Projects
Testing of Learning Robots (T-Largo)

About
The future of industrial robotics is rooted in the development of robots that can collaborate and learn with humans. These collaborative robots would have the ability to evolve and improve their behaviours through the usage of machine learning algorithms. However, understanding how to control and test the learning skills of uncaged, single- or multi-arm robots and their ability to safely interact with humans is challenging as their expected improvements is not precisely known. Testing such robots is becoming a crucial research area where the combination of expertise in software testing, machine learning and robotics is strongly required. The ambition of the multi-disciplinary T-LARGO project is to develop a new scientific and technological foundation enabling the testing of learning collaborative robots. Its main objective is the construction of an open test platform dedicated to collaborative robots while its impact lies in major scientific breakthroughs on how to test and control robots equipped with artificial intelligence.
Relevance
Generalizing well-proved software testing methods to learning robots, the T-LARGO project will push forward cutting-edge research in testing and control of advanced industrial collaborative robots. The project will advance the scientific knowledge of how to develop safer collaborative robots with learning capabilities. This is crucial for placing Norway’s robotic ecosystem in the leading position of this research and innovation branch, where some prospective studies say that the market of collaborative robots will increase roughly tenfold between now and 2020.
Aspects relating to the research project
The design of more robust artificial intelligence was recently advocated as one of the most important research directions to follow in the upcoming years. Developing robust artificial intelligence means finding methods that guarantee the safety and dependability of learning processes in various contexts, including robotics. This topic has started to receive considerable attention since the growth of collaborative robotics and machine learning in almost every part of human activities.
However, the rapid development of collaborative robots with learning capabilities is threatened by the lack of open software testing methods able to ensure the safety of these robots. In particular, how to control and test the machine learning algorithms used in robotics is a wide-open scientific question. The T-LARGO project addresses this challenging question through a multi-disciplinary approach combining expertise in software testing, machine learning, and industrial robotics.
Implementation
The T-LARGO project will start in September 2018 and will last for 42 months. The implementation of the project includes 4 tasks, 4 milestones, and 10 technical deliverables. All the technical deliverables will be publically released.
T1: Machine Learning in Robotics
A comprehensive study of machine learning algorithms used in industrial robotics (reinforcement learning, imitation learning, inductive programming, etc.) is required to understand how the T-LARGO project can advance the State-of-the-Art in the area of testing of industrial collaborative robots. This task, crucial to focus the research on the most important techniques used in robotics, will require to digest an abundant literature extracted from three areas, namely software testing, artificial intelligence, and robotics. In parallel, it is important to examine real-world examples of collaborative robots such as YUMI, UR3 and Tiago, as they are among the most advanced collaborative robots. Understanding the design choices which have guided the development of their learning and optimization abilities is crucial to better test them.
T2: Controlling Constraint Acquisition for Robotic Systems
The theoretical background of constraint acquisition framework is particularly well-understood with many available complexity results. This is clearly an opportunity for developing its adoption in collaborative robotics, where safer actions are requested. Extending the framework for learning new constraints (outside of the bias) and restricted forms of the constraints (safe constraints) will be the key challenges addressed in this task. By controlling constraint acquisition in robotics motion control, this task will enable a fine-grain understanding of the testing challenge of machine learning algorithms.
T3: Testing Machine Learning Algorithms
Machine learning algorithms are notoriously hard to test as their results cannot predicted in advance. Even if the inductive nature of these algorithms makes them almost impossible to test accurately, we believe that combining automatic test generation with partial oracle checking in continuous integration is currently the most promising approach to test these algorithms. The challenge of this task is to propose a general-purpose theoretical framework for testing machine learning algorithms and to support it by the development of an open test platform. Carefully selected checking properties will be examined in the project to serve as partial test oracles and their usage for testing constraint acquisition, reinforcement learning, imitation learning and other learning techniques used in robotics will be explored in depth.
T4: Testing Collaborative Robots Real-World Conditions
A key aspect of T-LARGO is to bring real-world industrial robots into the project (YUMI, UR3, Tiago), so that the research advances will be evaluated and demonstrated on actual robots. This process for evaluating the contributions on real collaborative robots will permit the researchers to revise and update their proposition from the feedback obtained during initial evaluation.
Funding source:
Research Council of Norway (RCN), IKT PLUSS

Publications for Testing of Learning Robots (T-Largo)
Poster
T-Largo: Testing of Learning Robots
Simula Research Laboratory - KA23, Oslo, Norway, 2022.Status: Published
T-Largo: Testing of Learning Robots
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Poster |
Year of Publication | 2022 |
Place Published | Simula Research Laboratory - KA23, Oslo, Norway |
Type of Work | Poster for celebrating the 20th anniversary of Simula |
Proceedings, refereed
FoCA: Failure-oriented Class Augmentation for Robust Image Classification
In The 34th IEEE International Conference on Tools with Artificial Intelligence (ICTAI). IEEE, 2022.Status: Published
FoCA: Failure-oriented Class Augmentation for Robust Image Classification
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo), AutoCSP: Self-Supervised Neuro-Symbolic Solvers for Constraint Satisfaction |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | The 34th IEEE International Conference on Tools with Artificial Intelligence (ICTAI) |
Publisher | IEEE |
URL | https://ieeexplore.ieee.org/abstract/document/10098003 |
DOI | 10.1109/ICTAI56018.2022.00144 |
Automated Program Analysis: Revisiting Precondition Inference through Constraint Acquisition
In 31st International Joint Conference on Artificial Intelligence and the 25th European Conference on Artificial Intelligence (IJCAI-ECAI 22). IJCAI, 2022.Status: Published
Automated Program Analysis: Revisiting Precondition Inference through Constraint Acquisition
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | 31st International Joint Conference on Artificial Intelligence and the 25th European Conference on Artificial Intelligence (IJCAI-ECAI 22) |
Publisher | IJCAI |
Testing Deep Learning Models: A First Comparative Study of Multiple Testing Techniques
In Artificial Intelligence in Software Testing @ 2022 IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW). IEEE, 2022.Status: Published
Testing Deep Learning Models: A First Comparative Study of Multiple Testing Techniques
Deep Learning (DL) has revolutionized the capabilities of vision-based systems (VBS) in critical applications such as autonomous driving, robotic surgery, critical infrastructure surveillance, air and maritime traffic control, etc. By analyzing images, voice, videos, or any type of complex signals, DL has considerably increased the situation awareness of these systems. At the same time, while relying more and more on trained DL models, the reliability and robustness of VBS have been challenged and it has become crucial to test thoroughly these models to assess their capabilities and potential errors. To discover faults in DL models, existing software testing methods have been adapted and refined accordingly. In this article, we provide an overview of these software testing methods, namely differential, metamorphic, mutation, and combinatorial testing, as well as adversarial perturbation testing and review some challenges in their deployment for boosting perception systems used in VBS. We also provide a first experimental comparative study on a classical benchmark used in VBS and discuss its results.
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | Artificial Intelligence in Software Testing @ 2022 IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW) |
Publisher | IEEE |
ISBN Number | 978-1-6654-9628-5 |
Other Numbers | arXiv:2202.12139 |
URL | https://ieeexplore.ieee.org/abstract/document/9787976 |
DOI | 10.1109/ICSTW55395.2022.00035 |
GEQCA: Generic Qualitative Constraint Acquisition
In Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 36. AAAI, 2022.Status: Published
GEQCA: Generic Qualitative Constraint Acquisition
Many planning, scheduling or multi-dimensional packing problems involve the design of subtle logical combinations of temporal or spatial constraints. On the one hand, the pre- cise modelling of these constraints, which are formulated in various relation algebras, entails a number of possible logical combinations and requires expertise in constraint-based mod- elling. On the other hand, active constraint acquisition (CA) has been used successfully to support non-experienced users in learning conjunctive constraint networks through the gen- eration of a sequence of queries. In this paper, we propose GEQCA, which stands for Generic Qualitative Constraint Acquisition, an active CA method that learns qualitative con- straints via the concept of qualitative queries. GEQCA com- bines qualitative queries with time-bounded path consistency (PC) and background knowledge propagation to acquire the qualitative constraints of any scheduling or packing prob- lem. We prove soundness, completeness and termination of GEQCA by exploiting the jointly exhaustive and pairwise disjoint property of qualitative calculus and we give an ex- perimental evaluation that shows (i) the efficiency of our ap- proach in learning temporal constraints and, (ii) the use of GEQCA on real scheduling instances.
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo), Department of Validation Intelligence for Autonomous Software Systems, AutoCSP: Self-Supervised Neuro-Symbolic Solvers for Constraint Satisfaction |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | Proceedings of the AAAI Conference on Artificial Intelligence |
Volume | 36 |
Number of Volumes | 4 |
Pagination | 3690-3697 |
Date Published | 06/2022 |
Publisher | AAAI |
URL | https://ojs.aaai.org/index.php/AAAI/article/view/20282 |
DOI | 10.1609/aaai.v36i4.20282 |
Book Chapter
Testing Industrial Robotic Systems: A New Battlefield!
In Software Engineering for Robotics, 109-137. Cham: Springer Nature, 2021.Status: Published
Testing Industrial Robotic Systems: A New Battlefield!
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-LARGO) , Department of Validation Intelligence for Autonomous Software Systems, AI4EU, Testing of Learning Robots (T-Largo) |
Publication Type | Book Chapter |
Year of Publication | 2021 |
Book Title | Software Engineering for Robotics |
Chapter | 5 |
Pagination | 109-137 |
Date Published | 06/2021 |
Publisher | Springer Nature |
Place Published | Cham |
ISBN Number | ISBN 978-3-030-66493-0 |
URL | https://www.springer.com/gp/book/9783030664930#aboutBook |
Journal Article
Predictive Machine Learning of Objective Boundaries for Solving COPs
AI 2, no. 4 (2021): 527-551.Status: Published
Predictive Machine Learning of Objective Boundaries for Solving COPs
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, AI4EU, Testing of Learning Robots (T-Largo) |
Publication Type | Journal Article |
Year of Publication | 2021 |
Journal | AI |
Volume | 2 |
Issue | 4 |
Pagination | 527 - 551 |
Date Published | 12/2021 |
Publisher | MDPI |
Place Published | Basel/Switzerland |
Other Numbers | arXiv:2111.03160 |
URL | https://www.mdpi.com/2673-2688/2/4/33 |
DOI | 10.3390/ai2040033 |
Proceedings, non-refereed
Solve Optimization Problems with Unknown Constraint Networks
In PTHG workshop in CP, 2021.Status: Accepted
Solve Optimization Problems with Unknown Constraint Networks
In most optimization problems, users have a clear understanding of the function to optimize (e.g., minimize the makespan for schedul- ing problems). However, the constraints may be difficult to state and their modelling often requires expertise in Constraint Program- ming. Active constraint acquisition has been successfully used to sup- port non-experienced users in learning constraint networks through the generation of a sequence of queries. In this paper, we propose Learn&Optimize, a method to solve optimization problems with known objective function and unknown constraint network. It uses an active constraint acquisition algorithm which learns the unknown constraints and computes boundaries for the optimal solution during the learning process. As a result, our method allows users to solve op- timization problems without learning the overall constraint network.
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-LARGO) , Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, non-refereed |
Year of Publication | 2021 |
Conference Name | PTHG workshop in CP |
Proceedings, refereed
Constraint Programming for Itemset Mining with Multiple Minimum Supports
In 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence (ICTAI). IEEE, 2021.Status: Published
Constraint Programming for Itemset Mining with Multiple Minimum Supports
The problem of discovering frequent itemsets includ- ing rare ones has received a great deal of attention. The mining process needs to be flexible enough to extract frequent and rare regularities at once. On the other hand, it has recently been shown that constraint programming is a flexible way to tackle data mining tasks. In this paper, we propose a constraint programming approach for mining itemsets with multiple minimum supports. Our approach provides the user with the possibility to express any kind of constraints on the minimum item supports. An experimental analysis shows the practical effectiveness of our approach compared to the state of the art.
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo), Department of Validation Intelligence for Autonomous Software Systems |
Publication Type | Proceedings, refereed |
Year of Publication | 2021 |
Conference Name | 2021 IEEE 33rd International Conference on Tools with Artificial Intelligence (ICTAI) |
Pagination | 598-603 |
Publisher | IEEE |
DOI | 10.1109/ICTAI52525.2021.00095 |
Improving the Reliability of Autonomous Software Systems through Metamorphic Testing
In Proceedings of the 31st European Safety and Reliability Conference (ESREL). ESREL, 2021.Status: Published
Improving the Reliability of Autonomous Software Systems through Metamorphic Testing
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, refereed |
Year of Publication | 2021 |
Conference Name | Proceedings of the 31st European Safety and Reliability Conference (ESREL) |
Pagination | 1-page abstract |
Publisher | ESREL |
ISBN Number | ISBN: 978-981-18-2016-8 |
Publications
Proceedings, refereed
FoCA: Failure-oriented Class Augmentation for Robust Image Classification
In The 34th IEEE International Conference on Tools with Artificial Intelligence (ICTAI). IEEE, 2022.Status: Published
FoCA: Failure-oriented Class Augmentation for Robust Image Classification
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo), AutoCSP: Self-Supervised Neuro-Symbolic Solvers for Constraint Satisfaction |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | The 34th IEEE International Conference on Tools with Artificial Intelligence (ICTAI) |
Publisher | IEEE |
URL | https://ieeexplore.ieee.org/abstract/document/10098003 |
DOI | 10.1109/ICTAI56018.2022.00144 |
Testing Deep Learning Models: A First Comparative Study of Multiple Testing Techniques
In Artificial Intelligence in Software Testing @ 2022 IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW). IEEE, 2022.Status: Published
Testing Deep Learning Models: A First Comparative Study of Multiple Testing Techniques
Deep Learning (DL) has revolutionized the capabilities of vision-based systems (VBS) in critical applications such as autonomous driving, robotic surgery, critical infrastructure surveillance, air and maritime traffic control, etc. By analyzing images, voice, videos, or any type of complex signals, DL has considerably increased the situation awareness of these systems. At the same time, while relying more and more on trained DL models, the reliability and robustness of VBS have been challenged and it has become crucial to test thoroughly these models to assess their capabilities and potential errors. To discover faults in DL models, existing software testing methods have been adapted and refined accordingly. In this article, we provide an overview of these software testing methods, namely differential, metamorphic, mutation, and combinatorial testing, as well as adversarial perturbation testing and review some challenges in their deployment for boosting perception systems used in VBS. We also provide a first experimental comparative study on a classical benchmark used in VBS and discuss its results.
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, refereed |
Year of Publication | 2022 |
Conference Name | Artificial Intelligence in Software Testing @ 2022 IEEE International Conference on Software Testing, Verification and Validation Workshops (ICSTW) |
Publisher | IEEE |
ISBN Number | 978-1-6654-9628-5 |
Other Numbers | arXiv:2202.12139 |
URL | https://ieeexplore.ieee.org/abstract/document/9787976 |
DOI | 10.1109/ICSTW55395.2022.00035 |
Poster
T-Largo: Testing of Learning Robots
Simula Research Laboratory - KA23, Oslo, Norway, 2022.Status: Published
T-Largo: Testing of Learning Robots
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Poster |
Year of Publication | 2022 |
Place Published | Simula Research Laboratory - KA23, Oslo, Norway |
Type of Work | Poster for celebrating the 20th anniversary of Simula |
Proceedings, refereed
Improving the Reliability of Autonomous Software Systems through Metamorphic Testing
In Proceedings of the 31st European Safety and Reliability Conference (ESREL). ESREL, 2021.Status: Published
Improving the Reliability of Autonomous Software Systems through Metamorphic Testing
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, Testing of Learning Robots (T-LARGO) , Testing of Learning Robots (T-Largo) |
Publication Type | Proceedings, refereed |
Year of Publication | 2021 |
Conference Name | Proceedings of the 31st European Safety and Reliability Conference (ESREL) |
Pagination | 1-page abstract |
Publisher | ESREL |
ISBN Number | ISBN: 978-981-18-2016-8 |
Proceedings, refereed
Opening the Software Engineering Toolbox for the Assessment of Trustworthy AI
In 1st International Workshop on New Foundations for Human-Centered AI @ ECAI 2020. CEUR Workshop Proceedings, 2020.Status: Published
Opening the Software Engineering Toolbox for the Assessment of Trustworthy AI
Trustworthiness is a central requirement for the acceptance and success of human-centered artificial intelligence (AI). To deem an AI system as trustworthy, it is crucial to assess its behaviour and characteristics against a gold standard of Trustworthy AI, consisting of guidelines, requirements, or only expectations. While AI systems are highly complex, their implementations are still based on software. The software engineering community has a long-established toolbox for the assessment of software systems, especially in the context of software testing. In this paper, we argue for the application of software engineering and testing practices for the assessment of trustworthy AI. We make the connection between the seven key requirements as defined by the European Commission's AI high-level expert group and established procedures from software engineering and raise questions for future work.
Afilliation | Software Engineering |
Project(s) | Department of Validation Intelligence for Autonomous Software Systems, AI4EU |
Publication Type | Proceedings, refereed |
Year of Publication | 2020 |
Conference Name | 1st International Workshop on New Foundations for Human-Centered AI @ ECAI 2020 |
Pagination | 67-70 |
Date Published | 08/2020 |
Publisher | CEUR Workshop Proceedings |
Other Numbers | arXiv:2007.07768 |
URL | http://ceur-ws.org/Vol-2659/ahuja.pdf |
Proceedings, refereed
Challenges of Testing Machine Learning Based Systems
In Proceedings of the 1st IEEE Artificial Intelligence Testing Conference (AI Test 2019). San Francisco, CA, USA: IEEE, 2019.Status: Published
Challenges of Testing Machine Learning Based Systems
Afilliation | Software Engineering |
Project(s) | The Certus Centre (SFI), Testing of Learning Robots (T-Largo), T3AS |
Publication Type | Proceedings, refereed |
Year of Publication | 2019 |
Conference Name | Proceedings of the 1st IEEE Artificial Intelligence Testing Conference (AI Test 2019) |
Publisher | IEEE |
Place Published | San Francisco, CA, USA |
Technical reports
Survey on Testing of Deep Learning Systems
Simula Research Laboratory, 2019.Status: Published
Survey on Testing of Deep Learning Systems
Recent studies have shown that deep learning algorithms used for image classification or object recognition are not sufficiently reliable. These algorithms can be easily fooled by applying perturbations to images or generating artificial images that result in misclassification. In this paper, we provide an overview of software testing methods present in literature to test deep learning systems. We have explored different methods of testing deep neural networks, namely metamorphic testing, mutation testing, differential testing, and adversarial perturbation testing. We present the main findings available from the literature and compare these methods systematically and comprehensively. The results show that systematic testing of deep learning systems can further help to increase the performance of state-of-the-art systems.
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-Largo) |
Publication Type | Technical reports |
Year of Publication | 2019 |
Publisher | Simula Research Laboratory |
Talks, contributed
Training of Deep Learning Models with Reduced Training Dataset using Regression Testing
In 14th Certus User Partner Workshop (UPW), Larvik, Norway, 2019.Status: Published
Training of Deep Learning Models with Reduced Training Dataset using Regression Testing
Afilliation | Software Engineering |
Project(s) | Testing of Learning Robots (T-Largo) |
Publication Type | Talks, contributed |
Year of Publication | 2019 |
Location of Talk | 14th Certus User Partner Workshop (UPW), Larvik, Norway |