural network to segment ground from non-ground spaces, the system creates an intermediate representation that enhances low-height obstacle detection reliability without compromising overall detection capability.
2Measurement precision
If only depth sensor data is used for obstacle detection, then the system remains simple, but low-height obstacles cannot be reliably detected
Solution Approach 1:
The patent merges depth sensor data with RGB camera data in the semantic segmentation pipeline. By combining these data sources and processing them through a trained deep neural network, the system achieves superior low-height obstacle detection precision while managing complexity through integrated processing rather than separate systems.
Solution Approach 2:
The patent replaces traditional mechanical or simple sensor-based obstacle detection with an intelligent vision system using deep neural networks. This substitution enables the system to detect low-height obstacles that would be impossible for simple sensors to detect, achieving high precision through algorithmic intelligence rather than hardware complexity.
3Measurement precision
If semantic segmentation is implemented using machine learning and visual data, then low-height obstacle detection accuracy improves, but computational requirements and processing time increase
Solution Approach 1:
The patent performs preliminary action by training the deep neural network offline before deployment. The semantic segmentation model is pre-trained on labeled data to distinguish ground from non-ground spaces, so that during actual cleaning operations, the pre-trained model can quickly process RGB images without requiring real-time training computations, thus reducing processing time while maintaining high precision.
Data Source
AI summary
A system and method of semantic segmentation for an autonomous or semi-autonomous cleaning device. By using machine learning and visual data, one can teach an algorithm (e.g., a deep neural network) to distinguish between ground and non-ground spaces. The non-ground spaces are passed to the cleaning device path planning to avoid these areas. The algorithm may be programmed to detect low obstacles (e.g., cardboard boxes, low pallets, forklift tines) and other low-height obstacles. Semantic segmentation obstacle avoidance adds another layer of safety to perception safety protocols of a cleaning device.