An Improved Method of Pose Estimation for Lighthouse Base Station Extension

Yi Yang, Dongdong Weng, Dong Li, Hang XunSensors 2017 · DOI

TL;DR

This paper introduces a novel pose estimation algorithm designed to overcome the limitations of HTC/Valve's Lighthouse tracking system, specifically its inability to expand base stations and its susceptibility to occlusion. The proposed method unifies data from multiple base stations, enabling robust tracking even with a limited number of detected sensors, achieving millimeter-level precision.

What Problem Does This Paper Address?

The Lighthouse tracking system, while lauded for its accuracy and low latency in virtual reality, has significant limitations. Firstly, it doesn't natively support expanding the number of base stations, restricting the tracking area. Secondly, it struggles with occlusion – when moving targets block the line of sight to sensors – leading to lost tracking data and unstable poses. This makes it difficult to track objects reliably in dynamic environments or when using a minimal set of sensors. The paper aims to address these issues by developing a more flexible and robust pose estimation method.

Methods

The core of the paper is an improved pose estimation algorithm that treats each base station as a pin-hole camera. The key innovation lies in its ability to unify data from multiple base stations. Instead of requiring a fixed number of sensor detections from a single base station, the algorithm can successfully estimate an object's pose as long as a total of three or more sensors across any of the base stations detect a signal. This unified approach allows for greater flexibility in sensor placement and robustness against partial occlusions.

The algorithm leverages the known intrinsic parameters of the base stations and transforms the problem into a traditional Perspective-n-Point (PnP) problem. It then employs an iterative optimization method, specifically using MATLAB's optimization toolbox, to solve for the rotation and translation matrices that define the object's pose. This approach is designed to handle various scenarios, including those where only a few sensors are visible.

Datasets

Not discussed in the paper. The experiments were conducted using HTC official base stations and custom-built receivers, and simulations were performed in MATLAB.

Key Insights and Ideas

  • Unified Multi-Base Station Framework: The most significant insight is the development of a framework that seamlessly integrates data from multiple base stations. This moves beyond the single-base-station limitation of the original Lighthouse system.
  • Robustness to Limited Sensor Data: The algorithm's ability to achieve precise positioning with as few as three total detected sensors (regardless of which base station they are associated with) is a major breakthrough for handling occlusions and expanding tracking coverage.
  • Flexibility in Receiver Design: The paper highlights the use of customizable receivers with varying numbers and arrangements of photodiodes, demonstrating the adaptability of the proposed system.

Key Results

The experimental results demonstrate the effectiveness of the proposed algorithm:

  • Positioning Accuracy: The system achieves millimeter-level precision, with an average positioning error of 1.4653 mm. The maximum error observed was 10.6 mm at a distance of 5 meters from the base stations.
  • Angular Accuracy: The angular precision is within the minute level, with an average angle precision of 0.3199°.
  • Jitter: The jitter range is reported to be less than 0.4%, indicating a stable tracking performance.
  • Frame Rate: The hardware frame rate is 272 Hz, and the system frame rate is 70 Hz.
  • Latency: The average latency of the system is less than 0.047 seconds.
  • Simulation: Simulation experiments confirm that the algorithm is equally applicable to multi-base station tracking systems, with pose estimation precision remaining stable as the number of stations and points increase.

Limitations and Future Work

The authors acknowledge that while the system achieves high precision, there is room for improvement, particularly in hardware stability. Future work will focus on:

  • Improving hardware stability.
  • Increasing the number of processing unit ports.
  • Expanding the number of base stations to achieve tracking over a broader range.

Summary updated

Abstract

In 2015, HTC and Valve launched a virtual reality headset empowered with Lighthouse, the cutting-edge space positioning technology. Although Lighthouse is superior in terms of accuracy, latency and refresh rate, its algorithms do not support base station expansion, and is flawed concerning occlusion in moving targets, that is, it is unable to calculate their poses with a small set of sensors, resulting in the loss of optical tracking data. In view of these problems, this paper proposes an improved pose estimation algorithm for cases where occlusion is involved. Our algorithm calculates the pose of a given object with a unified dataset comprising of inputs from sensors recognized by all base stations, as long as three or more sensors detect a signal in total, no matter from which base station. To verify our algorithm, HTC official base stations and autonomous developed receivers are used for prototyping. The experiment result shows that our pose calculation algorithm can achieve precise positioning when a few sensors detect the signal.
BibTeX
@inproceedings{yang2017improved,
  author = {Yi Yang and Dongdong Weng and Dong Li and Hang Xun},
  title = {{An Improved Method of Pose Estimation for Lighthouse Base Station Extension}},
  booktitle = {Sensors},
  year = {2017},
  doi = {10.3390/s17102411},
}
Figure