Safe and Trustworthy Autonomous Systems in Real-World Environments

Éditeurs invités

• Heye Huang, Korea Advanced Institute of Science and Technology (KAIST)
• Inhi Kim, Korea Advanced Institute of Science and Technology (KAIST)
• Boqi Li, Beihang University
• Haotian Shi, Tongji University
• Lei Yang, Nanyang Technological University
• Kailai Sun, Massachusetts Institute of Technology

Synthèse

Autonomous systems operating in complex real-world environments face significant challenges in ensuring safety, reliability, and generalization. This special issue seeks research that advances trustworthy autonomy through integration of data-driven learning with structured knowledge, world models, formal verification, and uncertainty reasoning. Contributions should address both average-case and long-tail safety-critical scenarios, with emphasis on how autonomous systems can identify risks, adapt continually, and coordinate with other agents.

The scope includes autonomous vehicles, robotics, embodied AI, and other safety-critical platforms. Submissions may combine learning, optimization, control, simulation, and human-centered methods provided they clearly contribute to trustworthy autonomous behavior. The issue welcomes original research, reviews, benchmarks, and perspective articles demonstrating methodological novelty, rigorous evaluation, or insights transferable to real-world autonomous system deployment.

Thèmes proposés

• Trustworthy autonomy and safety principles
• Uncertainty modeling and robustness
• Runtime assurance for learning-enabled systems
• Navigation and manipulation robotics
• Mobile robotics and spatial intelligence
• Perception-action learning
• Active exploration and embodied decision-making
• LLM/VLM-enabled autonomy
• Agentic AI and multimodal perception
• Tool-using agents and knowledge-enhanced models
• Foundation model adaptation for physical systems
• Risk representation and uncertainty-aware perception
• Behavior prediction and occupancy modeling
• Scene modeling and world models
• Safety-critical interaction anticipation
• Episodic and semantic memory
• Experience retrieval and in-context adaptation
• Continual learning and self-evolving autonomy
• Recursive self-improvement under safety constraints
• Safe adaptation in changing environments
• Risk-aware planning and constrained optimization
• Safe reinforcement learning
• Decision making under uncertainty
• Model-predictive control approaches
• Closed-loop safety mechanisms
• Rare-event discovery and scenario generation
• Generative simulation and counterfactual testing
• Adversarial and stress testing
• Digital twins and safety assessment
• Connected and automated vehicles
• V2X-enabled cooperation
• Multi-agent coordination
• Traffic-level autonomy and infrastructure intelligence
• System-wide mobility optimization