Iran University of Science and Technology
Abstract:The development of self-driving cars has garnered significant attention from researchers, universities, and industries worldwide. Autonomous vehicles integrate numerous subsystems, including lane tracking, object detection, and vehicle control, which require thorough testing and validation. Scaled-down vehicles offer a cost-effective and accessible platform for experimentation, providing researchers with opportunities to optimize algorithms under constraints of limited computational power. This paper presents a four-wheeled autonomous vehicle platform designed to facilitate research and prototyping in autonomous driving. Key contributions include (1) a novel density-based clustering approach utilizing histogram statistics for landmark tracking, (2) a lateral controller, and (3) the integration of these innovations into a cohesive platform. Additionally, the paper explores object detection through systematic dataset augmentation and introduces an autonomous parking procedure. The results demonstrate the platform's effectiveness in achieving reliable lane tracking under varying lighting conditions, smooth trajectory following, and consistent object detection performance. Though developed for small-scale vehicles, these modular solutions are adaptable for full-scale autonomous systems, offering a versatile and cost-efficient framework for advancing research and industry applications.
Abstract:The use of machine learning (ML) inference for various applications is growing drastically. ML inference services engage with users directly, requiring fast and accurate responses. Moreover, these services face dynamic workloads of requests, imposing changes in their computing resources. Failing to right-size computing resources results in either latency service level objectives (SLOs) violations or wasted computing resources. Adapting to dynamic workloads considering all the pillars of accuracy, latency, and resource cost is challenging. In response to these challenges, we propose InfAdapter, which proactively selects a set of ML model variants with their resource allocations to meet latency SLO while maximizing an objective function composed of accuracy and cost. InfAdapter decreases SLO violation and costs up to 65% and 33%, respectively, compared to a popular industry autoscaler (Kubernetes Vertical Pod Autoscaler).