Pose Predictor for Moving Platform Using IMU Signal Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for presenting content via electronic devices fail to accurately account for movement-based attributes, particularly distinguishing between user motion and motion of the moving platform, leading to inaccuracies in rendering virtual content.
Innovation Solution
A machine learning model is trained to distinguish between user motion and platform motion using inertial measurement unit (IMU) data, incorporating loss functions for accuracy and smoothness penalties, and utilizes additional sensors for enhanced localization, enabling prediction of future device poses to synchronize user perception with rendered surroundings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion sensor data from IMU is used to determine device pose, then motion tracking is enabled, but user motion cannot be distinguished from platform motion leading to rendering inaccuracies
Solution Approach 1:
The patent segments the combined motion signal into distinct components: platform motion and user motion. The machine learning model processes IMU data to separate these two motion sources, allowing independent analysis and compensation of each component. This segmentation enables accurate attribution of motion to its correct source, resolving the fundamental problem of indistinguishable motion signals.
Solution Approach 2:
The patent introduces machine learning models as intermediary processing layers between the raw IMU sensor data and the device pose determination. These models act as mediators that analyze the combined motion signal, identify patterns characteristic of platform versus user motion, and produce separated motion components. This intermediary processing enables accurate motion source discrimination without requiring separate physical sensors.
2Productivity
If current device pose is used for rendering, then rendering simplicity is maintained, but synchronization between user perception and rendered surroundings deteriorates due to motion delay
Solution Approach 1:
The patent applies preliminary action by predicting future device poses before they actually occur. The machine learning model analyzes current and historical motion data to extrapolate device orientation and position at future time points. This predictive approach allows the rendering system to prepare and display content corresponding to the user's anticipated viewpoint, eliminating perceptible delay between user motion and visual feedback.
Solution Approach 2:
The patent transitions from static pose-based rendering to dynamic predictive pose rendering. Instead of rendering based on current or past device poses, the system continuously updates rendered content based on predicted future poses that adapt to the user's motion trajectory. This dynamic approach maintains synchronization with the user's perception while preserving rendering efficiency through optimized prediction algorithms.
Data Source
AI summary
Various implementations disclosed herein include devices, systems, and methods that isolates movement of a user from movement of a platform moving with the user. For example, a process may obtain motion sensor data corresponding to an electronic device while the electronic device is located on a moving platform. The motion sensor data includes a measurement representing a combined motion of a user and the moving platform. The process may further extract from the motion sensor data, user motion data representing motion of the user. The process may further allocate the extracted user motion data as input for user motion analysis.