Abstract: Systems and techniques are provided for detecting occurrence of traffic scenes and characterizing performance of autonomous vehicles (AVs) in relation to said traffic scenes. An example method includes receiving a collection of data compiled by one or more AVs while navigating a real-world environment, wherein the one or more AVs are configured to execute a first version of AV software; identifying, based on at least one traffic scene selector and the collection of data, a plurality of traffic scene datasets each corresponding to an instance in which the one or more AVs encountered a traffic scene; and determining, based on the plurality of traffic scene datasets, one or more metrics for characterizing an operation of the one or more AVs in relation to the traffic scene.

Abstract: The present disclosure generally relates systems and techniques for classification of signs and gestures pertaining to traffic and, more specifically, to autonomous vehicle based classification of signs and gestures of humans controlling traffic (HCTs). In some aspects, the present disclosure provides a process for identifying, using one or more sensors of an autonomous vehicle (AV), a human controlling traffic (HCT) and recognizing a set of instructions from the HCT, wherein the set of instructions comprises a set of sensory data. In some aspects, the process can further include operations for modifying a trajectory of the AV based on the set of instructions received from the HCT and based on a determination on whether the set of instructions are intended for the AV.