Self-Driving Cars: A Survey

Abstract

We survey research on self-driving cars published in the literature focusing on autonomous cars developed since the DARPA challenges, which are equipped with an autonomy system that can be categorized as SAE level 3 or higher. The architecture of the autonomy system of self-driving cars is typically organized into the perception system and the decision-making system. The perception system is generally divided into many subsystems responsible for tasks such as self-driving-car localization, static obstacles mapping, moving obstacles detection and tracking, road mapping, traffic signalization detection and recognition, among others. The decision-making system is commonly partitioned as well into many subsystems responsible for tasks such as route planning, path planning, behavior selection, motion planning, and control. In this survey, we present the typical architecture of the autonomy system of self-driving cars. We also review research on relevant methods for perception and decision making. Furthermore, we present a detailed description of the architecture of the autonomy system of the self-driving car developed at the Universidade Federal do Espírito Santo (UFES), named Intelligent Autonomous Robotics Automobile (IARA). Finally, we list prominent self-driving car research platforms developed by academia and technology companies, and reported in the media.

Affordance-based Reinforcement Learning for Urban Driving

“…The modular approaches aim to divide the entire task into different sub-tasks and sub-modules that include the core functional blocks of localization and mapping, perception, prediction, planning, and decision making, and vehicle control [12,30,39,45,48,70,74]. The localization and mapping subsystem senses the state of the world and locates the ego-vehicle with respect to the environment [11,35,6,50,74]. This is followed by the perception sub-system that detects and tracks all surrounding static and dynamics objects [4,50,53,74,18,6].…”

“…The localization and mapping subsystem senses the state of the world and locates the ego-vehicle with respect to the environment [11,35,6,50,74]. This is followed by the perception sub-system that detects and tracks all surrounding static and dynamics objects [4,50,53,74,18,6]. The intermediate representation produced by the perception subsystem then feeds into the prediction and planning subsystems that output an optimal plan of action [63,51,6,50,74] based on the future trajectories of all the agents in the environment [38,19,46] and the planning cost function.…”

“…This is followed by the perception sub-system that detects and tracks all surrounding static and dynamics objects [4,50,53,74,18,6]. The intermediate representation produced by the perception subsystem then feeds into the prediction and planning subsystems that output an optimal plan of action [63,51,6,50,74] based on the future trajectories of all the agents in the environment [38,19,46] and the planning cost function. Finally, the plan of action is mapped into low-level control actions that are responsible for the motor actuation.…”

“…Finally, the plan of action is mapped into low-level control actions that are responsible for the motor actuation. For more details on these approaches, we direct readers to [74,6,39]. Although these systems offer high interpretability, they employ a heavily engineered approach that requires cumbersome parameter tuning and large amounts of annotated data to capture the diverse set of scenarios that the autonomous driving vehicle may face.…”

End-to-end Autonomous Driving Perception with Sequential Latent Representation Learning

“…A typical autonomous driving system is organized as a two parts architecture [1], [32]: a perception system, and a decision making system, as shown in Fig. 1.…”