Abstract
This project aims to pioneer a risk-aware control methodology tailored for ego vehicles operating in freeway driving scenarios, such as ramp merging and lane changing. The dynamic nature of these situations, characterized by intricate interactions, demands a meticulous approach due to uncertainties stemming from various sources—human factors, sensor-based stochasticity, and contextual information such as road geometry and vehicle types. For instance, in the context of an ego vehicle performing a ramp merge, it is vital to consider not only the distance between vehicles on the main road but also their speed, intent to yield, and overall traffic observations. This understanding is crucial, as attempting to merge into a gap within a truck platoon might be riskier than with non-truck vehicles. In this highly interactive and uncertain environment, the safety of human drivers heavily relies on Advanced Driver Assistance Systems (ADAS) for accurate risk perception, even when an immediate collision is not imminent.
Our primary objective is to develop a comprehensive risk assessment tool capable of quantifying the diverse risk factors influencing ego vehicles within multi-vehicle interactions. This tool can significantly enhance overall safety, whether a human driver is utilizing active ADAS or opting for full automation by the controller. Seamlessly integrating with existing ADAS, this tool can alert drivers to potential dangers when risk assessment exceeds a certain threshold and can take necessary preventive actions to avert collisions. Building upon our prior research in safety-critical autonomous driving, our goal is to seamlessly integrate the proposed novel risk management toolbox with ADAS/safety controller development, culminating in a distinctive interaction-aware framework for ego vehicles.
We propose the application of Conditional Value at Risk (CVaR) to quantify the cumulative risk that autonomous vehicles face during interactions with surrounding vehicles. This approach accounts for stochasticity arising from human drivers, uncertainty from sensor inputs, and contextual information such as road geometry and vehicle types. CVaR, a robust risk management tool, provides a quantitative measure of potential losses beyond a specified confidence level, helping identify tail risks, compare strategies, and facilitate informed decisions in risk management and analysis. For human-driven vehicles, this risk management tool acts as a verification layer for existing ADAS, mitigating possible unsafe actions by reckless or distracted drivers. For fully autonomous vehicles, this tool can be seamlessly integrated with safety-critical controllers, such as our pre-existing Control Barrier Function (CBF)-based safe controllers, to obtain formally provable safety guarantees. Notably, these safe controllers have demonstrated remarkable performance in initial collision avoidance scenarios, and subsequent efforts will focus on developing risk-aware CBF-based controllers, followed by validation through simulation and real-world vehicle testing.
To validate our approach, we will conduct a rigorous testing phase in simulation and the real world by using a 1/10th scale autonomous race car. The significance of this approach is twofold: first, its implementation in existing ADAS can convey to drivers the comprehensive risk associated with desired actions within the current interactive environment; second, for fully autonomous vehicles, it offers interpretable risk-aware behavior with formally proven safety guarantees, enhancing overall operational safety.
Description
Timeline
Strategic Description / RD&T
This proposal addresses the “Zero Fatalities” Grand Challenge (Ch. 1, p. 3) and two of its desired outcomes (Ch. 2, p. 16): 1) “People no longer accept a high risk of fatality or serious injury as a cost of mobility.” 2) Vehicle...designs incorporate proven, active and passive safety features that protect vehicle occupants and non-occupants.” It aligns with the US DOT RD&T primary purpose of “Promoting safety” (Ch. 1, p. 5) and with the US DOT strategic goals of “Safety” and “Transformation” (Ch. 1, pp. 5-6). Each of these strategic goals is dealt with separately below.
Safety: The proposal addresses two of this goal’s research priorities: Human Factors & Data-Driven System Safety (Ch. 2, p. 17, Table 3). Under the Human-Technology Interactions heading of Human Factors, it “Explore[s] the effects of new technologies, including automation, on travel behaviors" (Ch. 2, p. 18).Under the Safety Technology heading of Data-Driven System Safety, it “Leverage[s] innovative technologies to monitor, predict, and plan ways to reduce injuries and fatalities among the transportation workforce and traveling public" (Ch. 2, p. 19).
Transformation: The proposal addresses one of this goal’s research priorities: New and Novel Technologies (Ch. 2, p. 50, Table 6). Under the Automation heading of this priority, it “Conduct[s] research to develop an effective and efficient safety assessment framework for automated systems across all modes of transportation” and “Develop[s] best practices for safe interaction of automated roadway vehicles with existing vehicles…” (Ch. 2, p. 60).
Deployment Plan
Dec. 31, 2023: Implementation and validation of safe merging/lane change behaviors in simulator
Mar. 31, 2024: Implementation and validation of safe merging/lane change behaviors on a 1/10th scale autonomous race car
June 30: 2024: To the extent possible dependent on PennDOT testbed capabilities, implementation of ADAS driver warnings in unsafe merging/lane change situations
Expected Outcomes/Impacts
Outcomes
• Creation of a risk management toolbox that can be integrated to existing ADAS/safe controllers with significant safety benefits via driver assistance and inclusion in autonomous driving algorithms
• Validation of the system in simulation and on-road driving via a 1/10th scale autonomous race car
• Documentation of the results in a form appropriate for hand-off to a partner capable of full-scale deployment
Impacts/Metrics
• Achieve 98% safe interaction with neighboring drivers in entrance ramp, distance keeping, and lane changing maneuvers under uncertainty
• Achieve risk-aware but efficient driving response to such maneuvers
Expected Outputs
a. Identify the various sources of uncertainty and contextual information that should be considered in risk evaluation.
b. Develop a Conditional Value at Risk (CVaR)-based comprehensive risk evaluation toolbox for interaction-intensive scenarios.
c. Refine the comprehensive risk evaluation toolbox to handle multi-vehicle interactions.
d. Develop risk-aware ADAS/safe controllers by integrating the comprehensive risk evaluation toolbox into ADAS/CBF-based safe controllers.
e. Improve and analyze the risk-aware ADAS/safe controllers’ robustness to sensor inputs and human error.
f. Test the risk-aware ADAS/safe controllers in the freeway driving domain with a focus on entrance ramp behaviors but including also freeway distance keeping and lane changing.
TRID
The main respect in which the proposed project differs from prior work is its: 1) definition of a comprehensive risk-assessment tool that guides vehicle planning and human driver notification; 2) provision of safety guarantees using Control Barrier Functions (CBF). Prior work involves a variety of control approaches that can be tested against traffic scenarios and empirically provide improved safety. Our risk-assessment/CBF approach provides deterministic mathematical guarantees of safety given perfect knowledge of the system model, or probabilistic guarantees given bounded uncertainty on the knowledge of the system model. Since all models are uncertain, bounding this uncertainty remains an active area of research, and part of the project will be to continue to refine human-driver models based on naturalistic driving data. Our lab has done prior work on realistic human-driver models that will be leveraged in this project.
Individuals Involved
| Email |
Name |
Affiliation |
Role |
Position |
| jdolan@andrew.cmu.edu |
Dolan, John |
Carnegie Mellon University |
PI |
Faculty - Tenured |
| yiweilyu@andrew.cmu.edu |
Lyu, Yiwei |
Carnegie Mellon University |
Other |
Student - PhD |
Budget
Amount of UTC Funds Awarded
$50000.00
Total Project Budget (from all funding sources)
$195372.00
Documents
Match Sources
No match sources!
Partners
| Name |
Type |
| PennDOT |
Deployment & Equity Partner Deployment & Equity Partner |
| RISS, Carnegie Mellon University |
Equity Partner Equity Partner |
