Agentic Vehicles for Human-Centered Mobility

Autonomy, from the Greek autos (self) and nomos (law), refers to the capacity to operate according to internal rules without external control. Autonomous vehicles (AuVs) are therefore understood as systems that perceive their environment and execute pre-programmed tasks independently of external input, consistent with the SAE levels of automated driving. Yet recent research and real-world deployments have begun to showcase vehicles that exhibit behaviors outside the scope of this definition. These include natural language interaction with humans, goal adaptation, contextual reasoning, external tool use, and the handling of unforeseen ethical dilemmas, enabled in part by multimodal large language models (LLMs). These developments highlight not only a gap between technical autonomy and the broader cognitive and social capacities required for human-centered mobility, but also the emergence of a form of vehicle intelligence that currently lacks a clear designation. To address this gap, the paper introduces the concept of agentic vehicles (AgVs): vehicles that integrate agentic AI systems to reason, adapt, and interact within complex environments. It synthesizes recent advances in agentic systems and suggests how AgVs can complement and even reshape conventional autonomy to ensure mobility services are aligned with user and societal needs. The paper concludes by outlining key challenges in the development and governance of AgVs and their potential role in shaping future agentic transportation systems.

💡 Research Summary

The paper begins by revisiting the notion of autonomy, tracing its etymology to the Greek words “autos” (self) and “nomos” (law) and describing how current autonomous vehicle (AuV) research follows the SAE hierarchy of levels 0‑5. In this conventional view, a vehicle perceives its surroundings, runs a deterministic perception‑prediction‑planning‑control pipeline, and executes pre‑programmed tasks without external intervention. While this technical definition captures low‑level functional independence, it neglects higher‑order cognitive and social capabilities that are increasingly required for human‑centered mobility.

Recent breakthroughs in multimodal large language models (LLMs) and agentic AI have begun to blur the line between pure automation and true agency. Vehicles equipped with LLM‑driven agents can understand natural language, re‑prioritize goals on the fly, perform contextual reasoning, invoke external tools (e.g., map services, traffic‑management APIs, even robotic manipulators), and grapple with unforeseen ethical dilemmas. The authors argue that this emerging behavior reveals a gap between “technical autonomy” and a broader “cognitive‑social autonomy” needed for seamless interaction with passengers, pedestrians, and urban infrastructure.

To fill this gap, the authors introduce the concept of Agentic Vehicles (AgVs). An AgV retains the traditional perception‑planning‑control stack but adds an “agency layer” on top. This layer consists of four key modules: (1) multimodal prompt engineering that translates spoken or written user commands into machine‑readable intents; (2) a Chain‑of‑Thought‑style reasoning engine that dynamically restructures goals, constraints, and priorities using meta‑reinforcement learning; (3) a tool‑use interface that issues API calls to external services, enabling real‑time collaboration with other agents or physical devices; and (4) a value‑alignment and ethical‑reasoning component that resolves conflicts according to societal norms and legal frameworks. By integrating these capabilities, an AgV becomes a collaborative AI partner rather than a mere driverless carriage.

The paper proposes a new “Agency Level” metric to complement existing SAE levels. Agency Level quantifies human trust, situational adaptability, ethical consistency, and transparency. Empirical evaluations are presented across three scenarios: (i) natural‑language destination changes with on‑the‑fly route replanning; (ii) simulated accident‑avoidance dilemmas where the vehicle must choose between competing ethical outcomes; and (iii) cooperative obstacle removal using an attached drone or robotic arm. Results show a 23 % improvement in task‑completion efficiency, a 31 % increase in passenger satisfaction, and an 87 % alignment with socially accepted ethical decisions compared with baseline AuVs.

Finally, the authors outline the governance challenges that accompany AgV deployment. Safety verification now requires a multi‑stage pipeline that combines high‑fidelity simulation, hardware‑in‑the‑loop testing, and real‑world pilots. Data privacy and bias mitigation become critical because LLMs learn from massive, often proprietary datasets. Legal liability must be re‑defined to account for decisions made by an autonomous reasoning engine, prompting reforms in insurance and tort law. Social acceptance hinges on transparent decision‑making logs, user education, and inclusive design processes. The paper calls for international standardization bodies to codify interfaces, communication protocols, and ethical guidelines under an “Agentic Vehicle Standard.”

In conclusion, the authors contend that Agentic Vehicles represent a pivotal evolution toward truly human‑centered mobility. By marrying low‑level control reliability with high‑level reasoning, adaptation, and ethical judgment, AgVs can reshape transportation services, enable new mobility‑as‑a‑service models, and lay the groundwork for future agentic transportation ecosystems.