System Architecture for Learning Machines (SALM)
|Coordinator||Luleå tekniska universitet - Institutionen för system- och rymdteknik|
|Funding from Vinnova||SEK 3 500 000|
|Project duration||August 2017 - August 2019|
|Venture||Electronics, software and communication - FFI|
|End-of-project report||2017-01958eng.pdf (pdf, 709 kB)|
Purpose and goal
The aim of the project is to develop and demonstrate solutions that make it easier to develop and test control engineering solutions for tele-remote controlled, semi-autonomous and fully autonomous work machines. The project has successfully demonstrated solutions for automated loading based on imitation and reinforced machine learning. The demonstrations have shown that a neuron network can be trained to imitate an expert operator and that this network can be automatically adapted to other materials with less than one hour of automated loading.
Expected results and effects
The solutions have been demonstrated in-house Volvo CE for competence building and presented with academic publications and a doctoral dissertation. It has contributed to how machine learning can be systematically developed and tested on a full-scale loading machine (i.e. faster and with less risk of unexpected behavior). Wireless communication has been evaluated, which has contributed to Volvo´s investment in a test site for 5G. Adaptive video coding (i.e. SCReAM) has been evaluated in collaboration with Ericsson Research and reported with an academic publication.
Planned approach and implementation
The project has been carried out in close collaboration between Luleå University of Technology (LTU) and Volvo CE in Eskilstuna. The collaboration has included several visits from LTU staff and several longer stays in Eskilstuna for preparation and execution of experiments, tests and demonstrations. This has enabled the project, which has been heavily dependent on access to experimental equipment, full-scale wheel loaders (Volvo L180H) and the ability to instrument and adapt the machine. It has also created good conditions for the exchange of skills and joint competence building.