SLAM System Using Neural Network for Indoor Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional navigation systems, such as GPS, struggle in indoor environments and suffer from precision limitations and camera drift issues in Simultaneous Localization and Mapping (SLAM) applications, which affect the ability to accurately map and localize within unrecorded spaces.
Innovation Solution
A SLAM system utilizing an artificial neural network that combines data from RGB and depth cameras with inertial measurement units to create a map of the environment, allowing for quick re-localization by associating keyframes and camera poses, and employing transfer learning to adapt the neural network for precise localization using low-cost sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional GPS-based navigation is used, then outdoor navigation is simple and reliable, but it fails completely in indoor environments
Solution Approach 1:
The system implements a universal navigation framework that works across both outdoor and indoor environments by integrating multiple sensing modalities (visual, inertial, barometric) and multiple localization approaches (GPS when available, visual SLAM when GPS is unavailable). The navigation system adapts its functionality based on environmental context, making it versatile across different operating conditions.
Solution Approach 2:
The patent introduces visual SLAM and sensor fusion as intermediary systems that bridge the gap between GPS-based outdoor navigation and the need for indoor navigation. When GPS signals are unavailable, the system transitions to using visual features and inertial measurements as mediators to maintain navigation functionality in indoor environments.
2Measurement precision
If traditional visual SLAM is used, then absolute location can be calculated, but camera drift and precision limitations occur
Solution Approach 1:
The system merges multiple data sources and localization methods by combining visual SLAM with inertial measurement unit (IMU) data and barometric pressure sensors. This fusion of multiple sensing modalities compensates for the weaknesses of individual systems, reducing camera drift and improving overall localization precision and reliability.
Solution Approach 2:
The system implements feedback mechanisms where localization results are continuously refined by comparing visual features across frames and correcting drift through inertial measurement data. The barometric pressure data provides additional feedback for vertical position estimation, creating a closed-loop system that maintains precision over time.
3Measurement precision
If high-precision mapping is achieved, then localization accuracy improves, but computing resources and time consumption increase
Solution Approach 1:
The system segments the mapping and localization process by maintaining a sparse set of keyframes and landmark features rather than processing all visual data. This segmentation allows the system to achieve sufficient mapping precision while reducing computational burden and processing time by focusing only on critical visual elements.
Solution Approach 2:
The system employs partial action by using a reduced set of visual features and a simplified map representation that provides sufficient localization accuracy without the computational overhead of complete environmental mapping. This approach achieves adequate precision with reduced processing requirements.
Data Source
AI summary
SLAM systems are provided that utilize an artificial neural network to both map environments and locate positions within the environments. In some example embodiments, a sensor arrangement is used to map an environment. The sensor arrangement acquires sensor data from the various sensors and associates the sensor data, or data derived from the sensor data, with spatial regions in the environment. The sensor data may include image data and inertial measurement data that effectively describes the visual appearance of a spatial region at a particular location and orientation. This diverse sensor data may be fused into camera poses. The map of the environment includes camera poses organized by spatial region within the environment. Further, in these examples, an artificial neural network is adapted to the features of the environment by a transfer learning process using image data associated with camera poses.