Instrumentation & Measurement Magazine 25-5 - 24

detection, and rescue in dangerous environments like storms or
fires. Military UAS can carry out reconnaissance, surveillance,
tracking, strikes, communication relays, joint operations, etc.
Due to the excellent task performance of UAS and the support
of various development plans like " Unmanned Systems
Integrated Roadmap 2017-2042, " the number of UAS has
shown a rapid growth trend. Moreover, swarm and intelligence
have become the main research interests for UAS in the
future. However, with the continuous expansion of UAS functions,
the complexity of the systems continues to increase,
resulting in raised risks of system anomalies or failures. Coupled
with the control of development costs and the lack of
real-time response from their pilots, the safety and reliability
of UAS are lower than those of manned systems.
Taking the UAV as an example, statistics show that system
failures are the main cause of accidents. Due to external interference
or self-degradation, abnormality or failure of sensors,
actuators, mechanical power transmission components, etc.,
will directly lead to attitude instability, unexpected actions, or
propulsion loss, and then cause a crash accident. The typical
accident cases given in the " accidents will happen " published
by the Drone Wars company in 2018 included [2]:
◗ In 2017, the inertial navigation system of an RQ-4 Global
Hawk failed, and the failure was not detected in time,
which led to attitude instability and a crash accident.
◗ In 2015, the left tail and tail insert of an MQ-1B Predator
fell off, and consequently the UAV became uncontrollable.
The UAV entered an unrecoverable spin and crashed.
◗ In 2015, a compressor bearing failure of an MQ-9A Reaper
led to engine loss. The impact destroyed the UAV, four
missiles, and one bomb.
Therefore, the condition monitoring of UAS has attracted
the attention of researchers, who are expected to design monitoring
systems that will grasp the abnormal or faulty operation
status of UAS in time and take mitigation measures to avoid
disaster accidents. Meanwhile, in the process of the UAS executing
the task, there are link delays in remote operation
and link loss in complex environments, which forces UAS
condition monitoring to be completed online to ensure the
timeliness of the response.
Condition monitoring is the process of monitoring the condition
parameters in UAS to recognize an undesired function
or performance change due to an incipient fault or degradation.
OCM means the process of identifying the current health
status of UAS through the real-time data stream of the condition
parameters [3] (Fig. 1), where perception, intelligent
assessment, and edge computing are the key components.
Generally, UAS contain sufficient sensors, such as a global positioning
system (GPS), an inertial measurement unit (IMU), a
vibration sensor, etc. The data generated by these sensors can
reflect the UAS status in real-time, which can be used as the
perception component of OCM. On this basis, an intelligent assessment
model is established based on AI methods to identify
the current condition of the UAS through analyzing perceived
data. For instance, anomaly or fault detection models based
on AI algorithms like convolutional neural networks (CNN),
24
recurrent neural networks (RNN), and Gaussian process regression
are proposed to achieve the monitoring of sensors
and actuators in UAS [4]-[6]. In addition, a series of remaining
useful life prediction models based on CNN is designed to
analyze vibration data to realize the condition monitoring of
bearings and gears in mechanical transmission systems [7], [8].
After constructing the intelligent assessment model, an edge
computing platform needs to be provided on the UAS side
to ensure the timeliness of condition monitoring. These three
components in OCM are necessary to provide an overview of
system real-time health status, and the benefits of OCM are
summarized as follows:
◗ Intelligent assessment is used to replace manual data
analysis, which greatly improves the efficiency of condition
monitoring.
◗ Edge computing provides online and continuous condition
monitoring capability for the life cycle of UAS.
◗ Timely abnormal or fault condition information is
provided to support UAS to quickly take fault mitigation
measures or arrange maintenance plans.
Guidelines for Designing Embedded
Online Condition Monitoring
Instrument
To achieve UAS OCM, the embedded OCM instrument needs
to be developed to accompany the operation of the UAS and
assess its condition in real-time. Since UAS have enough onboard
sensors, the embedded OCM instrument can collect
the sensor readings online through the data bus to realize the
perception. In addition, some researchers have carried out
studies on condition monitoring based on AI algorithms to
achieve functions such as anomaly detection, fault diagnosis,
and remaining useful life prediction [9]. The above-mentioned
works lay the methodological foundations for the intelligent
assessment model of the embedded OCM instrument.
AI algorithms usually bring high time complexity while
providing high performance, and the current UAS mainly relies
on the high-performance computing platform in the ground
control station to realize condition monitoring, such as a server
or a workstation with GPU. With the complexity of the working
environment and scenarios, problems such as link delay,
limited bandwidth, and link loss greatly restrict the timeliness
and availability of UAS condition monitoring, which in turn
poses safety threats due to untimely detection of abnormal or
fault status. Online intelligent analysis of the perceived data
will reduce the delay and avoid the other problems and risks
mentioned above. In addition, it also enhances the security of
private data. This means the development of the UAS OCM instrument
has an urgent need for edge computing to support the
online operation of the intelligent assessment model.
However, we cannot directly deploy servers and workstations
into UAS under the limitations of SWaP, which pose new
challenges to the timeliness of OCM. How to realize the online
operation of complex AI algorithms in the UAS has become
the bottleneck in developing the embedded OCM instrument.
Hence, this section focuses on the introduction of the currently
IEEE Instrumentation & Measurement Magazine
August 2022
Instrumentation & Measurement Magazine 25-5

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 25-5
