During the last decade, a combination of advancements in machine learning and the proliferation of low-cost hardware accelerators have resulted in massive progress in the domain of autonomous vehicles. While there are currently no commercially available vehicles that meet the SAE’s definition of level 5 autonomy [1], autonomous driver assistance system (ADAS) features are now available and required by law (c.f. automatic emergency braking [13]) in commercial off-the-shelf designs. Such features are usually incorporated into existing vehicle architectures to minimize program risk, however, failure to achieve level 5 autonomy with existing algorithms and hardware [14] indicates that radically different hardware and software architectures may be required to develop such a vehicle.

Automotive engineering is a complex discipline, which requires research inputs from many different fields. The goal of this meeting is to bring researchers from different communities together to discuss and generate future research directions that could eventually lead to the safe and cost effective development of fully autonomous vehicles or improve the performance of existing semi-autonomous vehicles. Currently, there are many novel technologies being researched, but the trend is to treat these technologies in isolation. For instance, spike neural networks show promise [4, 20, 27], but may require complementary system-on-chip (SoC) architectures for efficient execution [21] and novel vehicle-to-infrastructure data sharing schemes [19] to realize their full potential. Likewise, addressing the limitations of current vehicle architectures requires investigation of complex workload models [12, 22] alongside statistical methods for execution time analysis [23] to enable provable real-time guarantees for complex, uncertain workloads on heterogeneous multi-core platforms. Furthermore, development of end-to-end latency assurance techniques across interconnected autonomous driving nodes is critical to ensure scalable and predictable performance as system complexity increases. Such simultaneous innovations can only take place when researchers from machine learning (ML), hardware, real-time, and networking backgrounds have a mutual understanding of the challenges involved.

We are proposing a 4 day workshop to facilitate communication between disciplines with the aim of addressing challenges such as those mentioned above. The first day will be dedicated to a review of state-of-the-art (SOTA) in ML algorithms and hardware/software architectures for autonomous driving and connected vehicles. Subsequent sessions will be focused on the following themes:

2.1 Theme T1: New Hardware/Software Architectures for ML and Co-Design Opportunities

Most autonomous driving research is centered around existing hardware and software architectures, e.g., commercially available SoCs, centralized engine control unit (ECU) network topologies, and robotics middleware like the Robot Operating System (ROS) [17]. However, it is unclear whether such hardware/software architectures are optimal for autonomous driving, especially when used in conjunction with each other and with ML-based applications. This theme explores co-designs that jointly consider the application (ML), the base hardware architecture and the middleware, with the objective to improve overall performance.

Key Questions to Address:
1. SoC Co-Design: Are there SoC topologies that better serve the autonomous driving task than currently available hardware architectures given SOTA ML models? [26, 6]
2. ECU Topology Co-Design: How can future intra-vehicle networks and ECUs meet the need for robustness while balancing cost, energy consumption and heat generation? [3, 9]
3. Software Co-Design: What middleware-level software advancements (communication, scheduling, and execution) would be required to take advantage of such hardware architectures? [24]

2.2 Theme T2: New Hardware/Software Architectures for Connectivity and Edge Computing

Even though autonomous vehicles should be capable of safe operation in the absence of network activity, it is important to consider vehicles as part of a larger transportation system. Future designs should not be limited to the hardware and software running on a single vehicle [2], however, taking advantage of such distributed hardware/software architectures will require careful planning with regards to resource allocation and system design. For example, additional processing power could be provided by the cloud or neighboring nodes in an interconnected network [7], and sensor data could be shared between infrastructure and vehicles to improve safety, performance, and possibly reduce energy consumption [5].

Key Questions to Address:
1. Underlying Infrastructure: What road-side infrastructure advancements are required to achieve fully autonomous driving? [5]
2. Reliability: Given the potential for a lack of network connectivity, how can we ensure that a connected vehicle still performs safely when relying solely on its own sensor data? [18]
3. Middleware for Edge Computing: Are any novel applications opened up by connected autonomous vehicles? Can the information shared between vehicles and infrastructure improve transportation system efficiency? [15]

2.3 Theme T3: Safety for Future Vehicular Hardware/Software Architectures

While the previous themes focus on innovation in on-vehicle and vehicle-to-everything (V2X) hardware/software architectures, this theme scrutinizes the safety of such proposals. Specifically, it focuses on how aspects of these designs could lead to violations of functional or non-functional requirements that could lead to safety violations.

Key Questions to Address:
• Safety of Hardware/Software Co-Design: How does co-design impact vehicle safety. What approaches and metrics will be needed to evaluate safety in future vehicular architectures? [16]
• Safety of V2X Middleware and Applications: How does V2X impact safety of vehicles and their surroundings and how can future design methods account for this? [10]
• Real-time considerations and their impact on safe vehicle operations: What real-time considerations in terms of statistical guarantees, scheduling techniques, and end-to-end latency calculations are required to guarantee safety? [11, 23, 25, 8]

The proposed series of workshops are expected to have the following outcomes:
1. An agenda for future systems’ research with respect to autonomous and semi-autonomous vehicles. The agenda should cover problems across all levels of the hardware/software architecture (SoC, intra-vehicular design and connected vehicles).
2. Interdisciplinary and joint project proposals: These proposals should combine expertise from researchers working in a variety of areas with applications in autonomous driving.
3. Collection of abstracts and presentation artifacts used by the speakers during the workshops.

Difference to other Seminars

• #208: Trustworthy Machine Learning System Engineering Techniques for Practical Applications (2024-10)
The focus of this seminar is on ML system engineering in general safety-critical domains, not specifically tailored to vehicle architectures or real-time constraints which is the focus of this proposal.
• #204: DevOps for Cyber-physical Systems (2023-11)
This seminar discusses the application of DevOps to cyber-physical systems’ lifecycle management, whereas our proposal is centered on system architecture innovations rather than development processes.
• #178: Formal Methods for Trustworthy AI-based Autonomous Systems (2023-10)
While this seminar focuses on formal methods and safety assurance for AI in autonomous systems, our proposal emphasizes cross-layer architectural co-design, including hardware, networks, and real-time scheduling.
• #118: Modelling and Analysing Resilient Cyber-Physical Systems (2018-12)
Although this seminar deals with resilience in cyber-physical systems, it does not specifically address the hardware/software co-design and vehicle-to-infrastructure communication aspects that are core to our proposal.
• #114: Resilient Machine-to-Machine Communication (2018-03)
This seminar addresses machine-to-machine communications in general Internet-of-Things contexts, whereas our proposal explores their role specifically in the context of connected autonomous vehicles.
• #104: Software Engineering and Networked Control for Smart Cyber Physical Systems (2017-08)
The main theme of this seminar is coordination between software and control systems in cyber-physical systems; our proposal goes beyond this to include edge/cloud collaboration and many-core scheduling for autonomy.
• #076: Validated Numerics Meets Reachability Analysis for CPS Design (2015-09)
This seminar is methodologically oriented towards formal analysis techniques; our proposal instead concentrates on architectural challenges in autonomous driving systems.
• #073: Architecture-Centric Modeling, Analysis, and Verification of Cyber-Physical Systems (2016-03)
The focus here is on formal modeling and verification at the architecture level, while our proposal also tackles runtime aspects, system scalability, and deployment scenarios.
• #055: Science and Practice of Engineering Trustworthy Cyber-Physical Systems (2014-10)
This seminar covers trustworthy cyber-physical systems’ design at a conceptual level, while our proposal is more practically focused on integrated system components for autonomous vehicles.

Major Events/Activities by the Organization Team in the Past

• Arvind Easwaran
– Program Chair – IEEE Real-Time Systems Symposium (RTSS) 2022.
– Program co-Chair – ACM International Conference on Cyber-Physical Systems (ICCPS) 2025.
– General co-Chair – IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA) 2020.
– Steering Committee - IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA), since 2021.
• Takuya Azumi
– Artifact Evaluation Chair – IEEE Real-Time Systems Symposium (RTSS) 2024.
– Program Co-Chair – International Conference on Intelligent Software Methodologies, Tools, and Techniques (SOMET 2022)
– Program Chair – Summer Workshop on Embedded System Technologies (SWEST 15, 16, 17) .
• Georgios Fainekos
– Area Chair: ACM/IEEE Conference on Cyber-Physical Systems (ICCPS); part of CPSWeek 2025
– Awards Chair: Hybrid Systems: Computation & Control (HSCC) 2023
– Program co-Chair: AAAI 2024 Spring Symposium on User-Aligned Assessment of Adaptive AI Systems

[1] Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles. Standard J3016 202104, SAE International, Warrendale, PA, USA, April 2021.

[2] Peter Arthurs, Lee Gillam, Paul Krause, Ning Wang, Kaushik Halder, and Alexandros Mouzakitis. A taxonomy and survey of edge cloud computing for intelligent transportation systems and connected vehicles. IEEE Transactions on Intelligent Transportation Systems, 23(7):6206–6221, 2022.

[3] Victor Bandur, Gehan Selim, Vera Pantelic, and Mark Lawford. Making the case for centralized automotive e/e architectures. IEEE Transactions on Vehicular Technology, 70(2):1230–1245, 2021.

[4] Zhenshan Bing, Claus Meschede, Kai Huang, Guang Chen, Florian Rohrbein, Mahmoud Akl, and Alois Knoll. End to end learning of spiking neural network based on r-stdp for a lane keeping vehicle. In 2018 IEEE international conference on robotics and automation (ICRA), pages 4725–4732. IEEE, 2018.

[5] Christian Creß, Zhenshan Bing, and Alois C. Knoll. Intelligent transportation systems using roadside infrastructure: A literature survey. IEEE Transactions on Intelligent Transportation Systems, 25(7):6309–6327, 2024.

[6] Pudi Dhilleswararao, Srinivas Boppu, M. Sabarimalai Manikandan, and Linga Reddy Cenkeramaddi. Efficient hardware architectures for accelerating deep neural networks: Survey. IEEE Access, 10:131788– 131828, 2022.

[7] Sijing Duan, Dan Wang, Ju Ren, Feng Lyu, Ye Zhang, Huaqing Wu, and Xuemin Shen. Distributed artificial intelligence empowered by end-edge-cloud computing: A survey. IEEE Communications Surveys & Tutorials, 25(1):591–624, 2023.

[8] Chuanchao Gao, Niraj Kumar, and Arvind Easwaran. Energy-efficient real-time job mapping and resource management in mobile-edge computing. In 2024 IEEE Real-Time Systems Symposium (RTSS), pages 15–28, 2024.

[9] Georg Georgakos, Ulf Schlichtmann, Reinhard Schneider, and Samarjit Chakraborty. Reliability challenges for electric vehicles: from devices to architecture and systems software. In Proceedings of the 50th Annual Design Automation Conference, DAC ’13, New York, NY, USA, 2013. Association for Computing Machinery.

[10] Amrita Ghosal and Mauro Conti. Security issues and challenges in v2x: A survey. Computer Networks, 169:107093, 2020.

[11] Mario G¨unzel, Harun Teper, Kuan-Hsun Chen, Georg von der Br¨uggen, and Jian-Jia Chen. On the equivalence of maximum reaction time and maximum data age for cause-effect chains. In 35th Euromicro Conference on Real-Time Systems (ECRTS 2023), volume 262, pages 10:1–10:22, 2023.

[12] Qingqiang He, Jinghao Sun, Nan Guan, Mingsong Lv, and Zhenyu Sun. Real-time scheduling of conditional DAG tasks with intra-task priority sssignment. IEEE Transactions on computer-aided design of integrated circuits and systems, 42(10):3196–3209, 2023.

[13] Kartheek, Nedunuri, Khare, Rashmita, Sathyamurthy, Sainathan, Manickam, Praveenkumar, and Kuchipudi, Venkata Sai Pavan. Comprehensive study on challenges to introduce autonomous emergency braking (aebs) feature in indian perspective. In Brake Colloquium & Exhibition - 42nd Annual. SAE International, sep 2024.

[14] Manzoor Ahmed Khan, Hesham El Sayed, Sumbal Malik, Talha Zia, Jalal Khan, Najla Alkaabi, and Henry Ignatious. Level-5 autonomous driving—are we there yet? a review of research literature. ACM Comput. Surv., 55(2), January 2022.

[15] Shaoshan Liu, Liangkai Liu, Jie Tang, Bo Yu, YifanWang, andWeisong Shi. Edge computing for autonomous driving: Opportunities and challenges. Proceedings of the IEEE, 107(8):1697–1716, 2019.

[16] Karolin L¨oser, Emil Delic, Michael Schreiber, Ali Hayek, and Josef B¨orcs¨ ok. The benefits of middleware for safety-related systems-on-chip. In 2014 International Symposium on Fundamentals of Electrical Engineering (ISFEE), pages 1–5, 2014.

[17] Steven Macenski, Tully Foote, Brian Gerkey, Chris Lalancette, and WilliamWoodall. Robot operating system 2: Design, architecture, and uses in the wild. Science Robotics, 7(66):eabm6074, 2022.

[18] James J Martin, Gurcan Comert, Manveen Kaur, and Adil Alsuhaim. Uncertainty quantification of cyber attacks on connected vehicles and infrastructure (part 1). Technical Report 69A3551747117, Clemson University, 2020.

[19] Kennedy Chinedu Okafor, Bamidele Adebisi, and Kelvin Anoh. Memory-efficient spike time-dependent plasticity for future driverless car park infrastructure. In 2023 2nd International Conference on Multidisciplinary Engineering and Applied Science (ICMEAS), volume 1, pages 1–5, 2023.

[20] Chengwei Pan, Jiale Geng, Wenyu Li, Feng Duan, and Shengbo Eben Li. Spiking neural network-based approximate dynamic programming for vehicle tracking control. In 2024 IEEE International Conference on Unmanned Systems (ICUS), pages 1340–1345, 2024.

[21] Nitin Rathi, Indranil Chakraborty, Adarsh Kosta, Abhronil Sengupta, Aayush Ankit, Priyadarshini Panda, and Kaushik Roy. Exploring neuromorphic computing based on spiking neural networks: Algorithms to hardware. ACM Comput. Surv., 55(12), March 2023.

[22] Ourania Spantidi, Iraklis Anagnostopoulos, and Georgios Fainekos. Efficient resource management of clustered multi-processor systems through formal property exploration. In Design, Automation and Test in Europe Conference (DATE), 2021.

[23] Hayate Toba and Takuya Azumi. Deadline Miss Early Detection Method for DAG Tasks Considering Variable Execution Time. In 36th Euromicro Conference on Real-Time Systems (ECRTS 2024), pages 8:1–8:21, 2024.

[24] TianzeWu, BaofuWu, SaWang, Liangkai Liu, Shaoshan Liu, Yungang Bao, andWeisong Shi. Oops! it’s too late. your autonomous driving system needs a faster middleware. IEEE Robotics and Automation Letters, 6(4):7301–7308, 2021.

[25] Saehanseul Yi, Tae-Wook Kim, Jong-Chan Kim, and Nikil Dutt. Energy-efficient ddaptive system reconfiguration for dynamic deadlines in autonomous driving. In 2021 IEEE 24th International Symposium on Real-Time Distributed Computing (ISORC), pages 96–104. IEEE, 2021.

[26] Kh Shahriya Zaman, Mamun Bin Ibne Reaz, Sawal Hamid Md Ali, Ahmad Ashrif A Bakar, and Muhammad Enamul Hoque Chowdhury. Custom hardware architectures for deep learning on portable devices: A review. IEEE Transactions on Neural Networks and Learning Systems, 33(11):6068–6088, 2022.

[27] Rui-Jie Zhu, Ziqing Wang, Leilani Gilpin, and Jason K. Eshraghian. Autonomous driving with spiking neural networks. In A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak, and C. Zhang, editors, Advances in Neural Information Processing Systems, volume 37, pages 136782–136804. Curran Associates, Inc., 2024.