In the project, a deep learning model has been developed by Veoneer staff using Imagimob AI SaaS to classify lane changes to the right and left, respectively, with data from existing sensors.
The Pionate data platform was created to act as standalone and can therefore be installed in any car without permanent intervention in the car. The platform is capable of processing vehicle dynamics, position, gaze behavior, ADAS systems on / off and video streams from inside and outside the car. The platform has the ability to identify events and log the above data around these events. Using deep learning techniques, complex events can be detected.
The model has been trained on data previously stored with the data logging platform. The work of developing a model has explored that the sensors are sufficient to classify file changes to the right and left, respectively. The deep learning model was converted to c-code in Imagimob AI and then it was deployed and integrated on the Pionate platform.
The model is run "on the edge", meaning all processing of data is done in real-time on the Pionate platform in the car. No data needs to be sent to the cloud to make the classification. The classification results from the model are saved on the platform together with other data.
Car testing has been performed to verify the integration and evaluate the solution's performance. In summary, the implementation worked well and lane changes could successfully be detected.
Imagimob is a pioneer in Edge-AI and tinyML. The company was founded in Stockholm, Sweden in 2013, and has been working with Edge AI applications ever since. Their SaaS solution, Imagimob AI, is an end-to-end solution that simplifies and makes it possible to develop deep learning applications for tiny devices with resource constraints.
Pionate is specializing in complete end-to-end solutions for remote data logger systems, enabling wireless uploading of large scale data to Pionate’s cloud environment for further analysis and automated data enrichment.