We have postdoctoral and PhD student positions available for creative and ambitious candidates.


About the group: The Industrial Informatics group led by Zdenek Hanzalek is oriented towards embedded control systems, industrial communication protocols, scheduling and combinatorial optimization.  The group has strong collaboration and personal contacts with several high-tech companies (Porsche Engineering Services, Volkswagen, Siemens, Unicontrols, Air Navigation Services, Merica, Skoda, UNIS, AZD and Rockwell).


Location: Prague is very nice city with affordable accommodation and living expenses.


Postdoctoral Researcher

To apply, email the following materials to zdenek.hanzalek@cvut.cz :

The postdoctoral funding is 2000 Euro per month (including full medical and social insurance). We evaluate candidates on an ongoing basis, until the position is filled, so please submit the materials as soon as they are available.



To apply, email the following materials to zdenek.hanzalek@cvut.cz :

The PhD. student income is 1000 Euro per month (including full medical and social insurance). We evaluate candidates on an ongoing basis, until the position is filled, so please submit the materials as soon as they are available.


Research Topics


Title: Rescheduling algorithms for communication protocols


The growing complexity of distributed real-time embedded systems creates new challenges for scientific practices. These systems are expected to implement more and more complex features (mainly related to reliability and flexibility), while respecting strict non-functional constraints (e.g. response time and energy consumption [ZigBee2010]). Therefore, we are faced with an optimization problem, since these requirements are often in contrast. We assume that adaptive systems that are aware of resources and deadlines can provide an efficient way to cover these problems.

We define the adaptivity as a property of a computing component to adapt itself to (1) both the physical changes (power level, clock speed, communication bandwidth/latency) and the logical changes (network topology, software specifications, protocol requirements), (2) both the changes of application requirements and the variable performance of the resources, (3) both on-line changes, when it is performing its function, and off-line changes, prior to that.


The adaptivity of the schedule is often required by the nature of the application, but we are usually quite reluctant to change them, once they have proven to be correct. Therefore, a new schedule is often required to be similar to the original schedule (e.g. some tasks are at the same position).


Various resource reservation techniques have been integrated into modern communication protocols incorporating time-triggered schedules for high-critical messages (such as Profinet IO IRT, beacon enabled ZigBee, FlexRay, Wireless Hart, TTP, SAFEbus...) .Following these protocols and their future extensions, we consider dynamic reconfiguration of the application at run-time. However, in order to make these technologies really applicable, it is necessary to develop appropriate rescheduling algorithms [Profinet2010]. One of the possible solutions is to extend the RCPS/TC (Resource Constrained Project Scheduling with Temporal Constraints) model so that it can be applied to the dynamic scenario while keeping the same representation of time and resource constraints.


[Profinet2010] Hanzalek, Z. - Burget, P. - Sucha, P.: Profinet IO IRT Message Scheduling with Temporal Constraints. IEEE Transactions on Industrial Informatics, Volume 6, Number 3, Pages 369 - 380, August 2010.


[ZigBee2010] Hanzalek, Z. - Jurčík, P.: Energy efficient scheduling for cluster-tree Wireless Sensor Networks with time-bounded data flows: application to IEEE 802.15.4/ZigBee. IEEE Transactions on Industrial Informatics.  Volume 6, Number 3, Pages 438 - 450, August 2010.


Required skills: good background in scheduling.




Title: Mixed-criticality scheduling algorithms


Many modern applications are of mixed criticality, where safety-critical tasks have to co-exist with less critical ones that are not subject to hard constraints. Recent research in real-time systems has yielded some promising techniques for meeting the two aspects (timing properties and efficiency). The current research in real-time scheduling has centered on an event-triggered approach. However current practice in many safety-critical domains, favors a time-triggered (TT) approach where start times of the tasks are determined by the scheduling algorithms.


Mixed-criticality approach assumes multiple processing time values to be specified for each task depending on the levels of assurance. This simple observation, dealing with tasks executed on processors initiates our idea to use the same mixed-criticality approach while dealing with messages transmitted on TT networks. In this case we deal with non-preemptive scheduling, where the processing time of the message can account for its re-transmission depending on its criticality level. Assuming for example a case with two criticality levels (Hi and Lo), we will obtain two different schedules the first one (verifying Lo criticality level) using optimistic estimates of the processing time of both Lo criticality and Hi criticality messages, and the second one (verifying Hi criticality level) using pessimistic estimate of the processing time of Hi criticality messages only. Both schedules have to be synchronized, so that the system can switch from one to another during its execution, depending on the re-transmissions of Hi criticality messages.


Required skills: good background in scheduling.



Title: Hierarchical planning and scheduling


The main benefit of hierarchical planning and scheduling is that at each decision layer, only the most relevant information is used. E.g., when taking planning decisions, resource capacities are aggregated and the fine details of dealing with single resources are neglected. In contrast, when solving scheduling problems, only the weekly or daily assignments have to be scheduled [Kis2012]. However, these two types of decisions are strongly related and therefore only an efficient communication between the decision layers can guarantee successful method application.


The considered problem is RCPSP (Resource Constrained Project Scheduling Problem) with/without alternatives [Capek2012]. It is characterized by a set of activities to be planned and scheduled. Order of these activities is partly defined by precedence constraints. Each activity requires some amount of the resources while every resource has planned capacity. The goal is to find a sequence of the planned activities and capacity of resources (i.e. a schedule) with respect to two criteria. One objective is to minimize the earliness/tardiness penalty (costs incurred by completing some of the jobs before or after their due-dates). The second objective is to minimize costs connected with increasing capacity of resources (e.g. hiring more employees or working overtime). Then the solution of the problem is a Pareto front considering these two objectives.


For solving the above described problem, the first step will be to design a centralized and exact algorithm. The second step will be to propose a distributed hierarchical algorithm and to compare their performance with respect to the centralized one. In the case of the hierarchical algorithm the objective is to define the problem decomposition into layers and an efficient communication protocol between them. There are two requirements applied on the protocol. On one hand, the protocol has to carry enough information to allow the algorithm to find perspective solutions. On the other hand it should carry the only information needed to minimize the communication bandwidth.


[Kis2012] T. Kis, A. Kovacs, A cutting plane approach for integrated planning and scheduling, Computers & Operations Research, Volume 39, Issue 2, February 2012, Pages 320-327.


[Capek2012] R. Capek, P. Sucha, Z. Hanzalek, Production scheduling with alternative process plans, European Journal of Operational Research, Volume 217, Issue 2, 1 March 2012, Pages 300-311.


Required skills: good background in operational research or combinatorial optimization. Knowledge of scheduling and or distributed algorithms is an advantage.