Location-based VR Tracking Solution

12/07/2017
Blog, Product

LPVR Location-based VR Tracking Introduction

NOTE: In case you are looking for our LPVR middleware for automotive and motion simulator applications, please refer to this page.

UPDATE 1: Full SteamVR platform support

UPDATE 2: Customer use case with AUDI and Lightshape

UPDATE 3: Customer use case with Bandai Namco

Consumer virtual reality head mounted display (HMD) systems such as the HTC VIVE support so-called room scale tracking. These systems are able to track head and controller motion of a user not only in a sitting or other stationary position, but support free, room-wide motions. The volume of this room scale tracking is limited to the capabilities of the specific system, usually covering around 5m x 5m x 3m. Whereas for single user games or applications this space may be sufficient, especially multi-user, location-based VR applications such as arcade-style game setups or enterprise applications require larger tracking volumes.

Optical tracking systems such as Optitrack offer tracking volumes of up to 15m x 15m x 3m. Although the positioning accuracy of optical tracking systems are in the sub-millimeter range, especially orientation measurement is often not sufficient to provide an immersive experience to the user. Image processing and signal routing may introduce further latencies.

Our locations-based VR / large room-scale tracking solution solves this problem by combining optical tracking information with inertial measurement data using a special predictive algorithm based on a head motion model.

Compatible HMDs: HTC VIVE, HTC VIVE Pro
Compatible optical tracking systems: Optitrack, VICON, ART, Qualisys, all VRPN-compatible tracking systems
Compatible software: Unity, Unreal, Autodesk VRED, all SteamVR-compatible applications

This location-based VR solution is now available from LP-RESEARCH. Please contact us here for more information or a price quotation.

IMU and Optical Tracker Attachment

The system follows each HMD using a combination of optical and IMU tracking. A special 3D-printed holder is used to attach a high-quality IMU (LPMS-CU2) and optical markers to an HTC VIVE headset.

HTC VIVE with LP holder and IMU attached

To create a perfectly immersive experience for the user, optical information is augmented with data from an inertial measurement unit at a rate of 800Hz. Additionally to the high frequency / low-latency updates, a head motion model is used to predict future movements of the player’s head. This creates an impression of zero latency gameplay.

Overview of solution key functionality

Tracking Camera, HMD and Player Setup

In its current configuration the system can track up to 20 actors simultaneously, each holding a VIVE controller to interact with the environment. Players wear backpack PCs to provide visualization and audio.

Complete location-based VR system setup

Playground and Camera Arrangement

The solution covers an area of 15m x 15m or more. There is an outer border of 1.5m that is out of the detection range of the cameras. This results in an actual, usable playground area of 13.5m x 13.5m. The overall size of the playground is 182.25m². The cameras are grouped around the playground to provide optimum coverage of the complete area.

Location-based VR playground setup

 

Contact us for further information.

Optical-Inertial Sensor Fusion

31/03/2017
, , , Projects, Use Cases

Optical position tracking and inertial orientation tracking are well established measurement methods. Each of these methods has its specific advantages and disadvantages. In this post we show an opto-inertial sensor fusion algorithm that joins the capabilities of both to create a capable system for position and orientation tracking.

How It Works

The reliability of position and orientation data provided by an optical tracking system (outside-in or inside-out) can for some applications be compromised by occlusions and slow system reaction times. In such cases it makes sense to combine optical tracking data with information from an inertial measurement unit located on the device. Our optical-intertial sensor fusion algorithm implements this functionality for integration with an existing tracking system or for the development of a novel system for a specific application case.

The graphs below show two examples of how the signal from an optical positioning system can be improved using inertial measurements. Slow camera framerates or occasional drop-outs are compensated by information from the integrated inertial measurement unit, improving the overall tracking performance.

Combination of Several Optical Trackers

For a demonstration, we combined three NEXONAR IR trackers and an LPMS-B2 IMU, mounted together as a hand controller. The system allows position and orientation tracking of the controller with high reliability and accuracy. It combines the strong aspects of outside-in IR tracking with inertial tracking, improving the system’s reaction time and robustness against occlusions.

Optical-Inertial Tracking in VR

The tracking of virtual reality (VR) headsets is one important area of application for this method. To keep the user immersed in a virtual environment, high quality head tracking is essential. Using opto-inertial tracking technology, outside-in tracking as well as inside-out camera-only tracking can be significantly improved.

Robot Operating System and LP-Research IMUs? Simple!

01/03/2017
, , , , , , , Blog, Use Cases
NOTE: We have released a new version of our ROS / ROS 2 driver, please refer to this post.


Introduction

Robot Operating System (ROS) is a tool commonly used in the robotics community to pass data between various subsystems of a robot setup. We at LP-Research are also using it in various projects, and it is actually very familiar to our founders from the time of their PhDs. Inertial Measurement Units are not only a standard tool in robotics, the modern MEMS devices that we are using in our LPMS product line are actually the result of robotics research. So it seemed kind of odd that an important application case for our IMUs was not covered by our LpSensor software: namely, we didn’t provide a ROS driver.  We are very happy to tell you that such a driver exists, and we are happy that we don’t have to write it ourselves: the Larics laboratory at the University of Zagreb are avid users of both ROS and our LPMS-U2 sensors. So, naturally, they developed a ROS driver which they provide on their github site.  Recently, I had a chance to play with it, and the purpose of this blog post is to share my experiences with you, in order to get you started with ROS and LPMS sensors on your Ubuntu Linux system.

Installing the LpSensor Library

Please check our download page for the latest version of the library, at the time of this writing it is 1.3.5. I downloaded it, and then followed these steps to unpack and install it:

I also installed libbluettoth-dev, because without Bluetooth support, my LPMS-B2 would be fairly useless.

Setting up ROS and a catkin Work Space

If you don’t already have a working ROS installation, follow the ROS Installation Instructions to get started. If you already have a catkin work space you can of course skip this step, and substitute your own in what follows.  The work space is created as follows, note that you run catkin_init_workspace inside the src sub-directory of your work space.

Downloading and Compiling the ROS Driver for LPMS IMUs

We can now download the driver sources from github. It optionally makes use of and additional ROS module by the Larics laboratory which synchronizes time stamps between ROS and the IMU data stream.  Therefore, we have to clone two git repositories to obtain all prerequisites for building the driver.

That’s it, we are now ready to run catkin_make to get everything compiled and ready.  Building was as simple as running catkin_make, but you should setup the ROS environment before that.  If you haven’t, here’s how to do that:

This should go smoothly. Time for a test.

Not as Cool as LpmsControl, but Very Cool!

Now that we are set up, we can harness all of the power and flexibility of ROS. I’ll simply show you how to visualize the data using standard ROS tools without any further programming.  You will need two virtual terminals.  In the first start roscore, if you don’t have it running yet.  In the second, we start rqt_plot in order to see the data from our IMU, and the lpms_imu_node which provides it.  In the box you can see the command I use to connect to my IMU. You will have to replace the _sensor_model and _port strings with the values corresponding to your device.  Maybe it’s worth pointing out that the second parameter is called _port, because for a USB device it would correspond to its virtual serial port (typically /dev/ttyUSB0).

Once you enter these commands, you will then see the familiar startup messages of LpSensor as in the screenshot below. As you can see the driver connected to my LPMS-B2 IMU right away. If you cannot connect, maybe Bluetooth is turned off or you didn’t enter the information needed to connect to your IMU.  Once you have verified the parameters, you can store them in your launch file or adapt the source code accordingly.

Screenshot starting LPMS ROS node

Screenshot of starting the LPMS ROS node

The lpms_imu_node uses the standard IMU and magnetic field message types provided by ROS, and it publishes them on the imu topic.  That’s all we need to actually visualize the data in realtime.  Below you can see how easy that is in rqt_plot. Not as cool as LpmsControl, but still fairly cool. Can you guess how I moved my IMU?

animation of how to display LPMS sensor data in ROS

Please get in touch with us, if you have any questions, or if you found this useful for your own projects.

Update: Martin Günther from the German Research Center for Artificial Intelligence was kind enough to teach me how to pass ROS parameters on the command line.  I’ve updated the post accordingly.

New Miniature Sensor In: LPMS-ME1 Maker Edition!

02/11/2016
, Blog, Product
20160808lpmsme1onfinger600x400_1

The LPMS-ME1’s Maker Edition is miniature-sized with just 12 x 12 x 2.6 mm.

We proudly present you our latest development! The LPMS-ME1 is our smallest motion sensor so far, with just 12 x 12 x 2.6 mm it is tiny! Despite its size this powerful 9-axis inertial measurement unit (IMU) has several sensors integrated, for example a 3-axis accelerometer, a 3-axis gyroscope and a 3-axis magnetometer. And this miniature motion sensor certainly comes at low cost.

It is very easy to assemble and can be conveniently embedded in the system of your choice. Due to its size it is perfect for your design ideas and development projects. Just to to give you some inspiration, it can be used for human motion capture or sports performance evaluation, for various sorts of Internet of Things (IoT) devices, and can be used to control unmanned aerial vehicles. You can even fly a drone with it!

Have a look at more specifications in our data sheet. This sensor comes with our own LpmsControl utility software and a one-year warranty service.

Get the LPMS-ME1 Maker Edition for your own innovations! Find a distributor of your choice or order online at Zenshin Tech.

20160808lpmsme1_box_600x400

 

Machine Learning for Context Analysis

25/10/2016
, , Projects

Deterministic Analysis vs. Machine Learning for Context Analysis

Machine learning for context analysis and artificial intelligence (AI) are important methods that allow computers to classify information about their environment. Today’s smart devices integrate an array of sensors that constantly measure and save data. On the first thought one would image that the more data is available, the easier it is to draw conlusions from this information. But, in fact larger amounts of data become harder to analyze using deterministic methods (e.g. thresholding). Whereas such methods by themselves can work efficiently, it is difficult to decide which analysis parameters to apply to which parts of the data.

Using machine learning techniques on the other hand this procedure of finding the right parameters can be greatly simplified. By teaching an algorithm which information corresponds to a certain outcome using training and verification data, analysis parameters can be determined automatically or at least semi-automatically. There exists a wide range of machine learning algorithms including the currently very popular convolutional neural networks.

Context analysis setup overview

Figure 1 – Overview of the complete analysis system with its various data sources

Context Analysis

Many health care applications rely on the correct classification of a user’s daily activities, as these reflect strongly his lifestyle and possibly involved health risks. One way of detecting human activity is monitoring their body motion using motion sensors such as our LPMS inertial measurement unit series. In the application described here we monitor a person’s mode of transportation, specifically

  1. Rest
  2. Walking
  3. Running
  4. In car
  5. On train

To illustrate the results for deterministic analysis vs. machine learning for context analysis approach we first implemented a state machine based on deterministic analysis parameters. An overview of the components of this system are shown in Figure 1.

Deterministic approach overview

Figure 2 – Deterministic approach

The result (Figure 2) is a relatively complicated state machine that needs to be very carefully tuned. This might have been because of our lack of patience, but in spite of our best efforts we were not able to reach detection accuracies of more than around 60%. Before spending a lot more time on manual tuning of this algorithm we switched to a machine learning approach.

Machine learning approach overview

Figure 3 – Machine learning approach

The eventual system structure shown in Figure 3 looks noticeably simpler than the deterministic state machine. Besides standard feature extraction, a central part of the algorithm is the data logging and training module. We sampled over 1 milion of training samples to generate the parameters for our detection network. As a a result, even though we used a relatively simple machine learning algorithm, we were able to reach a detection accuracy of more than 90%. A comparison between ground truth data and classification results from raw data is displayed in Figure 4.

Context analysis algorithm result

Figure 4 – Result graphs comparing ground truth and analysis output for ~1M data points

Conclusion

We strongly belief in the use of machine learning / AI techniques for sensor data classification. In combination with LP-RESEARCH sensor fusion algorithms, these methods add a further layer of insight for our data anlysis customers.

If this topic sounds familiar to you and you are looking for a solution to a related problem, contact us for further discussion.

1 … 6 7 8 9 10 … 14