swerve driving strategy

[USeebi]

In "Simulating a Self-Driving Car Using End-to-End Deep Learning with Keras" Ch 16 contributed by guest authors: Aditya Sharma and Mitchell Spryn from Practical Deep Learning for Cloud, Mobile, and Edge_ Real-World AI & Computer-Vision Projects Using Python, Keras & TensorFlow-O'Reilly Media (2019), there is a graph showing two training strategies, namely normal and swerve. Could anyone enlighten what exactly is a swerve driving strategy in training autonomous driving, how we are going to setup in simulation like AirSim?