AGV Navigation Using RGB-D Sensor Fusion Without Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated navigation technologies for automated guided vehicles (AGVs) face challenges in convergence of learning results due to massive image data, high background noise, and high image similarity, making existing machine learning technologies not applicable for AGV navigation, and require complex and costly integration of positioning systems.
Innovation Solution
An AGV navigation device equipped with a RGB-D camera, IMU, rotary encoder, and processor that captures depth and color image data, generates training data, and inputs it into a machine learning model for deep learning, allowing the AGV to navigate without relying on positioning technologies, using a dimension-reduction method to enhance learning efficiency and stability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple positioning technologies are integrated to satisfy different environment requirements and accuracy levels, then navigation reliability is improved, but device complexity and installation cost increase significantly
Solution Approach 1:
The patent combines multiple sensing capabilities (depth sensing, color imaging, inertial measurement, and rotary encoding) into a single integrated navigation system. The RGB-D camera integrates both depth and color sensors, while the processor fuses data from all sensors to achieve reliable navigation without requiring multiple separate positioning systems.
Solution Approach 2:
The navigation device is designed with multi-functional sensors that can operate across different environments. The RGB-D camera and sensor suite serve multiple purposes: depth perception for obstacle detection, color recognition for path identification, inertial measurement for motion tracking, and rotary encoding for position measurement, eliminating the need for environment-specific positioning systems.
2Adaptability or versatility
If machine learning technologies are applied to process massive image data for AGV navigation, then navigation adaptability is improved, but learning convergence fails due to high background noise and image similarity
Solution Approach 1:
The patent segments the navigation task into distinct processing stages: depth image processing, color image processing, sensor data fusion, and path recognition. The depth image data is processed separately to extract spatial information, while color images are processed for visual cues, and these are combined with sensor data to achieve reliable learning convergence.
Solution Approach 2:
The patent introduces depth information as an additional dimension to traditional 2D image processing. By processing depth images alongside color images, the system creates a 3D spatial understanding that distinguishes between similar-looking paths and reduces the impact of background noise, enabling successful learning convergence.
3Measurement precision
If traditional positioning systems are installed to achieve precise navigation, then navigation precision is improved, but installation cost and system complexity increase
Solution Approach 1:
The navigation system uses the AGV's own sensors and onboard processors to achieve precise navigation without external positioning infrastructure. The system self-calibrates using inertial measurement unit data and rotary encoder feedback, eliminating the need for expensive external positioning systems while maintaining high precision.
Solution Approach 2:
The patent replaces mechanical positioning systems (such as track guides or physical markers) with optical and sensor-based detection. The RGB-D camera and sensor suite detect path features and calculate position through image processing and sensor fusion, substituting complex mechanical positioning infrastructure with lighter, more cost-effective sensing technology.
Data Source
AI summary
An AGV navigation device is provided, which includes a RGB-D camera, a plurality of sensors and a processor. When an AGV moves along a target route having a plurality of paths, the RGB-D camera captures the depth and color image data of each path. The sensors (including an IMU and a rotary encoder) record the acceleration, the moving speed, the direction, the rotation angle and the moving distance of the AGV moving along each path. The processor generates training data according to the depth image data, the color image data, the accelerations, the moving speeds, the directions, the moving distances and the rotation angles, and inputs the training data into a machine learning model for deep learning in order to generate a training result. Therefore, the AGV navigation device can realize automatic navigation for AGVs without any positioning technology, so can reduce the cost of automatic navigation technologies.