“Helping Machines Make Sense of Nonsense”

“Helping Machines Make Sense of Nonsense” - New robotics technology for autonomous fault
diagnosis and fault tolerant control.
Modern society relies on the availability and smooth operation of complex engineering systems.
Examples include electric power systems, water distributions networks, transportation systems,
robotic systems, intelligent buildings, and many more. These systems have gradually evolved to
become very large scale and more complex whilst at the same time there are greater
expectations for these systems to function more intelligently, reliably and efficiently. We rely on
the uninterrupted operation of such systems and when they fail, the consequences can be
immense, in terms of societal, health, and economic effects. Think for example a power blackout
or a contamination event of a water distribution network. It is therefore important that these
systems are able to work, very often continuously and with little disruption if there is a failure.
The correct and efficient operation of these systems relies on monitoring and control
infrastructures that consist of sensor/actuator networks. The scale and complexity of these
systems requires the development of several sophisticated monitoring and control applications
where a large amount of real-time data about the monitored environment is collected and
processed to activate the appropriate actuators and achieve the desired objectives. However, in
many cases the collected data from all sensors may not make much sense! For example in a
mobile robotic scenario, one robot’s proximity sensor detects an obstacle while its distance
sensor measures a long distance to the obstacle. What does the robot do?
For humans, the decision may be easy because we have redundant sensory information which
we are able to process in real-time and correctly assess the situation. Furthermore, we have
very good confidence in the state of our sensory organs. For machines however, the decision is
not as easy.
In machines, information redundancy is limited due to cost and, in addition, the state of the
sensors and actuators in not always known and several things can go wrong. For example, some
measurements may be missing, sensor performance may be deteriorating due to aging or
environmental conditions, sensors may be drifting, etc. In some cases, data coming from
different sensors (or actuators), on the same unit (e.g., with different resolution) or residing in a
cluster, may become inconsistent. To complicate matters machines usually operate in
environments that can be subject to nonstationarity phenomena and the electronics are prone
to drifts and soft/hard faults. Such problems, which will generate “nonsense data,” are
generally the result of some faults in the sensor/actuator system itself or an abnormality in the
monitored environment, which may be either permanent or temporary, developing abruptly or
incipiently. These problems become more pronounced as sensing/actuation systems get older.
As complex engineering systems are integral to many
areas of everyday life, it is important that they are
equipped with the necessary tools to “make sense of
nonsense” and ultimately help them to respond reliably
and autonomously to the increasingly complex and
multiple functions they are being expected to perform.
The project iSense has developed tools and designed
methodologies that will prevent situations where a
relatively “small” fault in one or more components (e.g., sensor, actuator, communication link)
may cause an overall system failure. This technology would enable complex engineering systems
to detect and isolate faults whilst continuing to perform their function even if a percentage of
the individual components have failed, minimizing the detrimental effects of complete system
failure. The project research team focused on innovative cognitive fault diagnosis approaches
that can learn characteristics or system dynamics of the monitored environment and adapt their
behaviour and predict missing or inconsistent data to achieve fault tolerant monitoring and
control.
The challenge
Large-scale complex engineering systems, equipped with a wide range of sensors can generate a
large volume of real-time data that correspond to the state of the environment. Thus there is a
pressing need for these data to be processed in real-time to extract meaning and knowledge out
of the ever-increasing electronic information that becomes available. A huge problem is that
often the data are inconsistent or some data are actually missing. Such problems, which will
generate “nonsense data,” are generally the result of some faults in the sensor/actuator system
itself or an abnormality in the monitored environment, which may be either permanent or
temporary, developing abruptly or incipiently. These problems become more pronounced as
sensing/actuation systems get older. Automatically detecting and identifying the faulty
components is a challenging problem. Suppose that two sensors are measuring the same
property. If they provide a different measurement, it is evident that one of the two is faulty but
which one? Adding more sensors significantly increases the system cost, thus such solution is
not preferred. The challenge is therefore to algorithmically correlate the measurements of
different sensors such that the faulty components are correctly identified.
The technology
The i-Sense platform can help machines “make sense of nonsense”. This is achieved through a
set of algorithms to detect faults (and where possible anticipate them), identify and isolate
them as soon as possible, and accommodate for them in future decisions or actuator actions.
The algorithms can be integrated with the sensors, actuators and feedback control system for
making the overall monitoring and control system more robust, adaptive and fault-tolerant to
sensor/actuator faults and system faults or abnormalities in the environment.
Most design approaches for autonomous fault diagnosis rely on the concept of analytical
redundancy, where fault detection and isolation is obtained based on an available mathematical
model of the healthy system. With this approach fault diagnosis is achieved by comparing actual
observations with the prediction of the model. In practice, however, such mathematical models
may not be accurate or not available at all. Therefore, there is a need to develop fault detection
methods, where the time history of the observed data and the inter-relations between spatially
distributed sensing are exploited. This project has adopted this view and has developed an
innovative cognitive fault diagnosis framework with learning algorithms for approximating,
during operation, key correlations between measured variables. This has allowed us to handle
highly unstructured environments, which goes well beyond the current state-of-the-art in the
area of fault diagnosis.
The methodologies developed have been applied and tested in various scenarios included
mobile robotics, environmental monitoring, intelligent water distribution networks and smart
buildings. These applications have shown that the derived methodologies can significantly
improve the performance, in terms of autonomy, of the monitoring and control systems in many
different contexts.
Applications & Relevance to Industry
The i-Sense prototype can provide the enabling technology for a wide range of potential
applications of great relevance to society. The derived methodologies have the potential to
influence applications for homeland security, critical infrastructure protection, sensor networks,
intrusion detection in computer networks and many more. In fact, most applications that have
an intrinsic need for cognitive methods could benefit from the fault tolerant architecture
developed in this project.
Several of the developed algorithms are ready to be deployed in industrial prototype systems
for further evaluations. The consortium’s industrial partner, STMicroelectronics, is a world
leader in building automation with a wide range of products aimed at increasing performance,
reducing power and simplifying the design of digital automation in buildings and factories. These
results are also expected to impact on many other technologies like for example the Wireless
Sensor Networks (WSNs) of the future.
i-Sense Robots detecting, isolating faults whilst continuing to work.
The i-Sense platform was applied to mobile robots for diagnosing possible faults affecting the
sensors used for monitoring and control of the robots.
The first robotic demo (on the left ) involved a single robot with fault tolerant control algorithms.
The second demo (on the right ) showed that the application of the developed fault diagnosis
algorithms to interconnected systems satisfying a leader-follower formation, assuming that, in
the general case, hundreds of robots could be operated in this formation (e.g. for transportation
of large objects) constituting a large-scale system.
The prototype was developed by a team of international interdisciplinary experts, with
integrated expertise in systems and controls, fault diagnosis, machine learning, evolutionary
computation, fault modelling, neural networks, sensor hardware design and manufacturing
within a common framework.
Coordinated by: KIOS Research Center for Intelligent Systems and Networks, University of
Cyprus.
Funded: FP7 ICT
Project website: http://www.i-sense.org/
i-Sense Function Library:
http://www.i-sense.org/open_library.html