Deep learning (DL) continues to play a pivotal role in a wide range of intelligent systems, including autonomous machines, smart surveillance, industrial automation, and portable healthcare technologies. These applications often demand low-latency inference and efficient resource utilization, especially when deployed on embedded or edge devices with limited computational capacity. As DL models become increasingly complex, selecting the right inference framework is essential to meeting performance and deployment goals. In this work, we conduct a comprehensive comparison of five widely adopted inference frameworks: PyTorch, ONNX Runtime, TensorRT, Apache TVM, and JAX. All experiments are performed on the NVIDIA Jetson AGX Orin platform, a high-performance computing solution tailored for edge artificial intelligence workloads. The evaluation considers several key performance metrics, including inference accuracy, inference time, throughput, memory usage, and power consumption. Each framework is tested using a wide range of convolutional and transformer models and analyzed in terms of deployment complexity, runtime efficiency, and hardware utilization. Our results show that certain frameworks offer superior inference speed and throughput, while others provide advantages in flexibility, portability, or ease of integration. We also observe meaningful differences in how each framework manages system memory and power under various load conditions. This study offers practical insights into the trade-offs associated with deploying DL inference on resource-constrained hardware.

The scheduling of mixed-criticality (MC) systems with graceful degradation is considered, where LO-criticality tasks are guaranteed some service in HI mode in the form of minimum cumulative completion rates. First, we present an easy to implement admission-control procedure to determine which LO-criticality jobs to complete in HI mode. Then, we propose a demand-bound-function-based MC schedulability test that runs in pseudo-polynomial time for such systems under EDF-VD scheduling, wherein two virtual deadline setting heuristics are considered. Furthermore, we discuss a mechanism for the system to switch back from HI to LO mode and quantify the maximum time duration such recovery process would take. Finally, we show the effectiveness of our proposed method by experimental evaluation in comparison to state-of-the-art MC schedulers.

Many safety critical real-time systems are considered certified when they meet failure probability requirements with respect to the maximum permitted incidences of failure per hour. In this paper, the mixed-criticality task model with multiple worst case execution time (WCET) estimations is extended to incorporate such system-level certification restrictions. A new parameter is added to each task, characterizing the distribution of the WCET estimations -- the likelihood of all jobs of a task finishing their executions within the less pessimistic WCET estimate. An efficient algorithm named LFF-Clustering is derived for scheduling mixed-criticality systems represented by this model. Experimental analyses show our new model and algorithm out-perform current state-of-the-art mixed-criticality scheduling algorithms.

This paper considers for the first time end-to-end response-time analysis for DAG-based real-time task systems implemented on heterogeneous multicore platforms. The specific analysis problem that is considered was motivated by an industrial collaboration involving wireless cellular base stations. The DAG-based systems considered herein allow intra-task parallelism: while each invocation of a task (i.e., DAG node) is sequential, successive invocations of a task may execute in parallel. In the proposed analysis, this characteristic is exploited to reduce response-time bounds. Additionally, there is some leeway in choosing how to set tasks' relative deadlines. It is shown that by resolving such choices holistically via linear programming, response-time bounds can be further reduced. Finally, in the considered use case, DAGs are defined based upon just a few templates and individually often have quite low utilizations. It is shown that, by combining many such DAGs into one of higher utilization, response-time bounds can often be drastically lowered. The effectiveness of these techniques is demonstrated via both case-study and schedulability experiments.

This paper discusses challenges in computer systems research posed by the emerging autonomous driving systems. We first identify four research areas related to autonomous driving systems: real-time and embedded systems, machine learning, edge computing, and cloud computing. Next, we sketch two fatal accidents caused by active autonomous driving, and uses them to indicate key missing capabilities from today’s systems. In light of these research areas and shortcomings, we describe a vision of digital driving circumstances for autonomous vehicles and refer to autonomous vehicles as "clients" of this digital driving circumstance. Then we propose a new research thrust: collaborative autonomous driving. Intuitively, requesting useful information from a digital driving circumstance to enable collaborative autonomous driving is quite sophisticated (e.g., collaborations may come from different types of unstable edge devices), but it also provide us various research challenges and opportunities. The paper closes with a discussion of the research necessary to develop these capabilities.