Over the last few decades, our nation has experienced a significant increase in the amount of traffic congestion on freeways and intra-city roads. The current technology designed to control and improve this national problem has met with some real success. Based on inductive loops, that system’s ability to detect the presence of cars and, therefore, the frequency of use of roads has been a great boon. However, the system is not perfect; moreover, based on projections for future use of road infrastructure, the inductive loops solution will not be adequate for long. Continued smooth operations of our road infrastructure require that some more effective solution be developed.

Towards that end, this report proposes a new approach, building on the strengths of its predecessors. This new method utilizes an in-node Machine Learning-based Vehicle Classification (MLVC) system and promises to offer more comprehensive and valuable metrics than has been previously possible. Moreover the system is extremely efficient, boasting attributes such as low power consumption, minimal space requirements, and small associated costs; these features enable the system to technically outperform many existing vehicle-detection systems. Based on its features and performance, this system promises to be an ideal choice for monitoring and alleviating traffic congestion. Used on a wide scale, it could greatly increase the efficiency of our nation’s infrastructure and improve the quality of life of the millions who use our roads and highways.