Internet of Things

Designing adaptive cyber-physical systems – from undersea monitoring to landing on Mars

Share this post:

cyber-physical

Autonomous space exploration

Cyber-physical systems, or CPS for short, are sophisticated computer devices that work together to perform functions, control physical elements, and respond to human control. They are already being used in auto-pilot systems for aircraft, advanced robotic systems, smart grids, medical monitoring, and search and rescue. In fact, most Internet of Things (IoT) devices are CPS.

These systems  play a vital role in helping  realize new technologies like self-driving cars, advanced medical treatments, energy-efficient processes, and even lunar landing crafts. As the IoT becomes more sophisticated, we’ll need more of these systems that allow the devices themselves to interact and communicate with each other. The problem is that, today, there are simply no tools that support the simplified design or fast prototyping of adaptive cyber-physical systems – systems that understand internal and external operational context and adapt their behavior accordingly.

IBM Research – Haifa is leading a consortium of European partners known as CERBERO to bridge this gap. We’ll  define a new architecture and set of services to help others more easily design cyber-physical systems. And almost all the tools are open source or being designed that way. The architecture for these tools will be based on a fascinating group of use cases:

  • Ocean monitoring – This scenario, being led by Ambisense, will use smart video-sensing unmanned vessels that can monitor the surface and undersea environment. Our use case will define algorithms for data analysis and information fusion that lets the system quickly adapt its strategy to maintain the robots’ positions, even under changing environmental conditions. The undersea robots will be equipped with new sensing and processing abilities for navigation and operation, along with unique real-time data processing. Our goal is to make the robots remain charged via battery, solar, or wind energy to ensure communication during difficult arctic conditions when these systems usually go down.
  • Smart traveling for electric vehicles – This scenario embraces a virtual simulation environment for driver support inside a network of semi-autonomous electric vehicles. It also considers passenger preferences, energy saving factors, and mobility functions such as charge points, parking spaces, smart home, and congestion. Our goal is to create a framework for designing cyber-physical systems that have multiple dependencies and need to consider many – often conflicting – requirements. This case study involves numerous partners and is led by TNO and Centre Recerche Fiat.
  • Self-healing system for planetary exploration – This use case, led by Thales, focuses on t planetary exploration missions, including those combining the ability to move across the surface and to study Mars at depth. Our goal is to build a computer system that can dynamically reconfigure its own hardware and software to overcome failures caused by radiation or harsh conditions, and interact smoothly with astronauts, satellites, earth control stations, and more.

What’s the best way for complex systems to communicate and interact?

The big question we are pondering is: How can we build a model for huge complex systems so they know how to communicate with each other and interact with their environment. The end result will likely resemble a middleware platform that allows teams to easily build and operate adaptive cyber-physical systems – without it taking years to do. By providing teams with a methodology and framework for the design, it will also be easier to continuously improve the system and change it through advanced DevOps – without rebuilding.

cyber-physical

Ocean Monitoring

Take, for example, the use case for ocean monitoring. These robotic vessels work in fleets and can be equipped with video sensing or photographic equipment, along with many other sensors. To accomplish their tasks in the most optimal way, these vessels need to communicate with the human operator, ‘understand’ what to monitor, where to navigate, how to divide the task up between the fleet when there is no operator—and even how to split up the power available.

The challenge of designing artificially intelligent cyber-physical systems is especially exciting because the tool chain goes from the user all the way to the system and to run-time implementation—all using one sleek line.

The contribution from IBM Research is fascinating, offering a revolutionary approach to the concept of optimization. We’re starting at a very basic level with new theories to build a proof-of-concept over the next two and a half years. As coordinators of the project, responsible for the architecture of the solution, IBM Research  breaking new ground for systems that are adaptive, open source, and open to anyone that wants to participate.

 

Cyber-physical

CERBERO Team

IBM Research – Haifa is coordinating the consortium of 12 partners from 7 countries: University of Sassari, University of Cagliari, the Polytechnic University of Madrid, Università della Svizzera italiana, the Institut National des Sciences Appliqués de Rennes (France), the Dutch research centers TNO and Science and Technology, Abassula, the FIAT research center, Thales Alenia Space, and the UK group Ambisense LTd., all high-profile companies in the ICT.

More Internet of Things stories

Four Papers Advance Computational Argumentation in IBM’s Project Debater

The latest work on computational argumentation from the IBM Project Debater research team group is being presented at the ACL 2019 conference. Three papers will be presented at the main conference and one more paper will be presented in the co-located Argument Mining Workshop.

Continue reading

Label Set Operations (LaSO) Networks for Multi-Label Few-Shot Learning

Data augmentation is one of the leading methods to tackle the problem of few-shot learning, but current synthesis approaches only address the scenario of a single label per image, when in reality real life images may contain multiple objects. The IBM team came up with a novel technique for synthesizing samples with multiple labels.

Continue reading

AI Models Predict Breast Cancer with Radiologist-Level Accuracy

Our team of IBM researchers published research in Radiology around a new AI model that can predict the development of malignant breast cancer in patients within the year, at rates comparable to human radiologists.

Continue reading