The topic of self-driving cars is receiving a lot of press lately. This paper on safe autonomy is an interesting read for people researching autonomous vehicles or more generally autonomous control systems. While the authors provide readers with some background on autonomous control system, the paper is certainly not a primer for beginners.

The focus of so-called chance constrained temporal logic (C2TL) is safety requirements. As stated in the paper, “chance constraints provide a natural way to specify probabilistic correctness properties” and “capture perception uncertainty.” After providing formal definitions, the paper continues with how to formulate and synthesize autonomous vehicle control subject to C2TL specifications.

As proof of concept, the authors present case studies on “navigation in an uncertain map,” changing lanes, “passing a vehicle using oncoming traffic lane,” and risk distribution.

In reality, environments are dynamic and noisy. This is not addressed. While a more adaptive approach is needed, the paper presents “a first-step towards [the] design of autonomous systems with assurance in [the] presence of perception uncertainty.”