Double-Shift: A Low-Power DNN Weights Storage and Access Framework based on Approximate Decomposition and Quantization

One major challenge in deploying Deep Neural Network (DNN) in resource-constrained applications, such as edge nodes, mobile embedded systems, and IoT devices, is its high energy cost. The emerging approximate computing methodology can effectively reduce the energy consumption during the computing process in DNN. However, a recent study shows that the weight storage and access operations can dominate DNN's energy consumption due to the fact that the huge size of DNN weights must be stored in the high-energy-cost DRAM. In this paper, we propose Double-Shift, a low-power DNN weight storage and access framework, to solve this problem. Enabled by approximate decomposition and quantization, Double-Shift can reduce the data size of the weights effectively. By designing a novel weight storage allocation strategy, Double-Shift can boost the energy efficiency by trading the energy consuming weight storage and access operations for low-energy-cost computations. Our experimental results show that Double-Shift can reduce DNN weights to 3.96%–6.38% of the original size and achieve an energy saving of 86.47%–93.62%, while introducing a DNN classification error within 2%.

A machine learning approach for predictive warehouse design

Warehouse management systems (WMS) track warehousing and picking operations, generating a huge volumes of data quantified in millions to billions of records. Logistic operators incur significant costs to maintain these IT systems, without actively mining the collected data to monitor their business processes, smooth the warehousing flows, and support the strategic decisions. This study explores the impact of tracing data beyond the simple traceability purpose. We aim at supporting the strategic design of a warehousing system by training classifiers that can predict the storage technology (ST), the material handling system (MHS), the storage allocation strategy (SAS), and the picking policy (PP) of a storage system. We introduce the definition of a learning table, whose attributes are benchmarking metrics applicable to any storage system. Then, we investigate how the availability of data in the warehouse management system (i.e. varying the number of attributes of the learning table) affects the accuracy of the predictions. To validate the approach, we illustrate a generalisable case study which collects data from sixteen different real companies belonging to different industrial sectors (automotive, manufacturing, food and beverage, cosmetics and publishing) and different players (distribution centres and third-party logistic providers). The benchmarking metrics are applied and used to generate learning tables with varying number of attributes. A bunch of classifiers is used to identify the crucial input data attributes in the prediction of ST, MHS, SAS, and PP. The managerial relevance of the data-driven methodology for warehouse design is showcased for 3PL providers experiencing a fast rotation of the SKUs stored in their storage systems.

Energy Storage Configuration Optimization Strategy for Islanded Microgrid Interconnection Based on Energy Consumption Characteristics

As a result of distributed energy development, the demand for energy storage grows more rapidly. The optimization of energy storage allocation is urgently needed. The economic benefits and characteristics of storage allocation have been thoroughly explored, and a mathematical model of economy was established. The centralized configuration of energy storage can make the best use of surplus electricity and reduce charging loss. The peak cutting and valley filling effect is quite obvious. Furthermore, some allocation strategies of storage were proposed. Based on the data of a user-side transformer and photovoltaic power generation, the cost under different strategies and the configuration scheme with the lowest cost is calculated. The example was used to analyze and compare the benefits under various strategies, and the results show it works in cost-saving. The strategy of similarity storage allocation has a positive effect on energy and cost-saving.

Dynamic Picking and Storage Optimization of Robotic Picking Systems

With the development of e-commerce and the improvement of logistics requirements, more and more ‘parts-to-picker’ picking systems begin to replace the inefficient ‘picker-to-parts’ picking systems in various scenarios. As the mainstream ‘parts-to-picker’ system, the robotic mobile fulfillment system has been attracting much attention in recent years. In addition to the customer's changing requirements, the rapid response of the picking system to the order is particularly important. In the above context, to seek a breakthrough in the picking system's picking efficiency without increasing the cost of additional equipment, the storage allocation of the pods becomes very important. This article focuses on the dynamic storage allocation of robotic mobile fulfillment system, which has positive theoretical and practical significance. By analyzing the pod storage process of the robotic mobile fulfillment system, a dynamic pod storage allocation model suitable for the robotic mobile fulfillment system is established with the goal of minimizing the pod handling distance. Two dynamic pod storage allocation strategies are proposed for the system. By simulating the picking systems of different scales, the effectiveness of the dynamic storage allocation strategy is verified, which has a certain reference to the operation of the robotic mobile fulfillment system in practice.

Derived Storage Allocation Rules to Meet Shared Water Demands for Optimal Operation of an Inter-basin Multi-reservoir Water Supply System

The priority principle of storage allocation rules of serial cascade reservoirs within an inter-basin water supply can reduce water loss and reduce water supply times. Reasonable balancing curves for reservoirs in parallel are proposed and the proportional distribution of water is determined to illustrate the optimal allocation rule for different scheduling periods of reservoirs. The mutation point and slope are used to describe the segmentation of reservoirs in parallel. In addition, the optimization model is established with the objective function to minimize times of water shortage while the particle swarm optimization algorithm based on the immune evolutionary algorithm is applied to calibrate the balancing curves. Finally, the relative optimal water supply rule is obtained, providing a larger water supply capacity and higher storage synchronization of member reservoirs. The reservoir groups downstream of Luan River are used as an example, with the results showing that the suggested method can effectively improve the operational performance and meet shared water demands in an inter-basin multi-reservoir. This article highlights the superior results obtained compared to the current storage allocation rules to meet shared water demands.

Improving Order-picking Operations with Precedence Constraints through Efficient Storage Location Assignment

This paper is inspired by a manual picking retail company where shape and weight constraints affect the order-picking process. We proposed an alternative clustering similarity index that considers the similarity, the weight and the shape of products. This similarity index was further incorporated in a storage allocation heuristic procedure to set the location of the products. We test the procedure in a retail company that supplies over 191 stores, in Northern Portugal. When comparing the strategy currently used in the company with this procedure, we found out that our approach enabled a reduction of up to 40% on the picking distance; a percentage of improvement that is 32% higher than the one achieved by applying the Jaccard index, a similarity index commonly used in the literature. This allows warehouses to save time and work faster.