6DoF Inside-Out Tracking Controller Using Multi-Processor VI-SLAM
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current VR/AR systems face challenges with the SLAM algorithm, including scale factor issues, drift problems, and long processing times, particularly in obtaining real-time 6DoF data for game controllers, which are not efficient enough to provide accurate 3D translation and rotation data.
Innovation Solution
A multi-processor structure is employed for VI-SLAM, using a splitting algorithm to distribute computation across vector processors for efficient 6DoF output generation, with a pre-generated static map for real-time localization and a calibrated stereo camera to address scale factor issues, and a client-server topology to reduce power consumption and processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If SLAM algorithm is used for localization, then localization accuracy is improved, but processing time increases and drift occurs
Solution Approach 1:
The SLAM system is divided into two independent parts: a pre-built static map generation phase and a real-time localization phase. The static map is constructed offline using historical data, while online localization uses pre-computed features and descriptors, eliminating the need for continuous map building and reducing real-time processing requirements.
Solution Approach 2:
The environment map and feature descriptors are pre-computed and stored before actual localization occurs. During real-time operation, the system retrieves pre-processed information from storage rather than computing it on-the-fly, significantly reducing processing time while maintaining accuracy.
2Measurement precision
If bag of words approach is used for loop-closure, then localization accuracy is improved, but CPU consumption increases
Solution Approach 1:
Feature descriptors and map data are pre-computed and stored during an offline phase. During online operation, the system simply retrieves and matches pre-processed information, eliminating the need for continuous bag-of-words updates and reducing CPU consumption significantly.
Solution Approach 2:
Instead of maintaining a dynamic bag of words that requires continuous updates, the system creates a static copy of the environment map with pre-computed features. This static representation can be efficiently queried without the computational overhead of maintaining and updating a dynamic vocabulary.
3Measurement precision
If game controller provides 6DoF data, then tracking accuracy is improved, but real-time performance deteriorates
Solution Approach 1:
The 6DoF tracking system is segmented into independent modules: IMU-based pose prediction, visual feature matching, and map lookup. Each module processes specific tasks independently, allowing the system to achieve real-time performance by parallelizing computations and avoiding sequential processing bottlenecks.
Solution Approach 2:
A pre-built static map acts as an intermediary between the sensor inputs and the final 6DoF output. The map stores pre-computed spatial relationships and feature correspondences, enabling rapid lookup and inference without requiring complex real-time computations, thus achieving both accuracy and real-time performance.
Data Source
AI summary
Methods and apparatus are provided for 6DoF inside-out tracking game control. In one novel aspect, a multi-processor architecture is used for VI-SLAM. In one embodiment, the apparatus obtains overlapping image frames and sensor inputs of an apparatus, wherein the sensor inputs comprise gyrometer data, accelerometer data and magnetometer data, splits computation work onto a plurality of vector processors to obtain six degree of freedom (6DoF) outputs of the apparatus based on a splitting algorithm, and performs a localization process to generate 6DoF estimations, and a mapping process to generate a cloud of three-dimensional points associated to the descriptors of the map. In one embodiment, the localization process and mapping process are configured to run sequentially. In another embodiment, the localization process and mapping process are configured to run in parallel.