Case Study: PathPartner achieves 99.97% radar-based classification and reduces development time from months to days with MathWorks (MATLAB and Statistics and Machine Learning Toolbox)

PathPartner Develops Machine Learning Algorithms for Radar-Based Automotive Applications

PathPartner, a provider of radar algorithm packages for automotive applications, needed a fast, accurate classifier to detect pedestrians and other vulnerable road users (VRUs) in challenging conditions. Their early prototype ran too slowly (taking 5–8 seconds to detect a human) and required embedded implementation and real-world verification. To accelerate development and improve performance, PathPartner used MathWorks tools—specifically MATLAB and the Statistics and Machine Learning Toolbox.

Using MathWorks software, the team rapidly iterated on design, increasing the frame rate from 3 to 5 fps and creating moving-average features, then deploying and testing the classifier on an embedded platform. The result: design changes that once took months now take days (classifier tweaks in about an hour), algorithms can be implemented and evaluated in minutes, and classification accuracy improved from 97% to 99.97%, producing an automotive-ready classifier. MathWorks enabled the accelerated development and measurable performance gains.

PathPartner