IEEE Intelligent Vehicles Symposium 2026
Plenary Speakers
Date & Time
Tuesday, June 23, 2026
8:30 – 9:15 AM
Session I
Christian John
President · TIER IV/Autoware Foundation
“Autoware – Democratizing Access to Physical AI”
Abstract
Physical AI, the convergence of machine learning with systems that perceive, reason, and act in the real world, represents one of the most consequential technological frontiers of our time. Yet its development has remained concentrated in resource-rich organizations, creating barriers that slow progress and limit the diversity of contributors shaping the future of mobility. This presentation will cover the work being done by the Autoware Foundation to democratize Physical AI for autonomous driving through open collaboration, transparent tooling, and a globally distributed innovation model. Central to this strategy is the development of open-source foundation models for autonomous driving, along with development and validation pipelines of AI models built from within the community. The talk with demonstrates that open-source software can compete at the frontier of the field, enabling commercialization of safe autonomous driving without proprietary lock-in. Complementing these efforts, the Autoware Open AD Kit, the first SOAFEE blueprint for Software-Defined Vehicles, provides the infrastructure for hardware-software co-design that the ecosystem requires to scale optimized autonomous driving solutions. The talk will reflect on what it means to build trusted AI for physical systems in the open, the architectural choices that enable both safety and accessibility, and the collaborative model that allows a global community to advance what no single organization could achieve alone.
Date & Time
Wednesday, June 24, 2026
8:30 – 9:15 AM
Session II
Avinash Balachandran
Vice President · Toyota Research Institute
“Beyond Autonomy: Rethinking Intelligent Vehicles Through Human-Centered Collaboration”
Abstract
Autonomous driving has traditionally been framed as a problem of replacing the human driver. In this talk, we argue for a complementary paradigm: human-centered collaboration, where AI systems are designed to work with drivers rather than independently of them. We present Physical AI systems that operate at the limits of vehicle dynamics, using shared control to augment driver inputs in steering, throttle, and braking—combining human intent with machine precision to improve safety and performance. We further introduce AI-driven training and coaching systems that leverage multimodal feedback, including augmented reality, to help drivers acquire expert skills while adapting assistance to their ability. Together, these approaches illustrate a shift from assistance to collaboration, where intelligent vehicles act as partners that amplify human capability in complex, real-world environments.
Alberto Broggi
Pier Paolo Porta
Date & Time
Wednesday, June 24, 2026
4:45 – 5:20 PM
Session III
Alberto Broggi | Pier Paolo Porta
General Manager · VisLab | Marketing Director · VisLab
“Bringing Autonomous Driving to reality”
Abstract
We present a domain-specific system-on-chip (SoC) for autonomous driving that combines high computational throughput with stringent energy efficiency, enabling sustained real-time operation under air-cooled conditions. The proposed architecture is explicitly co-designed with a full-stack autonomous driving software framework, integrating specialized hardware accelerators for perception, multi-modal sensor fusion, trajectory planning, and vehicle control. The platform supports high-bandwidth acquisition and processing of heterogeneous data streams from large-scale camera, radar, and vehicle chassis sensor arrays. Low-level sensor fusion is performed directly on-chip, followed by end-to-end AI-based inference pipelines for scene understanding, behavioral planning, and closed-loop control. The system is optimized for deterministic latency, functional safety, and robustness in complex real-world environments. Beyond the autonomous driving stack itself, we emphasize the critical role of data engineering and dataset generation in achieving reliable learning-based performance. Leveraging decades of field experience, we have developed a highly efficient end-to-end data pipeline for large-scale data acquisition, automated annotation, and systematic data selection. This pipeline enables the construction of balanced and task-specific training datasets through configurable sampling strategies and domain-driven recipes, supporting robust and scalable neural network training. Our autonomous driving stack and data infrastructure are fully proprietary and have been developed in-house over more than ten years of stealth-mode research and iterative validation. The presented chip represents the 6th generation within a family of custom processors, incorporating architectural refinements derived from extensive field testing and previous deployments. The design reflects nearly three decades of experience in autonomous driving systems and a decade-long transition toward fully learning-based approaches, enabling systematic handling of rare events and long-tail corner cases. We describe the hardware–software co-design methodology, accelerator microarchitectures, memory management, and interconnect mechanisms that enable efficient end-to-end processing. Experimental results demonstrate competitive performance-per-watt and scalable throughput across perception and planning workloads. This work illustrates how deep vertical integration—from custom silicon and sensing technologies to large-scale data infrastructure and application-level deployment—enables the development of fully optimized autonomous driving systems. By integrating in-house radar technology with advanced vision processing and AI pipelines, our platform delivers an end-to-end solution spanning hardware, sensing, learning, and scalable deployment tools, providing strong guarantees in terms of efficiency, reliability, and long-term evolvability.)
Date & Time
Wednesday, June 24, 2026
5:25 – 6:00 PM
Session IV
Ali Peker
Chief Executive Officer · ADASTEC
“Scaling Automated Public Transit: Why Bus Automation Requires a New Deployment Model”
Abstract
This session will explore why automating buses is fundamentally different from extending passenger-car autonomy to public transit. Buses carry more passengers, operate under fixed service expectations, and must deliver safety, reliability, accessibility, and comfort as part of a continuous mobility service. In this context, automation must go beyond the driving task to address the full operational role traditionally performed by a professional driver, from precise bus-stop docking and passenger experience to remote supervision, infrastructure interaction, and service continuity. The presentation will introduce ADASTEC’s scalable SAE Level-4 automated transit deployment model through its flowride.ai platform. It will examine how clearly defined Operational Design Domains, redundant and fail-safe architecture, remote operations, and integration with transit systems enable public transport operators to transition selected conventional bus services into automated operations. Drawing on real-world use cases including airports, campuses, BRT corridors, city routes, and on-demand services, the session will highlight how automated transit can support driver shortage mitigation, improved service reliability, reduced operational expenditure, and long-term automation roadmap.
Date & Time
Thursday, June 25, 2026
8:30 – 9:00 AM
Session V
Linda Cadwell Stancin
Executive Director and Head of Research & Development · General Motors
“Your Car, Your Co-Pilot: Research for Vehicles as Intelligent and Efficient Assistants”
Abstract
The vehicle is evolving into a context-aware companion capable of sensing and reasoning. When chosen by the driver, it can provide more intelligent, safer, less intrusive, and more personalized support. This shift is driven by the convergence of connectivity, sensing, and AI across personal devices, cloud services, and cabin intelligence. Conversational AI together with advanced sensing capability shows how the cabin can evolve from a set of discrete features into a foundation for driver-state understanding, occupant support, and more natural human-vehicle interaction. The next research frontier lies in focused advances in multimodal fusion, driver and occupant state inference, and bounded in-vehicle AI, with close academia–industry collaboration needed to translate these research ideas into practical systems.
