Honda Research Institute

Applying Graph Data Model and AI to Sensor Data

Helping an automotive research branch to visualize and manage autonomous driving data

0 +
sensor data types

Challenge

Honda Research Institute (HRI) conducts research in advanced driver assist and autonomous driving technology. As part of their research, the institute has created of a large repository of annotated vehicle sensor data from their platform driving sessions—including camera, LIDAR, radar, and GPS—that is used for benchmarks in the evaluation and training of machine learning algorithms. HRI needed to correlate that sensor data across a wide variety of driving conditions and transform the data into visual models so engineers could gain insight into emerging patterns to refine the algorithms. HRI desired the solution to deliver the information through a web interface.

Solution

For the design process, Base22 conducted workshops with teams from EMEA and the US active in managing the driving platforms to define specific requirements and review specifications of all the technical products, protocols, and platforms used on the autonomous driving vehicles. Since the data collected was transformed and stored in Big Data architectures due to its volume and variety, Base22 designed a method to extract the data based on parameters specified by the engineers. Base22 designed and implemented a Graph Data Model to create metadata about the Big Data store. The application that managed the graph data model received parameters from the Base22-designed web-based interface and utilized the metadata model to organize and access the data requested by the engineers. The web-based application provided the ability to search driving session data, select data sets from the driving sessions, and deliver those data sets to the engineers for further analysis. By using the Graph Data model, the metadata layer is able to construct and infer relationships of data elements as they manifest themselves.

Services

Start Typing