Collaborative Research: CPS-FR: CSR: MAVERICKS: Mobility-Centric Autonomous System with Full-Stack Software/Hardware Co-Design for Real-Time High-Fidelity Situational Environments

Autonomous driving systems must operate safely in complex and changing environments. These systems rely on perception, planning, and control software to understand their surroundings and make driving decisions in real time. Recent advances in machine learning, computer vision, and embedded computing have greatly improved the capabilities of autonomous vehicles. However, a major challenge remains. These systems must adapt to unexpected events while meeting strict timing requirements. Road conditions, traffic patterns, weather, and pedestrian behavior can change quickly. This may require vehicles to respond immediately. At the same time, critical software tasks such as sensor processing, object detection, path planning, and vehicle control must complete within predictable time limits. These tasks work together in a processing pipeline, where delays in one stage can affect later stages. As a result, delays in processing can reduce system responsiveness and increase safety risks. Both adaptability and predictable timing are essential for the safe autonomous operations. This project establishes autonomous driving platforms that provide both adaptability and verifiable real-time performance. This project develops the Mobility-Centric Autonomous System with Full-Stack Software/Hardware Co-Design for Real-Time High-Fidelity Situational Environments (MAVERICKS). It is a framework. It unifies application software, operating systems, and hardware design. MAVERICKS uses a coordinated co-design approach. Its purpose is to ensure that adaptation and predictability are achieved together. The framework introduces three key innovations. First, it develops a context-aware perception framework. It dynamically adjusts vision processing based on driving conditions. For example, it adjusts to bad weather, dense urban traffic, and highway environments. Second, it creates a modular capsule-based scheduling approach. It manages computing tasks. It moreover provides formal guarantees about safety-critical operations. Third, the project designs customized hardware accelerators. They work closely with the software stack to improve performance, reduce timing variability, and enable efficient resource use. Together, these innovations create a unified approach. It allows autonomous vehicles to respond to changing conditions as well as maintain predictable and verifiable system behavior. The project advances the foundations of real-time autonomous systems. It provides new methods for integrating adaptive software with predictable computing platforms. The project also has broad societal and educational impacts. It enables safer and more dependable transportation systems. Collaboration with industry partners through the NSF eCAT Industry-University Cooperative Research Center helps transfer research results into practice. The project also contributes to workforce development through research training, student mentoring, and outreach activities. It introduces high-school students to computer science, engineering, and autonomous mobility technologies. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.