Research

Projects AntOM and BioMPE

Projects AntOM and BioMPEAntOM aims to promote the robustness, responsiveness of fully distributed P2P overlays at low overhead traffic. The project adopts Ant Colony Optimisation, a bio-inspired method, to discover and optimise nodes in overlay networks. By utilising pheromones and heuristic values, a colony of individual ants, each of whom performs actions based on local information only, would exhibit collective behaviours with interesting properties. The phenomenon enables each peer in a P2P overlay to select its relations with other peers. The macro-effect would be an optimised and adaptive overlay network.

AntOM is part of BioMPE, a project addresses adaptive and reliable communication services based on P2P overlay networks. BioMPE is composed of two parts: network monitoring (AntOM) and network management. The research of BioMPE is led by Dr. Apostolos Malatras and supervised by Prof. Béat Hirsbrunner.

Pervasive Computing Environments

Pervasive means everywhere. The term pervasive computing environment means that computing exists everywhere in our daily life. It includes desktop computers, laptops, tablets, smartphones, etc. It could also includes, in the near future I believe, everyday objects such as our digital cameras, our pens, or even if our key rings. All these devices are networked. However, they have different characteristics such as CPU power, memory capacity, and mobility. In other words, pervasive environments are full of heterogeneity and dynamicity. Moreover, people have diverse requirements to such environments. Someone might want to remotely monitor his home via his mobile phone, while others are sending an 100 page document to the printer. So there is a need to introduce a transition between the complexity of the underlying network and the diverse requirements of applications.

P2P overlays have been widely used in our “computerised life”. The great thing is that we almost don’t need to worry about the physical characteristics of the underlying network because the P2P overlay provided an abstraction of it. It hides the detail of the physical network behind the scene. It other words, from the application point of view, the network has become simpler. So if we can construct a P2P overlay such that on one hand the complexity of pervasive environments can be hidden, on the other hand the overlay can adjust itself to the changes of the environment, then deploying and running applications in pervasive environments would be much easier.

There are many existing solutions of P2P overlays. They all have good performance with regard to robustness, responsiveness and overhead traffic. These are the fundamental measurements for a P2P overlay. However, in highly heterogeneous and dynamic environments, it becomes much more challenge to satisfy these requirements. Therefore it comprise the goal of our AntOM project.

Ant Colony Optimisation

In the nature, there are a group of so called social insects. They find food in a collaborative way. If one of them found a food source, it will deposit a special chemical called pheromone on on its way back to the nest. The amount of the depostied pheromone is proportional to either the length of the path or the quality of the food. As long as the food is still there, the pheromone will be constantly reinforced again and again by ants. If there are multiple food sources, as you might have awared, the path to a better food source will receive a stronger reinforcement each time an ant passes. The concentration of pheromone then becomes an indicator of the attraction of the path because stronger pheromone means shorter path / better food. In the same time, the pheromone also constantly evaporates. If a food source is depleted, its pheromone will eventually completely disappear, which causes the path being unattractive at all.

The theory of Ant Colony Optimisation was developed based on the above mentioned behaviours of ants. Each ant starts its journey from a node. It collects some local information. Then it chooses its next stop from the neighbours of current node. After it has arrived a predefined destination (food source), it begins its backward journey and reinforces pheromones along its way home. The increment of pheromone is related to the quality of links / nodes.

In the same time, each node periodically reduce the amount of pheromones to simulate the evaporation process. A consequence of this procedure is that pheromone concentration becomes a combination of quality and validity. An example could be that, a perfect neighbour whose information was updated 3 minutes ago has the same pheromone concentration with a moderate neighbour whose information was updated just before 1 minute. This feature is important in pervasive environments because the network condition changes all the time. So validity of information must be considered.

 

Conferences

Peng, F., A. Malatras, M. Courant, and B. Hirsbrunner, “AntOM: Constructing Multi-layer Overlays for Pervasive Environments”, 8th International Workshop on Mobile Peer-to-Peer Computing (MP2P 2012) (accepted), Lugano, Switzerland, IEEE, 03/2012.

Malatras, Apostolos, Peng, Fei and Hirsbrunner, Béat, A Self-Management Framework for Efficient Resource Discovery in Pervasive Environments, in: 8th International Conference on Autonomic Computing, pages 1-2, ACM, Karlsruhe, Germany, June, 2011. {Accepted}

Malatras, Apostolos, Peng, Fei and Hirsbrunner, Béat, Bio-inspired monitoring of pervasive environments, in: 9th IEEE International Conference on Pervasive Computing and Communications, Work In Progress, PerCom 2011, IEEE, PerCom, Seattle, USA, March, 2011.

Peng, Fei and Malatras, Apostolos, Optimising P2P Overlays for Pervasive Environments, in: AIMS ’11 Proceedings of the 5rd International Conference on Autonomous Infrastructure, Management and Security (PhD Workshop), Springer, AIMS 2011, June, 2011. {Accepted}

 

Technical reports

Malatras, Apostolos, Peng, Fei, and Hirsbrunner, Béat, Exploiting bio-inspired approaches for the monitoring of pervasive environments, March, 2011.

Peng, Fei, Malatras, Apostolos, Courant, Michèle and Hirsbrunner, Béat, Topology Management in Heterogeneous P2P Overlay Networks, November, 2009.

I am / was the teaching assistant for the following courses:

Operating Systems

  • Linux (Ubuntu 10.04)
  • Minix 3 with two projects per semester
    • Modifying the scheduling policy.
    • Implementing a new system call (this post).

Programming Paradigms

  • All major programming paradigms with an emphasis on logic programming.
  • project per semester with Prolog.
Robotic Programming
  • The design and development of robot programs based on LEGO NXT.
Bachelor Thesis Co-supervision
  • Brice Courtet – Shortest Path Discovery in Static and Dynamic Graphs using Ant Colony Optimization.
  • Pat Kläy - Bio-inspired Information Broadcasting in P2P Overlays.
  • Georg Schaller - Bio-inspired On Demand Optimal Path Searching.

Comments & Responses

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>