Salt Spreader Controller Program Used Machine-Sensed Roadway Weather Data to Reduce Salt Usage by 34 Percent in Massachusetts.
A Pilot Study Was Field-Tested in Massachusetts to Evaluate the Effectiveness of Salt Rate Prediction Model Over Two Other Conventional Methods.
Amherst, Massachusetts, United States
Auburn
Development of a Salt Spreader Controller Program Using Machine-Sensed Roadway Weather Parameters
Summary Information
Effective material spreading is essential for minimizing winter storm impacts on roads while ensuring efficient use of salt, sand, or deicers. Massachusetts DOT (MassDOT) deploys mobile Road Weather Information System (RWIS) sensors to better monitor the road surface and local weather conditions. This study introduced and pilot-tested an intelligent salt application system to optimize material utilization by taking advantage of the mobile RWIS sensors through hardware, software, deep learning algorithm, and prediction model development. Four experiments were conducted using University of Massachusetts and MassDOT vehicles during winter months in 2024 to validate the developed intelligent system.
METHODOLOGY
In this study, hardware development focused on data collection, input/output systems, and power supplies, while software development encompassed video, GPS, and sensor data logging, synchronization, and data fusion. Researchers also developed the Road Surface Classification (RSC) algorithm and the Salt Rate Prediction (SRP) model to optimize salt application decisions for road treatment. Several experiments were designed to validate and refine these components: Experiments 1 and 2 tested hardware and software functionality, Experiment 3 collected 10,000 road surface images to train a machine learning model for RSC evaluation, and Experiment 4 gathered data during salt treatment to assess SRP performance against MassDOT's existing practices. This experiment tested three spread controller modes: auto-grip mode, manual mode, and the proposed SRP-based mode. During heavy snowstorms, operators manually controlled the salt spreader, adjusting salt application based on visual assessment of road and weather conditions to restore traction, while in moderate snowstorms, the system operated in auto-grip mode, automating salt application based on surface grip levels.
FINDINGS
- Experimental results showed that the newly developed and pilot-tested system which integrated the RSC outcome with a comprehensive salt rate decision tree can potentially reduce salt usage by approximately 34 percent and 37 percent, compared with the auto-grip mode and the manual mode, respectively, while maintaining treatment performance.
- The SRP model simulation revealed an approximately 18 percent decrease in salt usage compared to auto-grip mode and performed efficiently, especially in moderate to heavy weather and sleet-mixed snow conditions.
