Data-Driven, Automated Digital-Design Workflow for Safety Roads in Korea

HYOUNSEOK MOON: Hello! My name is Hyounseok Moon, I'm the director of the BIM Research Cluster at the Korea Institute of Civil Engineering and Building Technology.

The topic of my presentation today is about Data-Driven, Automated Digital Design Workflow for Safe Roads in Korea. I'll provide an overview of the technology development, and then explain the specific development details.

Actually, in Korea the traffic accidents on roads have been increasing steadily. To address these issues, the government is implementing the hazardous road improvements project on roads where accidents occurred frequently.

However, the traditional approach is based on 2D drawings, and has a limitation in directly incorporating information about traffic accidents, resulting in inefficient linear improvements.

To address this inefficiency, we are developing technology for optimal design and decision making for [INAUDIBLE] road using [INAUDIBLE] technology to determine the priority of hazardous road selections.

Typically in Korea it takes about two months and a cost of approximately $50,000 to improve one hazardous road linearly. The following diagram shows the procedure for feasibility analysis of domestic hazardous road linear improvements projects, and improve the actual results for roads with frequent accidents.

In this technology, we are applying PtD, which is the prevention through design concept technology to digital road design to prioritize hazardous roads and minimize traffic accidents. To achieve this, we have developed a new PtD engineering model, which is called PEM, and applied it to this technology development.

The goal is to evaluate the risk of traffic accidents on hazardous roads using building information modeling technology, propose the optimal road alignment alternatives using statistical techniques, and optimization algorithms, and develop a PtD engineering model to fundamentally eliminate the accident triggers by verifying alternatives in a virtual road environment.

This technology follows the following steps: selection of hazardous roads to construct the PEM model from 2D drawings, evaluation of [INAUDIBLE] risk for these hazardous roads, generation of optimal road alignment alternatives to minimize these risk indices, and decision making.

It also includes the verification of compliance with Korean road design standards, during optimization, not just the creation of a simple optimal road model. People are diving into the detailed technical development. We intend to analyze whether there is a correlation between traffic accidents and geometric factors using the Korean Traffic Accident Analysis System, which is a TAAS system-- which provides access to domestic traffic accident statistics big data.

We analyze approximately 37,128 fatal accidents from 2012 to 2020 as part of this traffic accident statistics big data. Typically, it is known that as the curvature radius of road increases, the number of traffic accidents decreases. However, surprisingly, our analysis showed that accidents are more frequent in roads with a curvature radius of 900 meters to over 1,000 meters.

Similarly, while it is believed that gentle longitudinal slopes result in fewer accidents, our statistics revealed that accidents are more frequent in road sections with almost no longitudinal gradient. This might initially suggest that traffic accidents are not correlated with geometry factors. However, when we differentiate between national and expressways, we find that this phenomenon is more prevalent in highways with large curvature radius or gentle longitudinal slopes.

Furthermore, when we display the location of accidents on a real map, we notice that the many outlier data points are not on actual roads but in areas like mountains, rivers, or the sea. This indicates that these outliers may result from the GPS errors at the location of accident investigations. To address this, we excluded the point with the significant outliers and the verified accident counts based on points in nearby area.

This figure presented shows the weather accident locations with traffic volume observation points. It indicates a lack of collaboration between the actual accident location and traffic volumes. Among the total analyzed 37,128 accidents, we have information about the design speeds for approximately 15,173 cases, with the design speed of 60 kilometers per hour, accounting for over half approximately 8,000 cases.

Generally, one fatal accident can be considered accidental. But two or more may suggest a problem with the road. Consequently, we found approximately 1,227 cases where there were fatalities among which the majority occurred on curved sections.

This figure is based on a sample of 1,227 fatal accidents, showed that most of them occurred on curved section, including curves. These analytical results indicates that the intersections and straight sections, which are excluded from the linear improvement targets, don't play a significant role in accidents. Instead, the remaining cases we removed the intersection and the straight sections, demonstrate that the geometric structures of the road affects traffic accidents.

In summary, accidents are more likely to occur with smaller than curvature radius and steeper longitudinal slopes, confirming that geometry conditions indeed influence the occurrence of the traffic accidents.

In this chapter, we conducted the development of a hazardous road index, which is the fitted index based on an evaluation criteria for determining hazardous roads. Let's dive into the details. This table represents the evaluation criteria provided by the Korean Ministry of Land Infrastructure and Transport of Korea for selecting hazardous road. Among these criteria, the conformity assessment score of geometric elements accounts for 55 out of a total of 100 points, or 55% of ratio, making geometric conditions a crucial factor in prioritizing hazardous road improvements.

Additionally, consideration can be given to factors such as the number of traffic accidents, traffic volume, regional demand, and the project costs. Extracting each geometric piece of information from 2D drawings is quite challenging, which is why creating a BIM model of hazardous road simplifies information extraction.

Furthermore, when using Civil 3D, it becomes possible to easily extract then support the evaluation of design geometric elements. From the previously analyzed accident cases, we can see that there are 1,138 cases, roughly 30% of the total, where the planned curve radius is less than or equal to 140 meters and the longitudinal gradient is greater than or equal to 7% slope.

By utilizing road ID, we can identify 77 road cases where accidents have occurred two or more times. These roads are considered to have a higher risk of accidents and are selected as high priority roads for linear improvements.

This figure provides an example of some of these target area, allowing us to confirm the shape the shape and condition of the roads through satellite imagery and road view.

In this chapter, we have developed a traffic accident risk index based on the evaluation of traffic accident risk. This index considers the geometric condition of roads. That can be used on optimization criteria for road alignment.

To determine how risky these roads are in terms of traffic accidents, we have developed PtD, Prevention through Design index, which serves as a traffic accident risk index. We have utilized the evaluation criteria for hazardous roads provided by the Ministry of Land, Infrastructure, and Transport of Korea, and calculated an overall score considering variables through A to G for each evaluation factor, and their respective weights. These weights are determined through expert evaluations and may vary depending on the shape and characteristics of the road.

In this figure, the overall scores range from 1 to 1.5 or higher. Finer score indicates a higher risk of accidents on the road. While a lower score signifies lower risk, greater safety. Based on this, we conduct our PtD evaluations for the previously analyzed 77 hazardous roads, with three of them being classified as high risk and one as low risk.

This table demonstrates how PtD scores for case A changes when there are improvements in each design evaluation factor, as previously presented. To track the segment-specific risk level, alignment-specific risk level of the road traffic accidents during the BIM design optimization process. We have established a color scheme representing appropriate risk level. Matching these results with four specific cases revealed that three of them were high risk, while one was classified as low risk, low danger.

In this chapter, we present the procedure for creating a digital model of roads with each using existing 2D drawings, digital terrain data, aerial photographs, and road view information. We have developed a simplified road design automation modules using Autodesk Civil 3D and Dynamo at each nodes. And the result of this development are shown.

This figure illustrates the process of creating a detailed design model of hazardous road, performing compliance check with road design standards, and generating optimal alternatives considering PtD scores and the earthwork quantities during the road alignment optimization process. Our research team has developed custom nodes for road design automation, compliance check, and optimization processes.

To achieve this kind of concept, we utilized the Autodesk Civil 3D 2022 and Dynamo modules provided by Civil 3D. This integrated system is expected to become crucial as such cases are not common.

The traditional road design procedure in Civil 3D requires some level of expertise and experiences. However, our research team has simplified the road design procedure performed by Civil 3D into Dynamo, streamlining the process. This allows for rapid automation of road design and easier updates in response to alignment changes.

All data related to road alignment is managed in Excel sheets. And the road design information in Excel is input into the Dynamo module. The road module is then swiftly generated based on the linear and cross-sectional creation algorithms. Our research team can implement this process rapidly using nine Dynamo scripts, as you can see the processes.

The following figure is the Dynamo script for drawing road horizontal alignment and extracting coordinate from the horizontal alignment. The next script creates the polylines based on the extracted coordinates from the created coordinates of the lines.

This figure representing a script for creating road alignment, including transition curves, like crossoid and circular curves based on the polylines and the result of execution. This figure shows the script for generating circular curve, French curves for terminal iterations.

These are screen displays for creating view of the longitudinal curves. This screen presents the script for creating longitudinal curves, including planned grades and in the longitudinal views. This screen display a Dynamo script that allows the modification of longitudinal curves by adding planned grade points, like the PVI point.

With the creation of horizontal alignments and longitudinal alignments completed as described above, a Dynamo script for generating the road's cross-section, as shown in this screen, was developed and applied. Once the horizontal alignment and the longitudinal alignment are completed, a visual check of the automatically generated of corridor can be performed, as shown in this screen.

After the digital road model is generated, earthwork quantity to take off can be calculated using Civil 3D algorithms. In addition, a Dynamo algorithm for automatically placing features, such as street lights and [INAUDIBLE] on the road corridor is provided, like this figure. As you can see, this is the street light.

In this chapter, we have implemented technology that allows us to validate the design criteria for changing road design elements when optimizing the linear aspects of our digital road model. First, you can select the design criteria. I'll leave you validation items, and extract the corresponding design criteria from the Pset database. You can then compare the model with the ruleset to assess compliance with the design criteria.

This screen displays design criteria tables for the geometric structures of road in Korea, which are legally mandated in Korea. We have developed design criteria rulesets using XML or language that computers can understand, like the right side XML scripting.

To facilitate the review, we have developed a Dynamo script that can validate the design criteria for horizontal alignments, longitudinal alignments, and road corridors. When using Dynamo Player, design criteria validation becomes straightforward.

The left side of this screen shows the Dynamo script for exploring flooring design elements from the generated road BIM model, while the right side presents an algorithm for checking whether the BIM model complies with the design criteria. This processing method allows for the validation of design criteria for each design element. This figure demonstrates running Dynamo through the horizontal alignment design criteria nodes and selecting the common property, what kind of project of this road and design criteria review items for the roads.

This enables you to determine whether the design speed for the roads' alignment is appropriate, like this figure. In addition, the full horizontal alignment, you can check the compliance with the criteria for installing the transition curves for the roads, as shown in this figure.

Validation of longitudinal slopes can be done through a custom node for longitudinal alignment design criteria, allowing you to check the conformity of longitudinal alignment installation. As a result, you can confirm the slopes which should be at minus 4 degrees according to the design criteria. It's installed at minus 4.7 degrees, indicating non-compliance with the design criteria.

The algorithm for the custom node for corridor design criteria allows you to check compliance with the design criteria, like this figure. As shown in this figure, you can verify that the minimum lane width for the road should be 3.25 meters, according to the criteria. But it is installed at 3 meters, indicating non-compliance with the design criteria. In this way, you can extract and validate the design element for risk road evaluation for horizontal and longitudinal and corridor elements like that.

Finally, I will introduce the development of a road alignment optimization model to minimize the risk of traffic accident. First, based on the corridor model of the road alignment, we randomly change the PI, a Point of Intersection, of the horizontal alignment to generate the various alternative alignments. The resulting alignments have different earthwork volume and PtD indices.

We then validate the compliance with the design criteria for each alternative alignment. The results for each alignment are recorded in Excel. And the optimal alternative is selected. This data is then reflected back into the model to create the optimal alignment model.

This figure illustrates a high-level algorithm for changing point of intersections to optimize the road alignment. Each PI value changes within certain boundaries, and is recorded for each alternative alignment. This figure shows the source and custom-node for automatic generation of alternative alignments.

Using this kind of logic, and the alternatives roads are recorded in Excel sheets. You can check the changes status in alignment for each alternatives-- one, five, and 10 alternatives, along with their corresponding earthwork volume and PtD indexes. With this information, you can determine the optimal alignment that satisfy PtD index and the earthwork volume and design criteria.

These results are visualized by incorporating the PtD index of visual scheme and into the algorithm, allowing you to visually assess the traffic accident risk and the change status for each segment of the alternatives. The following figure visually shows the variation in PtD for each alternative alignment. As the alignment of each segment changes, the PtD label for that segment also changes, providing a visual basis for determining which alignment to minimize traffic accident.

Consequently, the final alternative is selected. And it is observed that the evaluation PtD of the original alignment improved from 1.65 to 1.61, indicating an expected 6.2% reduction in traffic accident.

You can see the result in the following video, just with no any background voices. Just you can see the videos how the optimization process is being progressed.

[VIDEO PLAYBACK]

[END PLAYBACK]

This shows how to assess the design criteria based on the created 3D model with using the ruleset templates.

[VIDEO PLAYBACK]

[END PLAYBACK]

This video shows how to model, how to extract the optimized alternatives for rows with using this kind of the Dynamo script.

[VIDEO PLAYBACK]

[END PLAYBACK]

And also you can identify the optimized alignment, according to updating the road PtD data, and also volume and design criteria. And also, this video shows how PtD data of each segment for roads can be visualized.

[VIDEO PLAYBACK]

[END PLAYBACK]

When you change the alignment into the optimized another type of the alignment, so you can identify the changed alignment with the changed PtD colors.

[VIDEO PLAYBACK]

[END PLAYBACK]

OK, let me briefly explain the result of this technological development. Recently, these results have been widely covered in domestic newspapers and have been adopted as core technology for national Digital Twin Pilot Project. It is currently being advanced for practical use.

Additionally, it has received recognition in international news reports as a technical achievement, like this figure. And when you click this upper part of the links, then you can identify the technical details in these newspapers.

This is the final conclusion. This technology will be utilized as an advanced model that creates optimal road alignments to minimize traffic accidents, focusing on traffic accident data. It integrates Autodesk Civil 3D and Dynamo and represents a significant advancement. It can be used as a decision-making model for selecting and optimizing risk roads based on a digital model, moving away from the traditional 2D-based approaches.

We have digitalized the entire processes of risk road feasibility analysis process. In the future, with the development of 3D precision road maps, the applicability of this technology is expected to increase. And it can be integrated with Digital Twins to transition to the future's digital design method.

Thank you for your attention so far. Thank you very much.