# Tracking

Here, we will be explaining the necessity of 6DoF tracking for AAR, and the challenges we have faced with indoor tracking systems in our multi-room space with low ceiling. We'll go through the tracking systems we have tested as well as the current solution. We will also discuss the next step of using inside-out tracking.

:!: This page is still a work in progress. :!:

## Location tracking (indoors)

----
### Optical outside-in

----

#### Body tracking with ZED depth cameras
Our current solution.

**Advantages**
  -Enables kinesthetic interaction by tracking e.g. arm movements
  -Modular camera system with WiFi enables setup in delicate and complex indoor environments
  -No need for multisensor fusion; one camera per area is enough, although multiple cameras improve redundancy e.g. in case of occluded view

**Challenges**
  -Identification of different persons difficult
    -Fiducial markers (ArUco); our current solution
    -Face recognition; tested, but not good results with OpenCV; better results with Visage, but expensive; also, users' privacy concerns must be respected 
    -Feature recognition (clothes colours, etc); tested with bad results
  -Cameras + computers become rather expensive
  -Requires certain amount of light (doesn't work in dim or dark conditions)
  -Generates latency due to intensive computing

----

#### OptiTrack

**Advantage**  
  -Accurate
  -Tracks head orientation + other objects
  -Small latency

**Challenges**
  -Expensive
  -Trackable objects need to be equipped with optical markers
  -Not very discrete
  -Potentially not suitable for venues with complex multi-room layout
  -Potentially not possible to install in delicate buildings

----

#### VIVE Tracker

**Advantages**
  -Accurate
  -Tracks orientation
  -Inexpensive

**Challenges**
  -Limited tracking area when using four base stations
  -More than four base stations not officially supported
  -Potential issues with setting up and interface  

----

### Radio-frequency outside-in

----

#### Pozyx (UWB)

**Advantages**
  -Scalable system
  -Orientation tracking (IMU) integrated into the tag

**Challenges**
  -Potentially inaccurate, especially 3D tracking (although examples of well-working solutions exist, too) 
  -Sensitive to environmental interference leading to unpredictable tracking
  -Expensive when expanding the system

----

#### Quuppa (BLE AoA)

**Advantages**
  -Scalable system

**Challenges**
  -Doesn't work with low ceilings
  -2D tracking only

----
### Ultrasonic outside-in

----

#### Marvelmind

**Advantages**
  -Relatively accurate

**Challenges**
  -Makes audible tick sounds
  -Very picky with LOS

---
-
### Optical inside-out / Visual odometry

----

#### VIVE Ultimate Tracker
**Advantages**
  -Self-sufficient without a need for external base stations
  -Rather accurate
  -Small latency

**Challenges**
  -Limited tracking area
  -Integration with non-HMD solutions still in beta
  -Setting up potentially difficult and buggy  

----

#### Mixed reality headsets
Such as Apple Vision Pro, Microsoft Hololens, Meta Quest, Varjo, etc...

**Advantages**
  -Rather accurate tracking
  -Built-in processor capacity

**Challenges**
  -Distract vision
  -Heavy and bulky
  -Some are expensive

----

#### Iphone
**Advantages**
  -Self-sufficient without a need for external base stations
  -Quite accurate
  -Opens up many extra tracking features

**Challenges**
  -Expensive units
  -Rigging may be challenging on headphones (shape and weight)


----

#### Self-made custom rig
Head-mounted camera(s) connected to an onboard SBC -- or camera feed streamed wirelessy to an external computer -- for visual odometry calculation. Not tested yet, but has some potential.

----

## Head-tracking / Orientation tracking

----
### IMU

**Advantages**
  - Accurate
  - No need for external sensors

**Challenges**
  - Drifting issues when magnetometer cannot be used (often the case indoors)

----