Developing Reliable Machine Learning Methods for Autonomous Systems

Owing to the flourishing of machine learning (deep learning) techniques in the past decade, a transformative revolution has been cultivated in the development of autonomous systems, extending them from specific expertise in a restricted environment to more general intelligence in the open world. Autonomous systems are becoming omnipresent, ranging from aerospace, transport, manufacturing and agriculture to healthcare. New challenges also arise in this new era regarding the reliability of massively deploying autonomous systems into the safety-critical real world. For instance, an autonomous vehicle is expected to reliably perceive the new environment, decide and control its movement, as well as react to an emergency.

The study of autonomous systems concerns perception, prediction, reasoning, planning, control and the ability to move and interact with others. To push the envelope of reliability, this article collection will present cutting-edge research on the design of reliable machine learning methods for autonomous systems. Key to our envisioned methods includes multi-task learning for the general intelligence of autonomous systems, reinforcement learning for reasoning and planning, and adversarial learning for system security and privacy. Furthermore, this Article Collection is intended to represent a broad set of machine learning related topics to autonomous systems. We welcome submissions with theoretical, experimental, methodological, or dataset contributions, or systematic reviews if they provide substantial contributions to the state of the art.

The area of interest includes not limited to:

-Deep learning for machine perception and pattern recognition;

-Few/zero-shot learning and its application in autonomous systems;

-Explainable machine learning for automated planning and reasoning;

-Artificial intelligence for machine reasoning;

-Multi-task/modal learning for autonomous system;

-Robust and safe reinforcement learning and machine decision making;

-Trustworthy multi-agent learning and decision making;

-Safe human-agent interactions and imitation learning;

-Adversarial learning for robust and safe system;

-Trustworthy machine learning for systems security.