Security and safety of communication in distributed robotics

From Master Projects
Jump to: navigation, search

has title::Security and safety of communication in distributed robotics
status: ongoing
Master: project within::Technical Artificial Intelligence
Student name: student name::Florian Braam
Start start date:=2015/01/01
End end date:=2015/07/01
Supervisor: Guszti Eiben
Company: has company::Network Institute
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:Posternaam.pdf

Signature supervisor



Abstract KIM 1

In the last years robots have become smaller, cheaper and more accessible to the general public. Modern UAVs and self-driving cars are examples of semi-autonomous robots appearing in public spaces. At the same time even simple appliances now have networking capabilities and are using them to improve their performance.

Research in autonomous robot swarms is also benefiting from these technological developments, but barely considers the security implications of low powered wireless communications in autonomous robots. There is not enough knowledge on the influence of a malicious attacker on the behavior of a swarm. Consequentially it is not possible for developers to build swarms that are robust to intrusions. The main question of this project is: How does the security and safety of a networked robot swarm change when central control is decentralized?

In this project we will explore this question using a number of thymio robots, each with a raspberry pi and a bluetooth connection to perform a number of experiments. In each experiment the robots will have to perform a wall-following task with collision avoidance where certain aspects of the task are controlled by a central entity. For each experiment an attacker will attempt to disturb the swarm by jamming connections or by impersonating a robot and injecting misinformation.

We propose three experiments to cover the range from a fully centralized to a fully decentralized swarm: In the first two experiments the master receives the current speed from each follower and transmits an optimal speed back to each follower. The followers will always drive at the advised speed in the first experiment, but will be able to vary their speed in the second experiment to, for example, avoid collisions. The third experiment will be fully decentralized. Each robot will broadcast it's own speed, and calculate it's target speed based on what it receives from the surrounding robots in the swarm. For each experiment we will execute possible jamming, sniffing and traffic injection attacks to evaluate their effectiveness against a centralized, decentralized and hybrid approach.