Edge infrastructure and Industry 4.0 required services are offered by edge-servers (ESs) with different computation capabilities to run social application's workload based on a leased-price method. The usage of Social Internet of Things (SIoT) applications increases day-to-day, which makes social platforms very popular and simultaneously requires an effective computation system to achieve high service reliability. In this regard, offloading high required computational social service requests (SRs) in a time slot based on directed acyclic graph (DAG) is an NP-complete problem. Most state-of-art methods concentrate on the energy preservation of networks but neglect the resource sharing cost and dynamic subservice execution time (SET) during the computation and resource sharing. This article proposes a two-step deep reinforcement learning (DRL)-based service offloading (DSO) approach to diminish edge server costs through a DRL influenced resource and SET analysis (RSA) model. In the first level, the service and edge server cost is considered during service offloading. In the second level, the R-retaliation method evaluates resource factors to optimize resource sharing and SET fluctuations. The simulation results show that the proposed DSO approach achieves low execution costs by streamlining dynamic service completion and transmission time, server cost, and deadline violation rate attributes. Compared to the state-of-art approaches, our proposed method has achieved high resource usage with low energy consumption.

In cyber-physical systems, it is advantageous to leverage cloud with edge resources to distribute the workload for processing and computing user data at the point of generation. Services offered by cloud are not flexible enough against variations in the size of underlying data, which leads to increased latency, violation of deadline and higher cost. On the other hand, resolving above-mentioned issues with edge devices with limited resources is also challenging. In this work, a novel reinforcement learning algorithm, Capacity-Cost Ratio-Reinforcement Learning (CCR-RL), is proposed which considers both resource utilization and cost for the target cyber-physical systems. In CCR-RL, the task offloading decision is made considering data arrival rate, edge device computation power, and underlying transmission capacity. Then, a deep learning model is created to allocate resources based on the underlying communication and computation rate. Moreover, new algorithms are proposed to regulate the allocation of communication and computation resources for the workload among edge devices and edge servers. The simulation results demonstrate that the proposed method can achieve a minimal latency and a reduced processing cost compared to the state-of-the-art schemes.

The heterogeneous resource-required application tasks increase the cloud service provider (CSP) energy cost and revenue by providing demand resources. Enhancing CSP profit and preserving energy cost is a challenging task. Most of the existing approaches consider task deadline violation rate rather than performance cost and server size ratio during profit estimation, which impacts CSP revenue and causes high service cost. To address this issue, we develop two algorithms for profit maximization and adequate service reliability. First, a belief propagation-influenced cost-aware asset scheduling approach is derived based on the data analytic weight measurement (DAWM) model for effective performance and server size optimization. Second, the multiobjective heuristic user service demand (MHUSD) approach is formulated based on the CPS profit estimation model and the user service demand (USD) model with dynamic acyclic graph (DAG) phenomena for adequate service reliability. The DAWM model classifies prominent servers to preserve the server resource usage and cost during an effective resource slicing process by considering each machine execution factor (remaining energy, energy and service cost, workload execution rate, service deadline violation rate, cloud server configuration (CSC), service requirement rate, and service level agreement violation (SLAV) penalty rate). The MHUSD algorithm measures the user demand service rate and cost based on the USD and CSP profit estimation models by considering service demand weight, tenant cost, and energy cost. The simulation results show that the proposed system has accomplished the average revenue gain of 35%, cost of 51%, and profit of 39% than the state-of-the-art approaches.

Computation and memory resource management strategies are the backbone of continuous object tracking in intelligent vehicle orchestration. Multi-object tracking generates enormous measurements of targets and extended object positions using light detection and ranging (Lidar) sensors. Designing an adequate object-tracking system is a global challenge because of dynamic object detection and data association uncertainties during scene understanding. In this regard, we develop an intelligent multi-objective tracking (IMOT) system with a novel measurement model, called the box data association inflate (BDAI) model, to assess each target’s object state and trajectory without noise by using the Bayesian approach. The box object filter method filters ambiguous detection responses during data association. The theoretical proof of the box object filter is derived based on binomial expansion. Prognosticating a lower-dimension object than the original point object reduces the computational complexity of vehicle orchestration. Two datasets (NuScenes dataset and our lab dataset) are considered during the simulations, and our approach measures the kinematic states adequately with reduced computation complexity compared to state-of-the-art methods. The simulation outcomes show that our proposed method is effective and works well to detect and track objects. The NuScenes dataset contains 28130 samples for training, 6019 examples for validation and 6008 samples for testing. IMOT achieves 58.09% tracking accuracy and 71% mAP with 5 ms pre-processing time. The Jetson Xavier NX consumes 49.63% GPU and 9.37% average power and exhibits 25.32 ms latency compared to other approaches. Our system trains a single pair frame in 169.71 ms with affinity estimation time of 12.19 ms, track association time of 0.19 ms and mATE of 0.245 compared to state-of-the-art approaches