IEEE Systems, Man and Cybernetics Magazine - October 2020 - 16

Most of these approaches use an absolute reference
frame that provides a global position coordinate for the
device that is being localized. While this can be advantageous, a simple relative reference frame often is all that is
required. Because such systems rely on coordinating multiple devices located at different positions in the environment, they become dependent on some form of communication
or collaboration between different sensors/modules to
determine the relative location of the object being localized. This arrangement not only introduces communication-related issues like delay, signal-to-noise ratio, unintended
interruptions, congestion, medium-specific range, and so
on, but it also correlates the system's accuracy based on
the precise positioning of each such sensor/module. Positioning on a larger scale is subject to higher rates of human
error as opposed to a more contained single-point sensor-
actuator pair.
In this article, we present an infrared (IR)-based localization solution based on the HTC Vive tracker system
(VIVEPOSE) [26]. We show how this inexpensive system
can be used for accurate indoor localization. It does not
rely on communication networks and hence avoids all
communication-related issues. Human error in positioning the beacon is neutralized, since there is only one point
of reference from which the position of the object is
inferred. While the sensor array position is important to
accurately measure the object's location, it is much less
susceptible to positioning errors since they are manufactured on a printed circuit board, which has a much higher
level of accuracy over the scale of the localization volume.
We then show how the VIVEPOSE system can be
employed for a leader-follower-based formation controller
in a robotics application.
Localization Using the HTC Vive Tracker
Principal of Operation
The HTC Vive positioning system (VIVEPOSE) has several
benefits over other camera-based methods. The most useful is its ability to track any number of objects, as long as
they are not occluded and are within the allowable tracking range. According to the HTC Vive user manual, the recommended maximum tracking distance from the tracker
to the lighthouse is 5 m and the recommended minimum
tracking distance from the tracker to the lighthouse is
1.5 m [26]. This result was tested and verified by the
authors in a controlled laboratory environment, where reliable tracking was provided by using the system over a
5 # 5 # 3 m 3 volume, in which the restriction on height
was set to 3 m only due to the limitation of placing the Vive
lighthouse close to the ceiling.
The tests remain valid even in partial lighting conditions
and in the presence of other light-emitting sensors, since
the measurement of the light intercept is time sensitive.
This provides a form of error correction and, in case of
erroneous noise, the filter application such as extended
16

IEEE SYSTEMS, MAN, & CYBERNETICS MAGAZINE O ctober 2020

Kalman filtering (EKF) eliminates the noisy sample. Such
noise is acceptable for the formation controller since the
rate of localization acquisition is higher and because no
communication delay avoids sequencing errors. The VIVEPOSE system can infer location from sensor data with a
priori knowledge of sensor placement in a multisensor configuration as well as the use of a synchronous sweep pulse.
As a result, a computation of location can be performed on
the sensing device, allowing the VIVEPOSE system to avoid
the growth of linear computational complexity when the
number of tracked objects is increased. This also avoids
communication problems across devices, such as network
delays and interference.
The VIVEPOSE system uses IR laser projectors called
lighthouses, which synchronously sweep the area in front
of them by alternating between horizontal and vertical
planes. At the start of each sweep, IR synchronization
pulses are blasted into the area for the sensors to initiate.
The sweep area covered by the lighthouse's laser is
across a 90° symmetrical angle around the optical axis of
the frustum in view. The sensors within this frustum
measure the time interval between intercepting the synchronous IR burst and the sweep along an axis. When
there is more than one lighthouse, the sweeps alternate
between each.
The IR sync and laser sweeps are performed in a controlled frequency of 120 Hz, so each complete sweep takes
approximately 1/120 = 8.33 ms and is controlled to be linearly incremental with respect to angle. Hence, the corresponding dt (in seconds) measured between the intercept
of the sync pulse and the sweep is directly correlated to
the angular offset of the sensor in the frustum view of the
lighthouse. Given dt, the angle offset of the sensor i
along one axis of the lighthouse can be computed as in (1).
Using Figure 1 as a reference example, the interception of
the lighthouse's horizontal sweep with a sensor, where
z = 90° represents the extremes of the sweep:
i
dt
= 1
z
120
i = 120 zdt.

(1)

VIVEPOSE uses the HTC Vive trackers, which are built
using an array of 21 IR sensors, in different positions, carefully positioned to intercept IR rays along a 270° field of
view (FOV) [27]. As a result, each of these sensors provide
a pair of sweep angle readings from the lighthouse for the
horizontal and vertical sweeps, which offers the perspective of each sensor with respect to the lighthouse. Since
the tracker has six degrees of freedom and a rigid body,
and the model geometry of the tracker is known, the problem formulates to a model-based pose-estimation problem.
Solutions to such a problem can be achieved using several
methods, including perspective-n-point (PnP), to provide
absolute orientation and location with respect to the lighthouse coordinate system.

IEEE Systems, Man and Cybernetics Magazine - October 2020

Table of Contents for the Digital Edition of IEEE Systems, Man and Cybernetics Magazine - October 2020