Phd Project Internet-of-Things


The technological developments of sensor networks have resulted in the availability of small and cheap sensors that can communicate in an energy-efficient way (e.g., via the LoRaWAN protocol). This makes it possible to attach networked sensors to many objects in the real world and remotely measure many aspects of the environment. This development is sometimes described as the “Internet-Of-Things”. In combination with body area networks and mobile phones, this can be used to interpret and influence human behaviour, focusing for example on the improvement of health and wellbeing, sustainability or safety.

In this project, expertise on sensors and high-performance distributed computing (from the HPDC group of Henri Bal and Thilo Kielmann) will be combined with expertise on advanced modelling and interpretation (from the Behavioural Informatics group of Jan Treur and Michel Klein) to investigate how the human can be made an element in the Internet-of-Things. Specifically, the aim is to develop a generic framework for performing intelligent data interpretation tasks in a sensor network aiming at influencing human behaviours and evaluate this framework in several usage scenario’s, i.e. sustainability, health and safety.

We will take into account the research questions around networking, distributed data collection, data interpretation and computational modelling of human behaviour in the context of real-world scenarios.

Scientific challenges

The central research question in this project is: how can we design and develop frameworks based on wireless sensor networks for effectively supporting specific human behaviours? From this question, more specific challenges follow:

  1. How to collect, communicate, store, abstract, and summarize distributed sensor data?
  2. How to design systems in a robust way such that they can cope with incomplete and outdated data?
  3. How to computationally model social processes (e.g., of social influence) and factors that determine specific health / sustainable / safety behaviours?
  4. How to integrate sensor measurements and online data sources with high-performance parallel simulations (e.g. for parameter tuning) and how to decide where to store which sensor data and where to do the computations.
  5. How to use both data and simulation results for deciding about the actions of computer systems?

Approach and phases

In the first phase of the project, the PhD student will familiarize with sensor networks, human behaviour modelling, and/or parallel programming, depending on the background of the student (AI or Computer Science).
Secondly, a framework will be designed in which computation models can be used for interpreting sensor data and deciding about the actions. This framework will be based on the existing SWAN platform that is developed in Bal’s group by several PhD students (and is already applied within ACBA).
Subsequently, the framework will be applied in 2 to 3 different application domains. These domains will be chosen in order of increasing complexity, such that prototypes can be built upon each other. For each of the application domains, a selection will be made of relevant sensor or online data and computational models will be developed (or adapted from existing models) that are able to reason about this data. After that, a prototype of the sensor system will be developed according to the designed framework. The prototype will be tested in a real world scenario with actual users, for example using field labs.
Eventually, the project will result in a generic framework for performing intelligent data interpretation tasks in a sensor network aiming at influencing human behaviours.


The following domains are considered; the final decision will be made during the course of the project:

  • Sustainability: A change in the behaviour of people is seen as a key factor for energy saving. A relatively simple set of sensors will be used in this domain. Important data sources in this domain are sensors related to appliance use, presence of people in specific rooms, inside temperature, functioning of heating and cooling systems and weather data. Computational models can describe psychological and economical factors that influence behavioural choices and thermo-dynamic aspects of buildings. A sensor based system could reason about optimal usage of energy systems and appliances in combination with steering human choices.
  • Health and wellbeing: In this domain, physiological measurements, sensors related to the location, usage of sporting equipment and weather data are relevant data sources. This requires outdoor sensors via a LoRaWAN network in combination with Body Area Network. Computational models describe determinants of health behaviour and are able to predict the consequence of the combination of the different measurements on the health behaviours; this forms the basis for actions of the system (nudges, reminders, advices) towards the user.
  • Safety: During large events and especially in evacuation scenarios, the challenge is to manage the behaviour of large groups of people, who also influence each other. A large number of sensors is required, for measuring the crowdedness at specific locations, the emotion of people involved and the availability of specific routes. Moreover, complex and distributed reasoning is required to use this data as input to simulation models that predict the effect on the behaviour and the emotion of the crowd. These predictions can be used as basis for real time crowd management advices.