Department of Mathematics

Proposed Project for the MSU Industrial Math Students

Delphi Automotive PLC

Data Analysis for Identifying Bicycles and Pedestrians Using Radar Detection Signals

About Delphi:

Delphi Automotive PLC (NYSE: DLPH) is a leading global technology company for automotive and commercial vehicle markets delivering solutions that help make vehicles safe, green and connected. Headquartered in Gillingham, U.K., Delphi operates technical centers, manufacturing sites and customer support services in 33 countries.

Background:

Developing products and technologies that make vehicles smart and safe, Delphi uses 76 GHz radars to detect objects in front vehicles. Typically the radar signal will provide the target distance, speed, and relative direction to the vehicle. In real world situation, the detected signal includes noises, effects of target surroundings and target characteristics such as its size, surface, shape, and motion. The hope of the project is to develop a way to classify bicycles and pedestrians from other object by identifying unique signal characteristics from a given data set. In particular, we are interested in the micro-Doppler signature from the bicycles and pedestrians for this study. The micro-Doppler signal is an effect of radar reflections off a partially moving object such as the swing arms of a walking person and spinning wheels of a bicycle.

Project:

The objective of this project is to understand the possibility of identifying unique radar signal signatures to classify pedestrians and bicycles. A real world radar detection data set will be provided to MSU. The data set will include detection signals with and without pedestrians and bicycles. Through data analysis, the students will be asked to identify unique signatures of the pedestrian and bicycle reflections that can be used to classify these objects. Using the identified signatures, the students need to develop a model to test the accuracy of the classification. Independent data set will be provided for the testing of classification accuracy.