Monitoring and maintenance in robot networks
Supervisor: Dr. Faicel Hnaien and Mitra Fouladirad,
UTT/LOSI-LM2S, UMR CNRS
Email: d
Abstract
Surveillance applications are at the heart of the mobile sensor activities. Often, a static sensors configuration [1] [2] [3] is not able to appropriately cover a spatially-large environment. In this sense, the mobility offers a significant flexibility and adaptability, which must be guaranteed as better as possible. In this PHD thesis we aim to focus on the robots trajectory in order to monitor an environment under reliability constraints. These constraints are multiple such as the autonomy, the unavailability of a robot, ...etc.
Indeed, these robots have a limited autonomy and are subject to preventive maintenance and breakdowns by performing cyclical monitoring tasks. The difficulty is to control the movements of the robot in order to visit all points in the environment optimally taking into account their reliability.
The work will be done in collaboration with a local team of PhD students, Postdoc researchers and a CAPSECplatform.
I. Tasks description
I.1 Optimal trajectory of robots
Such mobile network is equipped with sensing, processing and communication capabilities in order to perform a task. Specifically, mobile agent has only a small footprint over which they can act in two different ways, either to sweep or to sense. Hence, the two primary areas of applications of mobile agent concern: monitoring and sensing environments. The difficulty is to control optimally the movements robots in order to visit all interest points in the environment. Indeed, spend more time in places where the environment changes quickly, without neglecting the places where it changes more slowly. In order to better meet the ultimate purpose of the mobile agent deployment, its movement policy should be established according to: (i) the level of persistent task fields, as well as, (ii) the decision level of trajectory planning. These decisions are distinct from the monitoring studies in the literature because the robot trajectory planning and the coverage time are not given in advance.
I.2 Health indicator modelling of fleet of robots
Ageing is a major deterioration factor of the robots. This latter decreases the robots efficiency and precision and van even lead to failure. The prediction and forecasting of robot aging phenomenon, is an important issue. To properly address this problem all available information on the phenomenon under consideration should be taken into account (sensor, environment, usage, etc.). A large number of sensor information is available at each second or microsecond on the robot trajectory or behavior. To take benefit of this information for the prediction and forecasting proper tools should be developed. The ability to manage, mine, analyze, and visualize the data is essential to the understanding of the phenomenon the data is translating. Nowadays, data intensive computing plays a major role in scientific discovery. Data volume, variety, velocity, and complexity all present challenges that must be faced to efficiently address data analysis at large scale. There is increasing need for new approaches and technologies that can analyze and synthesize very large amounts of data.
We consider large scale sensor data and based on the prior knowledge on the monitored system (the robot), the prediction of an undesirable event, such as a failure, should be carried out.The most important step is to build a health indicator easily tractable which incorporates the maximum information onthe system and as far as possible takes advantage of all available data.
I.3 Maintenance optimisation
Once the health indicator isavailable and its behavior isanalyzed, its evolution in time can bemodeled by a stochastic process. Based on theproposed model and its properties, the failure occurrence can be estimated within acceptable confidence bounds. Toavoid failure and itscostly and dangerous consequences, preventive maintenance operations should be planned. These operations should be planned before failure but not too early in thedeterioration (ageing) process. The optimal planification of these operations is the key idea of maintenance optimization. This latter is possible if the failure time can be predicted properly and the probability of failure can be measuredefficiently.The prediction of the failure time andthe calculation of the probability of failure based on the deterioration model permit us to propose optimal condition-based or predictive maintenance policy.
II Previous work and general layout of the thesis
Usually, for small scale data a stochastic process such as Wiener or Gamma are used to model the evolution of the system, see [4,5]. In the case of large scale data, the model can be represented by a large matrix. In this framework, multi-dimensional health indicators should be proposed and their behavior should be studied, refer to [6].
The different steps of this thesis can be resumed as follows:
- Sensor datacollection and analysis on robots’ behavior
- Health indicator extraction
- Healthindicatormodeling via stochastic models
- Evaluation of failure probability and the failure time
- Maintenance modelling
- Maintenance optimization
III Main Collaboration
The Ph.D subject concerns two departments of UTT. By its scientific requirements, the candidate will participate to the activities of the two departments. The candidate will organize and/or participate to meetings or seminars with the major industrial partners of the UTT on this subject.
IV Reseach team
IV.1 Involved reaserchers
Faicel Hnaien
research interests focus on stochastic optimization, supply planning, scheduling using operation researcher.
Mitra Fouladirad
research interests focus on maintenance modelling and joint maintenance/monitoring policies by using stochastic models to optimise maintenance and/or inspections policies
Contacts:
IV.2 Depatment
The Systems Modelling and Dependability Laboratory (webpage: and Laboratory of Industrial Systems Optimization ( are part of the Charles Delaunay Institute. This institute coordinates all the research activities in the university. The applicant will be involved in the two teams.
International collaborations
If necessary, a research stay in one of these universities can be organised. Moreover, if the quality of the work is correct, any Ph.D student of the
team attends international conferences during the thesis.
VI. References
[1] Rebai, M., Le berre, M., Snoussi, H., Hnaien, F., and Khoukhi, L. (2015). Sensor deployment optimization methods to achieve both coverage and connectivity in wireless sensor networks. Computers & Operations Research, 59 :11 – 21.
[2] Le berre, M., Rebai, M., Hnaien, F., and Snoussi, H. (2015b). A specific heuristic dedicated to a coverage/tracking bi-objective problem for wireless sensor deployment. Wireless Personal Communications, 84 :2187–2213.
[3] Rebai, M., Berre, M. L., Hnaien, F., and Snoussi, H. (2016). Exact bi-objective optimization methods for camera coverage problem in 3-dimensional areas. Sensors Journal, IEEE, 16(9) :3323–3331.
[4] S.L. Smith, M. Schwager, and D. Rus. Persistent robotic tasks: Monitoring and sweeping in changing environments. Robotics, IEEE Transactions on, 28(2):410–426, 2012. ISSN 1552-3098. doi: 10.1109/TRO.2011.2174493.
[5] Le Son, K., Fouladirad, M., Barros, A Remaining useful lifetime estimation and noisy gamma deterioration process Reliability Engineering and System Safety 149 (2016) 76-87.
[6] Nicolas Bousquet, Mitra Fouladirad, Antoine Grall, and Christian Paroissin, Bayesian gamma processes for optimizing condition-based maintenance under uncertainty, Applied Stochastic Models in Business and Industry, 31(3) November 2014.
[7] Xiaochuan Li, Fang Duan, David Mba, Ian Bennett, Multidimensional prognostics for rotating machinery: A review, Advances in Mechanical Engineering, 9 ( 2), 2017.