Plenary Speakers

Reinforcement learning (RL) is a prominent branch of artificial intelligence, achieving remarkable success across various domains. This talk offers a systematic introduction of RL in complex control tasks, covering its design principles and mainstream algorithms, as well as its applications in self-driving car and legged robot. How to address critical challenges in real-world deployment will be discussed, such as enforcing safety constraints, mitigating training divergence, and improving sample efficiency. Three recent advancements in algorithm design will be introduced, including: (1) distributional soft actor-critic (DSAC) algorithm, which mitigates value overestimation and achieves state-of-the-art policy performance; (2) relativistic adaptive gradient descent (RAD) algorithm, which preserves symplecticity to ensure long-term training stability; (3) diffusion actor-critic with entropy regulator (DACER) algorithm, which endows the policy with the capability of multi-modal action distributions. These contributions aim to advance the development of robust and efficient RL for real-world autonomous systems.

The impact of successful research in road traffic control spans across various domains, including the scientific, technological, social, and economic spheres. Its significance is profound, as it directly influences safety, quality of life, climate neutrality, energy resource utilization, and transportation costs. However, the development of effective methods and algorithms for road traffic control encounters notable methodological challenges. Traditionally, traffic control strategies have relied on infrastructure-based approaches. Yet, the rapid advancements in automotive technologies, traffic sensors, data processing, and communication have created unprecedented opportunities for the exploitation of an advanced type of vehicles, called connected and automated vehicles (CAVs), offering innovative solutions to longstanding traffic control challenges. This lecture will address these challenges and advancements, beginning with an overview of classical traffic control concepts. It will then focus on emerging research trends that exploit the multi-scale nature of traffic systems, from the microscopic scale of the individual CAV to the macroscopic scale of the traffic flow. Furthermore, the lecture will illustrate how efficiently controlled fleets of electric CAVs can be harmonized with the electric grid, fostering a synergistic integration between traffic and energy systems, and yielding societal benefits that extend well beyond those traditionally associated with traffic control.

The European CCAM Partnership brings together industry, research, and public stakeholders to accelerate the deployment of connected, cooperative and automated mobility across Europe and beyond. This talk presents the Partnership’s mission, major achievements, and its role in fostering innovation and large-scale testing within real-life environments. Emphasis will be placed on the knowledge exchange mechanisms established across projects, which facilitate collaboration, avoid fragmentation, and accelerate the uptake of CCAM solutions. In addition, the presentation will highlight how CCAM project results are feeding into international standardization bodies, ensuring interoperability, safety, and global impact. By showcasing success stories, lessons learned, and ongoing initiatives, this contribution aims to stimulate dialogue within the intelligent transportation systems community and strengthen cooperation across regions and domains for the benefit of future mobility.