Abstract
Bridge collapses, road closures, disruptions in the public transportation system, and major issues caused by autonomous vehicles (AVs) are everyday realities of our transportation infrastructure that not only cause inconvenience to the public but also constitute a major safety concern. When a particular component of the transportation system fails (e.g., due to an AV blocking a road), the failures and the associated congestion will likely be propagated to other parts of the transportation system, which may lead to further failures, and so on, potentially leading to a cascade of failures and a catastrophe in the whole city. A real-world example of this phenomenon took place on July 21, 2012, when a heavy rain shut down a metro line in Beijing and caused 100 bus routes to detour, skip stops, or cancel operation completely. Similarly, increasing deployment of AVs in the form of robotaxis by companies like Waymo and Cruise have not only led to several accidents but also events where seemingly confused AVs blocked certain roads for several hours. Cities such as Pittsburgh are particularly vulnerable to such cascade of failures and congestion propagation due to harsh weather conditions and existence of many bridges/tunnels creating bottlenecks. Given also the fact that increased congestion levels will likely lead to an increase in traffic incidents, there is a clear need for a better understanding of the impact of these cascading failures on the safety of the transportation system and the role that AVs play, both positive and negative, in them.
This project aims to study the cascading effects of transportation network failures with an eye towards developing mitigation policies that maximize overall public safety. We will be particularly interested in accounting for the increased presence of AVs, both to understand their impact on initiating or amplifying these failures, and to reveal how AVs can help mitigate cascading failures. For example, our prior project supported by Mobility 21/Big Ideas fund laid out the initial work demonstrating how AVs can help reduce congestion more effectively by their ability to react in real time to vehicles around them, and their ability to be remotely and centrally controlled by fleet owners. Building on these initial results where the goal was to minimize the overall delay/congestion, this project will seek to reveal the impact of AVs on the safety of the overall transportation system.
Our plan is to develop a comprehensive model that quantifies the safety impact of different failure events while taking into account the potential cascading effects. For example, a stalled robotaxi blocking an intersection in San Francisco would initially pose a safety threat to vehicles and pedestrians in its vicinity. In addition, depending on how long it blocks the road, this event may cause a congestion which can then cascade to neighboring roads, potentially leading to increased accident rates in the entire city. To the best of our knowledge, this project will be developing the first set of metrics for quantifying the safety impacts of these failures with their cascading effects also included.
We would like to add that this project is synergistic with our concurrently submitted proposal entitled “Evaluating Autonomous Vehicles’ Safety Benefits in Mixed Autonomy Scenarios,” where the goal is to evaluate the safety impact of AVs from the perspective of their accident rates with other vehicles and human pedestrians. The current project on other hand focuses on revealing the overall safety impact of AVs including their impact on congestion and cascading road failures. As such, the two projects will nicely complement each other can be combined together at a total budget of $150,000 if so preferred.
Description
Timeline
Strategic Description / RD&T
This project addresses several goals and priorities laid out in the US DoT Research, Development, and Technology Strategic Plan. Our project addresses two of the “primary purposes” mentioned on page 5 of the plan, namely “reducing congestion” and “promoting safety.” Our project is also directly aligned with two of the “strategi goals” mentioned on page 5, Safety and Transformation. By quantifying the safety impacts of (cascading) road failures and developing mitigation strategies to improve public safety, our work will “[m]ake our transportation system safer for all people. Also, by investigating impact of autonomous vehicles in both contributing to and mitigating these failures, our project will help “meet the challenge of the present and modernize a transportation system of the future that serves everyone today and in decades to come.”
Our project will also help realize the vision and desired outcomes of the “Safety Grand Challenge” outlined on page 16 of the plan by developing congestion mitigation strategies to “reduce the risk of severe crashes by managing speeds [and] reducing conflicts.” Our approach to these ends also matches the critical research topics outlined on page 16 of the plan, particularly to “safety risk analysis,” and “vehicle safety and automation.” In terms of the Safety Research Priorities and Objectives summarized in Table 3 (page 17), our project falls perfectly under “Data-Driven System Safety” and addresses several research objectives including “Safe Technology” (by “[a]dvanc[ing] transportation safety by evaluating the safety of existing transportation technologies and supporting the safe integration of emerging technologies) and “safe design” (focusing on “evaluating the safety performance of infrastructure design and developing and promoting the use of effective safety counter-measures”).
Deployment Plan
· Q1: In the first quarter of the project, we will collect data that might be useful in modeling the dependencies across different road components so that we can accurately model the cascading effects of these failures. In the first quarter, we also plan to build the metric for quantifying the safety impact of these (cascading) failures. This will go through understanding i) how congestion in a particular road affects road safety; ii) how a cascade congestions over a short period of time in a particular region might pose additional safety risks (with more frustrated drivers driving in unfamiliar roads); and iii) what additional safety risks are involved when AVs are involved (e.g., due to AVs initiating the failures in the first place or their inability to react “properly” to the instructions by the traffic authorities). We anticipate that most of the required data will be publicly available (e.g., bus routes), and we will work with our deployment partner, the Southwestern Pennsylvania Commission (SPC), to help identify additional data sources.
· Q2: In the next phase, we will develop a comprehensive analytical model for the transportation infrastructure reliability and propagation of congestion in the presence of AVs.
· Q3: Next, we will conduct simulations, mathematical analysis, and optimization on the congestion model developed, and evaluate different scenarios in terms of safety impact, e.g., extreme congestion, roadwork, a robotaxi blocking an intersection, etc.
· Q4: In the final phase of the project, we aim to convert our finding to a policy report focusing on the overall safety impact of AVs from the perspective of cascading road failures. On one hand, this report will shed a light on the current trade-offs involved in large-scale deployment of AVs (e.g., in the form of robo-taxis). On the other hand, it will provide policy guidelines and suggestions for mitigating cascading road failures and improving the overall safety impact of AVs.
Expected Outcomes/Impacts
This work will lead to a new set of metrics for evaluating the overall safety impact of autonomous vehicles that will take into consideration the cascading road failures that AVs might initiate and/or contribute to. These metrics will enable a new perspective on the safety impacts of AVs and can help develop new policy decisions and regulations to control their increased deployment in a manner that puts public safety at the forefront. We will also build on our previous “Big Ideas” Mobility 21 project, which revealed that AVs can help reduce congestion propagation if they can be routed intelligently. Combining these insights with the newly developed safety metrics, we will deliver a comprehensive evaluation of the current safety impacts of the AVs. More importantly, we will deliver guidelines for changes in existing practices and policies that can help improve the safety impact of AVs.
As our deployment and equity partner, The Southwestern Pennsylvania Commission (SPC) will aid us in identifying the policy implications of our findings as well as in identifying data sources that lay the foundation of our work. In addition, we will leverage SPC’s existing public participation processes to receive community feedback on our work and ensure that the safety metrics we develop take the safety of vulnerable populations particularly into account. We will also leverage SPC's input to help ensure that the policy guidelines we are developing are meaningful from an equity point of view as well.
Expected Outputs
Our expected deliverables include:
i) a comprehensive and realistic data-driven model of how congestion propagates through dif- ferent parts of the transportation network;
ii) a new metric for quantifying the safety impact of cascading road failures which will be customized for autonomous vehicles
iii) a comprehensive evaluation of the safety impact of current AV deployments and guidelines for changing existing policy and practices to improve their safety impact.
TRID
We conducted a TRID search with the keyword “Cascading Failures for Safety” which resulted in 28 results. Most of these results focused on analyzing the cascading failures in different components of the transportation system (e.g., railways-bus, public-transportation, railway-power interdependent network), and do not address safety impacts of these failures. Our search for the key word “autonomous vehicles and cascading failures” did not return any results. Finally our search for the key word “autonomous vehicles congestion propagation” produced 17 results. Most of these works focus on routing and intelligent control of AVs which is partially related to our goals but none of them focus on safety or address cascading failures. We plan to build on the findings of these prior works on congestion propagation in the era of AVs by considering the cascade of congestion-induced failures and by an overall evaluation from the safety point of view.
Individuals Involved
| Email |
Name |
Affiliation |
Role |
Position |
| cjoewong@andrew.cmu.edu |
Joe-Wong, Carlee |
Carnegie Mellon University |
Co-PI |
Faculty - Untenured, Tenure Track |
| oyagan@ece.cmu.edu |
Yagan, Osman |
ECE |
PI |
Faculty - Untenured, Tenure Track |
Budget
Amount of UTC Funds Awarded
$75000.00
Total Project Budget (from all funding sources)
$150000.00
Documents
Match Sources
No match sources!
Partners
| Name |
Type |
| Southwestern Pennsylvania Commission |
Deployment & Equity Partner Deployment & Equity Partner |
