Improving B-WIM performance with big data and Artificial Intelligence

The increasing number of heavy goods vehicles (HGV), such as lorries, buses, and future road trains and truck platoons, significantly affect road safety and infrastructure management. Overloaded vehicles are especially problematic as they raise security risks and prematurely deteriorate the road infrastructure. Therefore, preventing and efficiently enforcing illegal road use benefits all road infrastructure stakeholders. Overload prevention and enforcement start with traffic monitoring that collects axle loads, gross vehicle weights, vehicle types and frequency, etc. These results are vital for traffic studies and the design and assessments of existing highway structures. The traditional static weighing, which provides the most accurate results, is costly and inefficient for heavy traffic. Thus, weigh-in-motion (WIM) systems are used to collect traffic loading information. Installing most WIM systems is difficult without disrupting the traffic flow and damaging the pavement. The only technology that avoids cutting grooves for sensors into the road surface and does not require roadblocks during the equipment installation and maintenance is bridge weigh-in-motion (B-WIM). Unfortunately, the accuracy and reliability of current B-WIM results do not meet the requirements of legal metrology standards set by OIML (International Organisation of Legal Metrology). As a result, they cannot be used for legal purposes, which limits them to collecting statistical load data and only pre-screening the likely overloaded vehicles for static weighing. The primary source of B-WIM errors is the measurement principle: the axle loads are not measured from contact between the wheels and the sensors but are calculated indirectly from the bridge deflections. The errors rapidly increase with more axles on and the length of the bridge. A B-WIM system can accurately measure gross vehicle weights in such situations but often cannot correctly distribute them to individual axles. An open research question is how to reliably and accurately measure all vehicles’ axial loads on most bridge types. It is accepted within the B-WIM community that conventional research approaches used in the last 30 years have been exhausted and cannot deliver the substantial improvements required for legal enforcement applications. The proposed project solves the listed shortcomings with an alternative approach using big data and advanced Artificial Intelligence (AI) methods. We plan to: 1. Use AI-driven evaluation methods instead of the current heuristic-based B-WIM quality assessment factors to provide reliability scores. Implementing a fully automated quality control (QC) system will significantly expand the usability of B-WIM systems. 2. Design and evaluate a novel AI-driven approach (AI-BWIM) to enhance measurement accuracy. The key to improvement is to determine the number and position of axles of all vehicles on the bridge more reliably. We will achieve this by combining raw strain measurements, heuristic-based evaluation of B-WIM results and traffic camera images with AI methods. The objective is to attain the first OIML-compliant B-WIM system that will fulfil all direct enforcement criteria for at least 60% of all weighed vehicles, a percentage that has not been achieved to date by any other WIM system worldwide. 3. Verify the proposed methods with real measured data from the Slovenian road network and create a public repository of B-WIM data allowing other researchers to validate our findings and improve the B-WIM algorithms. The research community will consequently benefit from the opening of a new research area in the field of B-WIM technology.