PhD project title

Energy-Conscious and Fault-Resilient Autonomous Industrial Robots via Algorithm-Hardware Co-Design

Outline description, including interdisciplinary, intersectoral and international dimensions (300 words max)

Industry 4.0 is a new emerging paradigm aiming at transforming the manufacturing processes by automating them through the use of autonomous robots and smart edge/cloud computing infrastructure. The operation of robots relies on a range of computationally intensive algorithms from control theory to computer vision and machine learning for autonomously moving, deciding and performing critical operations in real time. Most of these operations need to be performed under stringent power envelopes and dynamically changing environmental conditions that include varying levels of heat, electromagnetism, radiation and humidity. Such conditions make the delicate electronic processing and storage circuits prone to intermittent failures, which in turn threaten the correct functionality of robots and eventually the reliable, non-disruptive and safe, accident-free operation of factories.  This project aims at addressing a truly interdisciplinary challenge by developing low cost adaptive mechanisms that enable the provably correct operation of the robot. This will be achieved through algorithm-hardware co-design that allows the adaptation of kernels typically found in robots based on feedback from the underlying hardware. To enable this we will first study the impact of potential hardware failures on such kernels using Machine Learning based failure prediction models. Task scheduling mechanisms will then be developed giving priority to critical tasks, while meeting stringent power and performance requirements. We plan to also exploit the dynamic reduction of arithmetic precision, which we have recently shown can reduce the power consumption but also the number and impact of failures on applications.  The project brings together experts on low-power fault-tolerant systems, control theory and robotics and machine learning for digital manufacturing from Queen’s Electronics, Communications and Information Technology (ECIT) Global-Research-Institute, the Intelligent Autonomous Manufacturing Systems (i-AMS) Pioneer Research Programme  and the Bosch Center for AI in Germany. The student will join an award winning research team in Queen’s ECIT-GRI and the i-AMS centre taking advantage of the unique infrastructure on system design, characterization, and robots while visiting Bosch in a six-month placement at the Center for Artificial Intelligence (BCAI) in Germany providing an international industrial experience. The project is expected to result in several publications in top-tier international conferences and journals venues and the ideas will be disseminated  through international networks of excellence and technical committees of which the supervisors are members and through a workshop at the last year that we aim at organizing in one of the relevant top-tier international conferences.     

Key words/descriptors

Robotics, Industry 4.0, Smart Manufacturing, Control Engineering, Software-Hardware co-Design,  Fault Detection, Approximate Computing, Energy Efficient Systems, Cyber-Physical Systems, Artificial Intelligence

Fit to CITI-GENS theme(s)

• Information Technology: the project aims at combining expertise on control theory, machine learning, low power system design, fault-tolerant and approximate computing, software-hardware co-design for re-designing fundamental robotic algorithms  by taking into consideration the sensitivity of electronics to environmental conditions, and power consumption.

• Advanced Manufacturing: the project goal is to tackle a major challenge in Robotics in digital manufacturing, namely efficient low-power computations in adversarial environments. This is a critical enabler in Industry 4.0, as typically robots are involved in safe-critical and time-critical applications. It is also a prerequisite for robot mobility and autonomy, leading to breakthroughs such as capacity for highly customizable products, and ant-hive type structures in the factory of the future.
• More details about the 3-‘I’ elements
• The project is purely interdisciplinary and intersectoral exploiting concepts and ideas from Systems Theory, Control Engineering, Machine Learning, Computer Science, Robotics to tackle a critical challenge in Advanced Manufacturing, as well as in other applications requiring robot’s deployment in harsh environments.

International Elements:

• Mobility: The PhD project involves (at least) a three-month placement in the Bosch Center for Artificial Intelligence (BCAI), Renningen, Germany.
• Robotics is an enabler for Industry 4.0. A major challenge in manufacturing at the moment is autonomy and safety of robots relying on hardware with a limited computational capacity and on a IIoT infrastructure. i-AMS has input to various such national and international organisations in manufacturing, such as EFFRA (European Factories of the Future Research Association), and the Catapult for High Value Manufacturing. Moreover, members of i-AMS are engaging with prominent researchers from academic institutions in Ireland, France, Belgium, Germany, Italy and USA that are performing research in manufacturing and or robotics, e.g., IIT Advanced Robotics institute. Members of the i-AMS centre can moreover advertise the results of the project via engagement with various non-HEI organisations on manufacturing/robotics, such as the IMR (Irish Manufacturing Research).

• The results of the project will be disseminated through articles in top tier international conferences and journals on design automation and circuits and systems like IEEE Design Automation Conference (DAC), DATE, ICCAD, IEEE Dependable Systems and Networks (DSN), IEEE Trans. on CAD , Trans. on Computer as well as in international control conferences, such as the IEEE Conference on Decision and Control, and top-tier robotics conferences such as the IEEE International Conference on Robotics and Automation. Presentation in such venues will enable the engagement of the Fellow with prominent members of the aforementioned fields.
• Furthermore, the first supervisor is member of the HiPEAC network of excellence which consists of more than 2500 researchers from academia and industry on embedded systems on which robots rely on. The Fellow will become member of this network and he will participate in annual international summer schools covering many topics relevant to the project, will have the opportunity to disseminate his work through a quarterly magazine distributed to all members and will have access to internships and jobs from industrial members.    
• The second supervisor is in the technical committee of the IEEE  Hybrid Systems subcommittee, comprising of more than 60 academics from the best international universities that perform research on similar projects. Recently, the committee extended membership to PhD students, which is something we can avail of to advertise the project and its outcomes in regular meetings.

• The first two supervisors are PIs/cIs of running projects relevant to the project goals. In particular, the first supervisor is the director of the 5M UniServer project looking into revealing and predicting the behaviour of computing systems under various conditions, and coI of the Oprecomp FET-Open project that looks into approximate computing collaborating with ARM, IBM, Ampere, Worldsensing, GreenWaves to whom the ideas will be disseminated. The second supervisor is participating in the CHIST-ERA, EPSRC -funded ( 1.3M) project DRUID-NET on co-design of communication networks and control algorithms for the control of complex cyber-physical systems. The result of the proposed PhD project will be communicated to the DRUID-NET consortium, consisting of five different academic institutes from Canada, Greece, France and Belgium.
• The co-design aim of the project corresponds to an emerging theoretical challenge, namely of formal verification and control of hybrid systems, that is sought by many research teams in US, EU, and elsewhere, e.g., Imperial, Oxford, UCLA, UIUC. The second supervisor is an active member of this community, serving as an Associate Editor in the most important respective journal, namely, "Nonlinear Analysis: Hybrid Systems", Elsevier. There are frequently advertised special issues on these topics, that the outputs of the PhD project could be published.

Supervisor Information

First Supervisor:                     Dr Georgios Karakonstantis                        School: EEECS / ECIT

Second Supervisor:                Dr Nikolaos Athanasopoulos                      School: EEECS / i-AMS

Third Supervisor:                    Dr Vien Ngo                                                   Company: Bosch,  Bosch Center for   

                                                                                                                          Artificial Intelligence, Renningen, Germany

Name of non-HEI partner(s)

Robert Bosch GmbH, or Bosch, is a German multinational engineering and electronics company headquartered in Gerlingen, near Stuttgart, Germany. It is the world's largest supplier of automotive components measured by 2011 revenues. Bosch Cooperate Research: International research network of the corporate sector for research and advance engineering, comprising 12 locations in 8 countries (Palo Alto, Pittsburgh, Boston, Hildesheim, Renningen, St. Petersburg, Moscow, Bangalore, Tel Aviv, Tokyo, Shanghai, and Singapore); other engineering activities related to the development of products and services at some 120 locations worldwide. Bosch Reseach in Renningen where the fouth supervisor is based, Germany. This goal is to encourage interdisciplinary collaboration (automobile, robotic, medicine), and in this way further enhance its innovative strength. At the  center of artificial intelligence in Renningen for research and advance engineering on the outskirts of Stuttgart, some 1,700 creative minds are doing applied industrial research.

Contribution of non-HEI partner(s) to the project:

The third supervisor, Dr Vien Ngo is a research scientist at The Bosch Center for Artificial Intelligence (BCAI). His team's research interest focuses on applying AI and ML theories for robotic manipulation. The team is in cooperation with Cyber Valley, which is Europe’s largest research consortium in the field of artificial intelligence with partners from science and industry. The state of Baden-Württemberg, the Max Planck Society with the Max Planck Institute for Intelligent Systems, the Universities of Stuttgart and Tübingen, as well as Amazon, BMW AG, Daimler AG, IAV GmbH, Dr. Ing. h.c. F. Porsche AG, Robert Bosch GmbH, and ZF Friedrichshafen AG are the founding partners of this initiative. He received the B.S. degree in Computer Engineering from Hanoi University of Science and Technology, Vietnam, in 2005, and the Ph.D degree in Computer Engineering from Kyung Hee University, Republic of Korea, in 2009. He was a postdoctoral researcher at the National University of Singapore from 2009 to 2011, at the Ravensburg-Weingarten University of Applied Sciences in Germany from 2011-2013, and a group leader at the Machine Learning and Robotics lab at the University of Stuttgart from 2013-2017. His research interests include Machine Learning, and Robotics.

Dr Ngo will continuously provide mentoring on the project part that requires formal analysis, and subsequently co-design, of ML and AI algorithms in learning of robots, and host the student in the BCAI labs. Specifically,  he will 1) jointly co-supervise the student during the whole PhD project; 2) Host the student to his research group for a 6 month period in the duration of the project; 3) allow access and testing to specialized Franka industrial robot systems of the lab, used for logistics and manipulations; 4) advice on the use of ML methods for failure prediction and robotic manipulation.

Profile of the non-HEI partner and the nature of the relationship.    

The Bosch Center for Artificial Intelligence was founded in early 2017 to deploy cutting-edge AI technologies across Bosch products and services creating solutions for industry related problems. It hosts 1700 researchers in various interdisciplinary fields including AI-based dynamics modeling, Explainable Deep Learning, Control through Reinforcement Learning and Dynamic multi-agent planning. The student will have the opportunity to gain significant practical knowledge during his placement on industrial research, and he will also be exposed to challenging industry relevant problems.

Research centre / School

 ECIT Institute and i-AMS centre

Subject area

Computer Science, Electrical Engineering