Demystifying Power and Performance Bottlenecks in Autonomous Driving Systems

Abstract

Autonomous Vehicles (AVs) have the potential to radically change the automotive industry. However, computing solutions for AVs have to meet severe performance and power constraints to guarantee a safe driving experience. Current solutions either exhibit high cost and power dissipation or fail to meet the stringent latency constraints. Therefore, the popularization of AVs requires a low-cost yet effective computing system. Understanding the sources of latency and energy consumption is key in order to improve autonomous driving systems. In this paper, we present a detailed characterization of Autoware, a modern self-driving car system. We analyze the performance and power of the different components and leverage hardware counters to identify the main bottlenecks.Our approach to AV characterization avoids pitfalls of previous works: profiling individual components in isolation and neglecting LiDAR-related components. We base our characterization on a rigorous methodology that considers the entire software stack. Profiling the end-to-end system accounts for interference and contention among different components that run in parallel, also including memory transfers to communicate data. We show that all these factors have a high impact on latency and cannot be measured by profiling isolated modules.Our characterization provides novel insights, some of the interesting findings are the following. First, contention among different modules drastically impacts latency and performance predictability. Second, LiDAR-related components are important contributors to the latency of the system. Finally, a modern platform with a high-end CPU and GPU cannot achieve real-time performance when considering the entire end-to-end system.

Autonomous Vehicle Safety through the SIFT Method: A Conceptual Analysis

“…The platform integrates hardware, software, and cloud services to provide a comprehensive solution for AV development and deployment. Another important development in AV technology is the improvement of sensor fusion and perception capabilities, which enable AVs to detect and recognize objects, lanes, traffic signs, and signals, and to estimate their position, speed, and orientation [12]. Sensor fusion is the process of combining and integrating data from different sensors, such as cameras, radars, lidars, and ultrasonic sensors, to create a more accurate and robust representation of the surrounding environment.…”

Direct Tilt Controller Design with Disturbance Compensation and Implementation for a Narrow Tilting Electric Vehicle

“…The same problem exists in autonomous vehicles. On-board computers and sensors increase the power and energy consumption in autonomous vehicles studied in [28]. The same problems should be considered for the tilt control system and its subcomponents.…”

“…Actuator load torque (disturbance load) in ( 26) is applied to a first-order observer as expressed in ( 27) and (28).…”

A Survey of FPGA-Based Vision Systems for Autonomous Cars

“…Some early prototypes on self-driving cars were consuming more the 2,000 W on computing alone [25]. More recent analysis [26]- [28] estimate that the power consumption of the computing platform can be below 200 W. Considering that the typical power consumption of the rest of the car is in the order of 10 to 20 kW; and that the power consumption of heating, ventilation, and air conditioning systems (HVAC) is generally around 500W [27], this value seems reasonable. Nevertheless, current computing power estimations are based on the execution of software stacks that are not at level 5 of autonomy, as defined by the National Highway Traffic Safety Administration (NHTSA) in the US.…”

“…Some naive end-to-end latency analysis measure each individual latency and aggregate them to get the critical path. However, as pointed out by Becker [28], the overhead of communication between the different functions and the effects of their simultaneous excution in the same computing platform have a significant impact that should not be overlooked. The author analyzes the Autoware software [285] latencies, not only in isolation, but end to end, concluding that current software and hardware platforms are not able to deliver the required performance.…”