The field of cyber-physical systems is moving towards developing more robust and reliable safety monitoring systems, particularly in out-of-distribution scenarios. Researchers are focusing on creating novel approaches that can directly monitor safety properties, rather than just detecting out-of-distribution data. This shift in paradigm is expected to improve the overall reliability and performance of learning-enabled cyber-physical systems. Noteworthy papers in this area include:
- A paper proposing a monitor and recover paradigm, which emphasizes robust safety monitoring and distribution shift recovery, offering a promising direction for future research.
- A study introducing TrustLoRA, a low-rank adaptation framework for failure detection under out-of-distribution data, which enhances failure detection ability effectively and flexibly.
- Research investigating the challenge of establishing stochastic-like guarantees when sequentially learning from a stream of i.i.d. data that includes an unknown quantity of clean-label adversarial samples, providing a theoretical analysis of a disagreement-based learner for thresholds subject to a clean-label adversary with noise.