Today is a big day. We have some great news from Teraki in Berlin. The ‘Teraki Platform’ has been launched! This Platform automatizes intelligent sensor processing for telematics, video and 3D point cloud. The Platform was developed with a single ideology in the back of our minds: “deliver scalability”. After months of testing we are finally launching the new Platform to the public.
Feedback from developers across the automotive and IoT industries pointed to the increasing need for handling sensor data from high volume of vehicles and devices and the demand to bring this data to the cloud for training for critical AI-models for detection, prediction or decision making. The Platform is introduced to complement the existing Teraki Device SDKs so customers can effortlessly develop embedded applications at scale. The Platform enables customers to bridge - in a fully automated way - high volumes of edge sensors to the cloud. With this ability, customers can efficiently and swiftly develop the best algorithms for any targeted use case.
The platform in its first release includes the following services:
File Service: Supports Sensor Data upload and export. It is the first step before using several of Teraki’s Services such as Telematics Training Service and Decoder Service.
Telematics Service: It includes the Training of Teraki’s intelligent data pre-processing models. It enables inferring data types and signals, and the test of individual Reduction models allowing customers to identify the ones that best serve their needs.
Model Service: Supports querying of Models and Signals. It also enables Developers to compile a given Data Reduction Model together with the latest version of Teraki Telematics encoder for integration in an Embedded application. In this first release linux64 platform will be supported.
Decoder Service: This service is responsible for reconstructing the encoded binary payloads generated by Teraki’s Telematics Encoder.
The Platform offers REST APIs to Developers for easy integration of the above services into 3rd Party applications. It is scalable and can support high volumes of data generated by large number of sensors. This Platform is also behind Teraki’s DevCenter which in its latest release supports Telematics and Video processing and will soon include 3D Point Cloud processing capabilities.
Sample Case Studies:
Case 1: The Data Science team at a major mobility company is interested in a speedy and good evaluation of Teraki’s solutions in efficient edge processing of high volume of telematics data from their next generation car model. They have over 60 types of signals ranging from pressure, voltage, GPS, speed, etc. which they want to evaluate. They are keen on configuring measuring the achieved reduction rates in relation to the accuracy of the reconstructed data. They will select between lossy and lossless processing as well as setting the hard accuarcy bounds per signal - depending on the sensor inaccuarcy and on the use case at hand.
Teraki’s Platform offers REST APIs that can be easily called using any scripting language or SDK. The Data Science team writes a simple Python script to invoke Teraki’s Files Service to upload and ingest the available sensor Data. This script would also call the Telematics Service to infer the types of signals and data types present in the uploaded data and can trigger the Model Training with its various configurations. Once the training is complete, also using the Telematics Service, the trained models can be tested, and data quality KPI reports can be generated.
Case 2: The R&D team is very happy with the high reduction rates achieved by Teraki’s solution as well as the quality of data. The product development team is tasked to integrate Teraki’s offerings in a Series Production integration that involves sensor data from a fleet of hundreds to millions of Electric Vehicles to train an AI model for increased EV range.
Using the Model Service API, the development team can download the data reduction model and the latest encoder library (SDK 3.3) to run on an embedded device. Alternatively, if they have any special hardware requirements, Teraki can provide them with the Intelligent Edge Processing Model(s) that run on the Embedded HW. Sensor data is read from the ECUs and fed into Teraki’s encoder library to generate binary payloads. By simply calling the Teraki Decoder API, signals can be reconstructed and fed to train a Machine Learning model to increase the range of Electric Vehicles.
What is next?
The above Case studies are real world examples of what the Platform can offer to our customers, but this is just the start. While we continue to improve the existing offerings, we will be introducing additional services to cover further use cases and types of sensors. This summer, we will be adding the 3D point cloud service to the Platform for processing 3D point cloud data. Following that, we will be expanding our Model and Training Services to leverage the “power of AI” to further downsize the data required for training Machine Learning models of our customers, saving them millions of miles of collected data. A related service we plan to introduce in the last quarter of 2020 is the Data Service which will provide Customers with the necessary car and drone data required to build AI models.
Teraki Platform is a powerhouse that delivers the exact solution to the industry to cope with exploding edge data volumes. The agile functionality provides flexibility for the customers to easily implement their own AI models. This platform is made to deal with high data volumes. Scalable, automated data ingestions bridge the gap between planning and execution of new AI-models. All available via a simple script and a few lines code.
How to get access?
To get your hands on Teraki’s Platform, simply sign up on DevCenter and request access to the Platform documentation.
While the Platform is hosted on Teraki’s cloud, its infrastructure technology enables easy On-Premise or Customer-Cloud deployments. For more details please do not hesitate to contact us!