Skip to the content

Thesis work - Machine Learning (ML) for Embedded Systems

Problem:

ACTIA Nordic AB is a leading designer and manufacturer of vehicle communication equipment. Through the years, ACTIA has delivered communication and diagnostics solution to customers such as Volvo Cars, Jaguar Land Rover, SCANIA, and Volvo Trucks.

  • It’s increasingly common with rental vehicles/machines. This separation of owning and operating a vehicle/machine drives the need for supervision systems for detecting misuse that leads to increased maintenance and operating costs (for the owner).
  • Telematics devices are highly connected devices with both wireless connections to external systems and internet, and also wired connections to communication networks in the vehicle. The highly connected nature of the telematics device means that it’s an interesting target for attackers (i.e. “hackers”), for this reason strong security is needed, one aspect of security is intrusion detection (typically based on anomaly detection).
  • It’s common to have a wide range of vehicles/machines in a fleet, the amount of manual work required for tuning parameters for each vehicle/machine type must be kept at a minimum.

Implementation:

The latest telematics platform developed at ACTIA is ACU6 which is a cross segment platform targeting both automotive and off highway applications. The ACU6 consists of two hardware platforms, a high-end variant called ACU6-Pro which is based on a high-end SoC, and a low-end variant called ACU6-Lite which is based on an MCU. The thesis is targeted at the ACU6-Pro hardware platform, but it’s desirable that some of the results can be transferred to ACU6-Lite in the future.

A thesis project could comprise the following parts:

  • Define data to be collected for training of ML misuse detection and intrusion detection
  • Collect the data needed for training of ML models for misuse detection and intrusion detection
  • Training of ML models for misuse detection and intrusion detection
  • Integrate the ML models in the ACU6-Pro device
  • Evaluate the effectiveness of the ML models

Department: System, Hardware, Software

Time period: Spring 2020

Location: ACTIA Tech-center in Mjärdevi, Linköping. 

Tutor at ACTIA: TBA

Contact: oscar.holm@actia.se