Average Energy Consumption Research Articles

Graph-based meta-learning for context-aware sensor management in nonlinear safety-critical environments

This study introduces a novel framework for optimizing energy efficiency and computational load in safety-critical robotic systems operating in nonlinear domains. Leveraging Graph Attention Networks for state awareness and decision-making, the framework employs adaptive sensor and filter toggling strategies to dynamically manage system resources through real-time inferential processes. Our framework maintains continuous robot operation in the presence of sensor noise and environmental disturbances by activating additional sensors, thus preventing system shutdowns or stalls. Few-shot meta-learning techniques further augment the model's adaptability, allowing it to generalize and make real-time decisions across varying operational conditions. An extensive evaluation reveals a reduced average energy consumption, compared to ‘always-on’ configurations, by 13.71% and CPU utilization by 29.07%, without compromising system performance and safety. We also introduce Matching Networks and Siamese Networks with different loss functions to assess the system's capability to adapt to different levels of criticality. Our experiments demonstrate that the system prioritizes performance and safety in high-critical scenarios while maximizing energy efficiency in less critical situations. The framework's real-time decision-making capability is particularly crucial in human–robot environments and holds significant implications for future applications in nonlinear control systems and resilient robotic systems.

Deadline-aware task offloading in vehicular networks using deep reinforcement learning

Smart vehicles have a rising demand for computation resources, and recently vehicular edge computing has been recognized as an effective solution. Edge servers deployed in roadside units are capable of accomplishing tasks beyond the capacity which is embedded inside the vehicles. However, the main challenge is to carefully select the tasks to be offloaded considering the deadlines, and in order to reduce energy consumption, while delivering a good performance. In this paper, we consider a vehicular edge computing network in which multiple cars are moving at non-constant speed and produce tasks at each time slot. Then, we propose a task offloading algorithm, aware of the vehicle’s direction, based on Rainbow, a deep Q-learning algorithm combining several independent improvements to the deep Q-network algorithm. This is to overcome the conventional limits and to reach an optimal offloading policy, by effectively incorporating the computation resources of edge servers to jointly minimize average delay and energy consumption. Real-world traffic data is used to evaluate the performance of the proposed approach compared to other algorithms, in particular deep Q-network, double deep Q-network, and deep recurrent Q-network. Results of the experiments show an average reduction of 18% and 15% in energy consumption and delay, respectively, when using the proposed Rainbow deep Q-network based algorithm in comparison to the state-of-the-art. Moreover, the stability and convergence of the learning process have significantly improved by adopting the Rainbow algorithm.

Gait Factor on the Energy Harvesting for a Simple Biped Robot

To design walk-fast and energy-efficient robots, there has been lots of work in the last decade examining the locomotion dynamics of a passive biped. As the walking environment or system parameter changes, an energy use efficient robot may become inefficient. A possible approach to increase the energy efficiency is through the ability to harvest the energy used during the locomotion. The paper’s main goal is to investigate the relations between walking speed, the locomotion energy consumption of a passive biped, and the ability to retrieve the lost energy as locomotion energy efficiency varies. Piezoelectric bimorphs were attached to the feet of the biped to harvest energy via exploiting the acceleration excitations induced vibrations at the instant foot lift and heel strike. It is found that as a foot-to-hip mass ratio increases, the stable periodic-1 (P1) walking gait becomes slower and more energy costing. Also it means more available energy to harvest, although the retrieved energy is much smaller compared to the locomotive energy. Once the foot-to-hip mass ratio passes the periodic doubling (PD) point, P1 walking gaits will become limped P2 walking gaits, and the high energy cost situation alleviates, which also means less available energy to harvest. On the other hand, if the foot-to-hip mass ratio is fixed and the slope angle increases, the walking will experience sequences of PD bifurcations, and the walking gaits go through P1, P2, P4, P8, and chaotic walking. As the walking gaits change, the average walking efficiency, average locomotion energy consumption, and average harvested energy grow as the slope becomes deeper.

Performance analysis of congestion-aware Q-routing algorithm for network on chip

<p>A network on chip’s performance is greatly impacted by network congestion due to the substantial increase in latency and energy utilized. Designing routing strategies that keep the network informed of the status of traffic is made easier by machine learning techniques. In this work, a reinforcement-based congestion-aware Q-routing (CAQR) technique has been presented. The proposed algorithm performed better in comparison to the conventional XY routing method tested against the SPEC CPU2006 benchmark suite in the gem5 NoC simulator tool. The suite used has 4 benchmarks, namely, namd, lbm, leslie3d and bzip2 which can be used for the cores in the network in any combination. The tests were run with 16 cores on a 44 network with the maximum instruction count supported by the system (here 5,000). The proposed Q-routing algorithm showed an average of 19% reduction for benchmark simulation as compared to the Dimension-ordered (X-Y) routing for readings of average packet latency which is a crucial factor in determining a network’s efficiency. The analysis also shows an average reduction of 24%, 10%, 23% and 47% in terms of average packet network latency, average flit latency, average flit network latency and average energy consumption across various benchmarks.</p>

Proximate node aware optimal and secure data aggregation in wireless sensor network based IoT environment

<span>Internet of things (IoT) has become one of the eminent phenomena in human life along with its collaboration with wireless sensor networks (WSNs), due to enormous growth in the domain; there has been a demand to address the various issues regarding it such as energy consumption, redundancy, and overhead. Data aggregation (DA) is considered as the basic mechanism to minimize the energy efficiency and communication overhead; however, security plays an important role where node security is essential due to the volatile nature of WSN. Thus, we design and develop proximate node aware secure data aggregation (PNA-SDA). In the PNA-SDA mechanism, additional data is used to secure the original data, and further information is shared with the proximate node; moreover, further security is achieved by updating the state each time. Moreover, the node that does not have updated information is considered as the compromised node and discarded. PNA-SDA is evaluated considering the different parameters like average energy consumption, and average deceased node; also, comparative analysis is carried out with the existing model in terms of throughput and correct packet identification.</span>

Utilization of Smart Agricultural Technology to Improve Resource Efficiency in Agro-industry

This study investigates the utilization of smart agricultural technologies to improve resource efficiency in the agro-industry, using a quantitative approach with a focus on Structural Equation Modeling - Partial Least Squares (SEM-PLS) analysis. A survey of 250 agro-industry stakeholders produced descriptive statistics showing a high mean adoption score (4.2) and a significant frequency of adoption (75%). Resource efficiency indicators, including average water use (32.5 gallons per hectare), average energy consumption (15.8 kWh per hectare), and average crop yield (2,800 kg per hectare), were also assessed. The SEM-PLS results showed strong reliability and validity of the measurement model, with positive path coefficients indicating substantial impacts of smart technology adoption on water use efficiency, energy consumption optimization, and crop yield. The model showed a satisfactory fit, and bootstrapping confirmed the robustness of the relationships. The discussion highlights practical implications for farmers, policymakers, and technology providers, emphasizing the potential for increased efficiency, reduced costs, and improved yields through the adoption of smart technologies. This study contributes valuable insights to the discourse of sustainable agricultural practices.

An energy recovery routing algorithm for opportunistic networks based on connection stability and node encounter type

Routing algorithms have been the focus of research on opportunistic networks to effectively deal with the connectivity problem of intermittent communication in opportunistic networks. Due to the frequent forwarding of messages by active nodes in the opportunistic networks leading to excessive energy consumption, opportunistic networks suffer from communication susceptibility to interruption and unstable network connectivity. In order to solve this problem, this paper proposes an opportunistic network routing algorithm (ST-ProPhet) based on connection stability and energy recovery, which takes into account the node’s historical average connection duration, encounter probability, and the node’s energy consumption. Firstly, a node encounter probability prediction model fusing the average connection duration is developed. Secondly, the base station is introduced and the node wireless energy recovery strategy is designed. Finally, the ST-ProPhet algorithm proposes a message forwarding strategy based on node encounter types on the basis of the encounter probability prediction model and energy recovery model. Simulation experiments show that the ST-ProPhet algorithm has an overall reduction of 5066.27mAh over the ProPhet algorithm in terms of average message energy consumption.

FASA-DRAM: Reducing DRAM Latency with Destructive Activation and Delayed Restoration

DRAM memory is a performance bottleneck for many applications, due to its high access latency. Previous work has mainly focused on data locality, introducing small-but-fast regions to cache frequently accessed data, thereby reducing the average latency. However, these locality-based designs have three challenges in modern multi-core systems: 1) Inter-application interference leads to random memory access traffic. 2) Fairness issues prevent the memory controller from over-prioritizing data locality. 3) Write-intensive applications have much lower locality and evict substantial dirty entries. With frequent data movement between the fast in-DRAM cache and slow regular arrays, the overhead induced by moving data may even offset the performance and energy benefits of in-DRAM caching. In this paper, we decouple the data movement process into two distinct phases. The first phase is Load-Reduced Destructive Activation (LRDA), which destructively promotes data into the in-DRAM cache. The second phase is Delayed Cycle-Stealing Restoration (DCSR), which restores the original data when DRAM bank is idle. LRDA decouples the most time-consuming restoration phase from activation, and DCSR hides the restoration latency through prevalent bank-level parallelism. We propose FASA-DRAM incorporating destructive activation and delayed restoration techniques to enable both in-DRAM caching and proactive latency-hiding mechanisms. Our evaluation shows that FASA-DRAM improves the average performance by 19.9% and reduces average DRAM energy consumption by 18.1% over DDR4 DRAM for four-core workloads, with less than 3.4% extra area overhead. Furthermore, FASA-DRAM outperforms state-of-the-art designs in both performance and energy efficiency.

Improving performance metrics in WBANs with a dynamic next beacon interval and superframe duration scheme

The advancement of networking, information, and communication technologies has fueled the popularity of Wireless Body Area Networks (WBANs) in both medical (remote patient monitoring) and non-medical sectors. Due to low, medium, and high data traffic requirements, WBAN performance suffers during the synchronization process that generates periodic beacon frames between sensor nodes and the coordinator. It also suffers when sensor nodes implicitly send data to the coordinator during the fixed time slot using the Contention Access Period (CAP). In this study, we propose a solution called Dynamic Next Beacon Interval and Superframe Duration Scheme (DNBISD) to tackle these issues of the IEEE 802.15.4 standard. This standard relies on a Beacon Interval (BI) and CAP for synchronization and data transmission between sensor nodes and the coordinator. However, the standard must adapt to BI and CAP's changing traffic load requirements, resulting in drawbacks such as prolonged packet delays, increased energy consumption, and potential data loss, particularly in real-time patient monitoring applications. In order to overcome these challenges, our DNBISD scheme employs a fuzzy approach to adapt the BI and CAP based on requested synchronization and data considering input parameters like packet received ratio and buffer ratio. The inference system utilizes the Takagi, Sugeno, and Kang (TSK) fuzzy model for rational quantitative analysis. Simulations demonstrate that our proposed scheme significantly enhances data transmission, boosts the average packet delivery ratio and throughput, and reduces the coordinator's average packet loss ratio and energy consumption. Consequently, this improvement allows for more efficient data transfer among numerous nodes within the specified superframe structure.

Design of Long-Life Wireless Near-Field Hydrogen Gas Sensor.

A compact wireless near-field hydrogen gas sensor is proposed, which detects leaking hydrogen near its source to achieve fast responses and high reliability. A semiconductor-type sensing element is implemented in the sensor, which can provide a significant response in 100 ms when stimulated by pure hydrogen. The overall response time is shortened by orders of magnitude compared to conventional sensors according to simulation results, which will be within 200 ms, compared with over 25 s for spatial concentration sensors under the worst conditions. Over 1 year maintenance intervals are enabled by wireless design based on the Bluetooth low energy protocol. The average energy consumption during a single alarm process is 153 μJ/s. The whole sensor is integrated on a 20 × 26 mm circuit board for compact use.

Multi-attribute Feature Fusion Algorithm of IoT Communication Big Data Based on Ant Colony Neural Network

Abstract To address the issues of time-consuming and inaccurate feature fusion in traditional Internet of Things (IoT) communication big data multi-attribute fusion algorithms, a new algorithm based on the ant colony neural network is proposed. Modeling and feature decomposition processing is carried out on IoT communication big data, communication data information and data attribute characteristics are optimized into different data groups according to a data clustering algorithm, and a rough set calculation method is used to calculate information attribute characteristics. The genetic algorithm is used to improve the ant colony neural network, improve the search ability, extract the multi-attribute characteristics of IoT communication big data, filter the current big data’s characteristic attribute demand through the comparison method, and achieve multi-attribute feature fusion of IoT communication big data through the similarity measurement method. The experimental results show that the fusion error of the IoT communication big data multi-attribute feature fusion algorithm based on the ant colony neural network is small and relatively stable. During the 600 MB data fusion process, the data fusion processing time of the method in this paper is 150 ms, and the average energy consumption ratio is below 2 %, which reflects better IoT communication big data multi-attribute feature fusion performance.

Changed pattern of physical activity and energy expenditure among mid-pregnancy women during China's COVID-19 lockdown: A cross-sectional study

BackgroundThis study examines the physical activity (PA) patterns and energy expenditure of Chinese pregnant women during the COVID-19 lockdown, their influencing factors, and explores potential improvement strategies. MethodsA survey using the Chinese version of the Pregnancy Physical Activity Questionnaire (PPAQ) was conducted among 610 women in their second trimester. It compared their PA patterns and energy expenditure during the lockdown with pre-pandemic levels. Factors influencing PA were analyzed using one-way ANOVA and regression. ResultsAlthough aware of PA benefits, respondents overestimated their PA levels during lockdown, showing significantly lower PA and energy expenditure than before pregnancy, primarily in housework and caregiving activities. Average energy expenditure was 34.3 MET-hrs/wk. Influential factors included concerns about miscarriage/preterm birth, parity, pre-pregnancy exercise habits, and family opposition. ConclusionsThe lockdown notably decreased PA patterns and energy expenditure in mid-pregnancy, highlighting a gap between perception and practice. Activities were mainly sedentary to light-intensity. Influencing factors suggest the need for targeted health education and support. Implementing adaptive strategies, personalized antenatal care, effective health education, family engagement, and digital health tools could enhance PA during restrictive periods like lockdowns.

Flight-Path Optimization for a Hybrid-Electric Aircraft

Abstract This study aims to illustrate a sequence that optimizes the flight-path trajectory for a hybrid-electric aircraft at mission level, in addition to identifying the respective optimum power management strategy. An in-house framework for hybrid-electric propulsion system modeling is utilized. A hybrid-electric commuter aircraft serves as a virtual test-bench. Vectorized calculations, decision variable count, and optimization algorithms are considered for reducing the computational time of the framework. Performance improvements are evaluated for the aircraft's design mission profile. Total energy consumption is set as the objective function. Emphasis lies on minimizing the average value and standard deviation of the energy consumption and timeframe metrics. The best performing application decreases computational time by two orders of magnitude, while retaining equal accuracy and consistency as the original model. It is employed for creating a dataset for training an artificial neural network (ANN) against random mission patterns. The trained network is integrated into a surrogate model. The latter part of the analysis evaluates optimized mission profile characteristics with respect to energy consumption, against a benchmark flight-path. The combined optimization process decreases the multihour-scale timeframe by two orders of magnitude to a 3-min sequence. Using the novel framework, a 12% average energy consumption benefit is calculated for short, medium, and long regional missions, against equivalent benchmark profiles.

Eating Behavior and Nutritional Profile of Children with Autism Spectrum Disorder in a Reference Center in the Amazon.

There is no single pattern for the evolution of the nutritional status of children with autism spectrum disorder (ASD). Previous studies have found a tendency towards food selectivity with food monotony and difficulties with food texture in children with ASD, but studies in this area, especially in Brazil, are still scarce. The nutritional profile and changes in eating behavior were analyzed in patients with autism spectrum disorder assisted at a reference center in Belém. Eating behavior was assessed using the Labyrinth Scale, nutritional status assessment through weight and height (to calculate body mass index-BMI), and consumption food through the 24 h reminder. A total of 80 children of both sexes participated in the study, the majority of whom were male (80%), 47.5% eutrophic, while for the food consumption of the children evaluated, there was an average energy consumption of 1911 kcal daily, with 57.3%, 15.4%, and 27.3% of carbohydrates, proteins, and lipids, respectively. In relation to eating behavior, the highest averages were demonstrated in the domains of food selectivity, behavioral aspects, and mealtime skills. Masticatory motor scores showed a positive correlation with weight, BMI, and the amount of energy consumed by the child. The gastrointestinal symptoms score showed a negative correlation with the child's age. Regarding mealtime skills, a negative correlation was observed with the proportion of carbohydrates in the diet and a positive correlation with the proportion of lipids consumed in the children's diet. Therefore, knowing the main changes in eating behavior is important to ensure a complete and safe approach for each patient.

Joint Dataset Reconstruction and Power Control for Distributed Training in D2D Edge Network

The intrinsic nature of non-independent and identically distributed datasets on heterogeneous devices slows down the distributed model training process and reduces the training accuracy. To settle this problem, we propose a dataset reconstruction scheme to transform the data distribution of training device’s dataset into independent and identically distributed dataset via data exchange among trusted devices. For energy efficiency, we further consider power control for the devices. We then formulate an optimization problem, which is a mixed integer non-linear programming problem, to minimize the total energy consumption for each round of distributed training. Due to the NP-hardness and coupling property of the optimization problem, we decompose it into two subproblems for dataset reconstruction and power control, respectively. An approximation algorithm is designed to obtain a near-optimal auxiliary devices set for dataset reconstruction with minimum energy consumption, while meeting the variance constraint of the optimization problem. We prove that approximation algorithm has a worst-case approximation ratio of 1+ln|Ωi(t)|, where |Ωi(t)| is the required data samples for dataset reconstruction of each training device. For power control, we design a dynamic programming algorithm to further reduce the energy consumption. For comparison, we propose three benchmark schemes that adopt either one of the algorithms or neither. We also customize three baseline algorithms based on the state-of-the-arts to compare with our proposed algorithm. Numerical results show that, our proposed algorithm outperforms three benchmarks on the average energy consumption for one round for different cases. When varying the labels that each device owns, our proposed algorithm outperforms the other three baseline algorithms on training accuracy. Besides, when setting a target accuracy, our proposed algorithm always has the lowest energy consumption.

Treatment of Monochlorobenzene from Polymers Process through Electrochemical Oxidation.

With the rapid development of the economy and the demands of people's lives, the usage amount of polymer materials is significantly increasing globally. Chlorobenzenes (CBS) are widely used in the industrial, agriculture and chemical industries, particularly as important chemical raw materials during polymers processes. CBS are difficult to remove due to their properties, such as being hydrophobic, volatile and persistent and biotoxic, and they have caused great harm to the ecological environment and human health. Electrochemical oxidation technology for the treatment of refractory pollutants has been widely used due to its high efficiency and easiness of operation. Thus, the electrochemical oxidation system was established for the efficient treatment of monochlorobenzene (MCB) waste gas. The effect of a single factor, such as anode materials, cathode materials, the electrolyte concentration, current density and electrode distance on the removal efficiency (RE) of MCB gas were first studied. The response-surface methodology (RSM) was used to investigate the relationships between different factors' conditions (current density, electrolyte concentration, electrode distance), and a prediction model was established using the Design-Expert 10.0.1 software to optimize the reaction conditions. The results of the one-factor experiments showed that when treating 2.90 g/m3 MCB gas with a 0.40 L/min flow rate, Ti/Ti4O7 as an anode, stainless steel wire mesh as a cathode, 0.15 mol/L NaCl electrolyte, 10.0 mA/cm2 current density and 4.0 cm electrode distance, the average removal efficiency (RE), efficiency capacity (EC) and energy consumption (Esp) were 57.99%, 20.18 g/(m3·h) and 190.2 (kW·h)/kg, respectively. The results of the RSM showed that the effects of the process parameters on the RE of MBC were as follows: current density > electrode distance > electrolyte concentration; the interactions effects on the RE of MBC were in the order of electrolyte concentration and current density > current density and electrode distance > electrolyte concentration and electrode distance; the optimal experimental conditions were as follows: the concentration of electrolyte was 0.149 mol/L, current density was 18.11 mA, electrode distance was 3.804 cm. Under these conditions, the RE achieved 66.43%. The response-surface variance analysis showed that the regression model reached a significant level, and the validation results were in agreement with the predicted results, which proved the feasibility of the model. The model can be applied to treat the CBS waste gas of polymer processes through electrochemical oxidation.

Investigation of the engineering and environmental properties of cement mortars incorporating ladle furnace steel slag

While the use of ladle steel slag in concrete offers several potential advantages, there are significant research gaps that need to be addressed to promote its safe and effective utilization in the construction industry, e.g. the variations in slag properties, such as chemical composition and physical characteristics, can influence concrete performance. Moreover, even there is some research on the mechanical properties of concrete containing ladle steel slag, more in-depth studies are needed to assess the long-term durability of such concrete and the research should focus more on evaluating the environmental impact of using ladle steel slag in concrete. The paper reports the fundamental findings of a research study focused on the treatment of ladle slag to be applied as a cement substitute in the manufacturing of cement products. The research aimed to determine the optimal proportion of admixture based on treated ladle slag and evaluate the environmental properties of the resulting cement composites. The treatment process involved mechanical activation, specifically grinding the ladle slag fraction of 0/8 mm using a vibrating mill (Coarse slag) and a ball mill (Fine slag). The treated ladle slag, as an admixture, was examined in conjunction with four types of cement (CEM I, CEM II, CEM III, and CEM V). The research findings revealed that the incorporation of coarse ladle slag enhanced the workability of fresh cement paste across all tested formulations. Moreover, the utilization of treated ladle slag as an admixture in cement paste (substituting cement) exhibited a retarding effect on the setting time. As the age of the test specimens progressed (90 to180 days), the strength properties of the samples with ladle slag admixture approached those of the cement composites prepared using natural aggregates. The climate change impact of cement mortars per strength unit spanned from 0.82 to 2.75 kg CO2eq/MPa for composites incorporating finely-ground slag, and from 0.9 to 3.1 kg CO2eq/MPa for composites utilizing coarse slag. The grinding of ladle slag, even up to 30 wt% cement replacement, did not significantly impact the overall global warming potential and climate change, considering the average mill energy consumption of 50 kWh per 1 t of material.

Service Function Chain Deployment Algorithm Based on Deep Reinforcement Learning in Space–Air–Ground Integrated Network

SAGIN is formed by the fusion of ground networks and aircraft networks. It breaks through the limitation of communication, which cannot cover the whole world, bringing new opportunities for network communication in remote areas. However, many heterogeneous devices in SAGIN pose significant challenges in terms of end-to-end resource management, and the limited regional heterogeneous resources also threaten the QoS for users. In this regard, this paper proposes a hierarchical resource management structure for SAGIN, named SAGIN-MEC, based on a SDN, NFV, and MEC, aiming to facilitate the systematic management of heterogeneous network resources. Furthermore, to minimize the operator deployment costs while ensuring the QoS, this paper formulates a resource scheduling optimization model tailored to SAGIN scenarios to minimize energy consumption. Additionally, we propose a deployment algorithm, named DRL-G, which is based on heuristics and DRL, aiming to allocate heterogeneous network resources within SAGIN effectively. Experimental results showed that SAGIN-MEC can reduce the end-to-end delay by 6–15 ms compared to the terrestrial edge network, and compared to other algorithms, the DRL-G algorithm can improve the service request reception rate by up to 20%. In terms of energy consumption, it reduces the average energy consumption by 4.4% compared to the PG algorithm.

Unravelling the Proportion of Food Expenditure and Energy Consumption on Food Security: Evidence from Smallholders’ Clove Farmers in Indonesia

Objective: The purpose of this study is to examine the proportion of food consumption expenditure to clove farmer household expenditures, as well as the energy and protein consumption and food security conditions of clove farmer households in Indonesia. Theoretical Framework: Hunger eradication and food security remain critical development challenges, particularly in developing nations such as Indonesia. Poorer households spend a large proportion of their income on basic needs, including food. However, as their income increases, they divert more of their income towards buying higher quality goods and services, thereby reducing the proportion that goes into food. Method: This research was conducted on 30 clove farming communities using a purposive sampling technique. In in-depth interviews, all informants responded to the researcher's questions using precise, in-depth, and easy-to-understand phrases. Data analysis was carried out by reducing data, presenting data and drawing conclusions and verification. Result and Conclusion: The results showed that the average proportion of non-food expenditure to total household expenditure of clove farmers in the study locations was 54.45%, while the proportion of food consumption expenditure to total expenditure was 45.55%. This suggests that non-food consumption spending accounts for a significant portion of farmer household expenditure. Meanwhile, the average energy and protein consumption of rice farming households was 1995.81 kcal/person/day and 88.65 grams/person/day, respectively. Implication Of Research: These findings imply that clove farmers have a sufficient amount of energy and protein. It may be stated that clove farmers in the research area have very excellent food security, with no food insecure households.

Medley deep reinforcement learning-based workload offloading and cache placement decision in UAV-enabled MEC networks

Internet of Things devices generate a large number of heterogeneous workloads in real-time that require specific application to tackle, and the inability to communicate between devices and communication base stations due to complex scenarios is a thorny issue. Service caching play a key role in managing specific-request workload from devices, and unmanned aerial vehicles with computation and communication functions can effectively solve communication barrier between devices and ground base stations. In addition, the joint optimization of workload offloading and service cache placement is a key issue. Accordingly, we design an unmanned aerial vehicle-enabled mobile edge computing system with multiple devices, unmanned aerial vehicles and edge servers. The proposed framework takes into account the randomness of workload arrival, the time-varying nature of channel states, the limitations of the hosting service caching, and wireless communication blocking. Furthermore, we designed workload offloading and service caching hosting decision-making optimization problems to minimize the long-term weighted average latency and energy consumption costs. To tackle this joint optimization problem, we propose a request-specific workload offloading and service caching decision-making scheme based on the medley deep reinforcement learning scheme. To this end, the proposed scheme is decomposed into two-stage optimization subproblems: the workload offloading decision-making problem and the service caching hosting selection problem. In terms of the first subproblem, we model each device as a learning agent and propose the workloads offloading decision-making scheme based on multi-agent deep deterministic policy gradient. For the second subproblem, we present the decentralized double deep Q-learning scheme to tackle the service caching hosting policy. According to the comprehensive experimental results, the proposed scheme is able to converge rapidly on various parameter configurations and whose performance surpasses the other four baseline learning algorithms.