查看文章

Autonomous cars (also known as, self-driving/driver-less cars) are vehicles that flaunts a fully-automated driving system and are believed to be the urban mode-of-transport for the future. Ideally, such cars should be capable of comprehending its environment and navigate amidst dynamic on-road situations, to exactly mimic an efficient human driver. The major sub-systems involved in self-driving cars for their efficient operation are Perception, Planning, Control and Coordination. In this report, we elaborate on the Perception and Planning subsystems, and provide salient observations regarding some of the recent developments in these fields. Such autonomous vehicles have a network of sensors and cameras that gather bulks of information about its surrounding. Perception involves intelligently interpreting such raw sensor data into meaningful internal representations so as to gain a clear picture of its environment. A perception subsystem can be further categorized into two main functionalities viz., detection and segmentation. While the detection block is responsible for identifying objects like pedestrians, nearby cars etc. and classifying traffic signals, the segmentation block identifies chunks of semantically meaningful segments eg asphalt roads, footpaths. Together, they enable the vehicle to delineate drivable areas alongside avoiding potential obstacles. Planning subsystem involves route, behaviour and motion planning. While route planning is responsible for computing an optimal global source-terminal route, behaviour planning takes into account smoothness of roads, velocity profiles and real-time traffic information to choose a safe …