Enabling Safe Innovation for Autonomy: Making Publish/Subscribe Really Real-Time

Funded by NSF Cyber-Physical Systems Program.

PI: Jim Anderson. Co-PIs: Ron Alterovitz and Shige Wang (Motional).

The Challenge.

In the automotive industry today, there is fierce pressure to innovate with respect to autonomous features. ROS, or the Robot Operating System, has emerged as a key technology to fuel such innovation. Despite its name, ROS is actually publish/subscribe (pub/sub) middleware that facilitates the modular construction of graph-based robotics applications. This modularity enables rapid innovation through code reuse. Looking ahead, however, ROS looms as a serious liability: for full autonomy to ever become a practical reality, real-time safety certification will be needed to ensure that critical computations complete on time. Here, ROS is fundamentally broken because it lacks sufficient features for supporting verifiable real-time execution. The ROS community has recognized this shortcoming and introduced some weak real-time features to ROS, but the choice of features to support was not guided by research-based evidence.

The Approach.

This project is directed at providing such evidence through multi-faceted research in the CPS Core Research Areas of Real-Time Systems, Safety, Autonomy, and CPS System Architecture. It will contribute to Real-Time Systems and Safety by re-thinking the design of real-time pub/sub middleware for autonomous vehicles. It will contribute to CPS System Architecture by implementing a new pub/sub middleware framework, informed by real-time analysis, and by experimentally evaluating it. It will contribute to Safety and Autonomy by targeting the design of autonomous vehicles that must exhibit certifiably safe and dependable behavior.

Significance.

ROS's success convincingly demonstrates the importance of pub/sub in fueling innovation in autonomy. However, pub/sub must be safe to apply. This project is directed at this very issue, specifically in the context of multicore+acclerator platforms as used in autonomous vehicles. It will address this issue by resolving fundamental resource-allocation concerns at the OS and middleware levels, producing analysis for validating response-time bounds in real-time pub/sub graphs, producing a reference pub/sub middleware implementation, and experimentally comparing this implementation to ROS. While evolving ROS itself is beyond the scope of this project, this project will expose fundamental tradeoffs of relevance to such an evolution.

It is inconceivable how full autonomy can become a common-case reality without stringent certification. A key problem here is excessive reliance on ``black-box'' software (such as ROS) that was originally designed for other purposes (diverse robotics applications, for ROS). The research community must find a way forward that allows such software to be used safely. Otherwise, a certification crisis is inevitable for the automotive industry. This project will tackle a subproblem in this space by bringing real-time safety to pub/sub.


Publications


J. Goh and J. Anderson, " Towards Principled Budget Enforcement in Real-Time Systems", Proceedings of the 45th IEEE Real-Time Systems Symposium, pp. 256-266, December 2024. PDF .


Z. Tong and J. Anderson, " Budgeting Processing Graphs Under Restricted Parallelism", Proceedings of the 14th IEEE International Symposium on Industrial Embedded Systems, pp. 172–181, October 2024 PDF .


S. Ahmed and J. Anderson, " Open Problem Resolved: The `Two’ in Existing Multiprocessor PI-Blocking Bounds is Fundamental", Proceedings of the 36th Euromicro Conference on Real-Time Systems, pp. 11:1–11:21, July 2024. Winner, outstanding paper award. PDF .


S. Ali, Z. Tong, J. Goh, and J. Anderson, " Predictable GPU Sharing in Component-Based Real-Time Systems", Proceedings of the 36th Euromicro Conference on Real-Time Systems, pp. 15:1–15:22, July 2024. PDF .


S. Liu, R. Wagle, J. Anderson, M. Yang, C. Zhang, and Y. Li, " Autonomy Today: Many Delay-Prone Black Boxes", Proceedings of the 36th Euromicro Conference on Real-Time Systems, pp. 12:1–12:27, July 2024. Winner, outstanding paper award. PDF .


J. Bakita and J. Anderson, " Demystifying NVIDIA GPU Internals to Enable Reliable GPU Management", Proceedings of the 30th IEEE Real-Time and Embedded Technology and Applications Symposium, pp. 294-305, May 2024. PDF .


Last modified 18 December 2024