arXiv subjects

Kaishun Wu

Publications and source records attributed to Kaishun Wu.

48 records · Page 3Linked to original sources

arXiv2024

On Designing Multi-UAV aided Wireless Powered Dynamic Communication via Hierarchical Deep Reinforcement Learning

This paper proposes a novel design on the wireless powered communication network (WPCN) in dynamic environments under the assistance of multiple unmanned aerial vehicles (UAVs). Unlike the existing studies, where the low-power wireless nodes (WNs) often conform to the coherent harvest-then-transmit protocol, under our newly proposed double-threshold based WN type updating rule, each WN can dynamically and repeatedly update its WN type as an E-node for non-linear energy harvesting over time slots or an I-node for transmitting data over sub-slots. To maximize the total transmission data size of all the WNs over T slots, each of the UAVs individually determines its trajectory and binary wireless energy transmission (WET) decisions over times slots and its binary wireless data collection (WDC) decisions over sub-slots, under the constraints of each UAV's limited on-board energy and each WN's node type updating rule. However, due to the UAVs' tightly-coupled trajectories with their WET and WDC decisions, as well as each WN's time-varying battery energy, this problem is difficult to solve optimally. We then propose a new multi-agent based hierarchical deep reinforcement learning (MAHDRL) framework with two tiers to solve the problem efficiently, where the soft actor critic (SAC) policy is designed in tier-1 to determine each UAV's continuous trajectory and binary WET decision over time slots, and the deep-Q learning (DQN) policy is designed in tier-2 to determine each UAV's binary WDC decisions over sub-slots under the given UAV trajectory from tier-1. Both of the SAC policy and the DQN policy are executed distributively at each UAV. Finally, extensive simulation results are provided to validate the outweighed performance of the proposed MAHDRL approach over various state-of-the-art benchmarks.

cs.NI↗

arXiv2023

Self-Convinced Prompting: Few-Shot Question Answering with Repeated Introspection

While large language models (LLMs) such as ChatGPT and PaLM have demonstrated remarkable performance in various language understanding and generation tasks, their capabilities in complex reasoning and intricate knowledge utilization still fall short of human-level proficiency. Recent studies have established the effectiveness of prompts in steering LLMs towards generating desired outputs. Building on these insights, we introduce a novel framework that harnesses the potential of large-scale pre-trained language models, to iteratively enhance performance of the LLMs. Our framework incorporates three components: \textit{Normal CoT}, a \textit{Convincer}, and an \textit{Answerer}. It processes the output of a typical few-shot chain-of-thought prompt, assesses the correctness of the response, scrutinizes the answer, refines the reasoning, and ultimately produces a new solution. Experimental results on the 7 datasets of miscellaneous problems validate the efficacy of the Self-Convince framework, achieving substantial improvements compared to the baselines. This study contributes to the burgeoning body of research focused on integrating pre-trained language models with tailored prompts and iterative refinement processes to augment their performance in complex tasks.

cs.CL↗

arXiv2023

Multi-Agent Graph Reinforcement Learning based On-Demand Wireless Energy Transfer in Multi-UAV-aided IoT Network

This paper proposes a new on-demand wireless energy transfer (WET) scheme of multiple unmanned aerial vehicles (UAVs). Unlike the existing studies that simply pursuing the total or the minimum harvested energy maximization at the Internet of Things (IoT) devices, where the IoT devices' own energy requirements are barely considered, we propose a new metric called the hungry-level of energy (HoE), which reflects the time-varying energy demand of each IoT device based on the energy gap between its required energy and the harvested energy from the UAVs. With the purpose to minimize the overall HoE of the IoT devices whose energy requirements are not satisfied, we optimally determine all the UAVs' trajectories and WET decisions over time, under the practical mobility and energy constraints of the UAVs. Although the proposed problem is of high complexity to solve, by excavating the UAVs' self-attentions for their collaborative WET, we propose the multiagent graph reinforcement learning (MAGRL) based approach. Through the offline training of the MAGRL model, where the global training at the central controller guides the local training at each UAV agent, each UAV then distributively determines its trajectory and WET based on the well-trained local neural networks. Simulation results show that the proposed MAGRL-based approach outperforms various benchmarks for meeting the IoT devices' energy requirements.

cs.IT↗

arXiv2023

Task-Oriented Integrated Sensing, Computation and Communication for Wireless Edge AI

With the advent of emerging IoT applications such as autonomous driving, digital-twin and metaverse etc. featuring massive data sensing, analyzing and inference as well critical latency in beyond 5G (B5G) networks, edge artificial intelligence (AI) has been proposed to provide high-performance computation of a conventional cloud down to the network edge. Recently, convergence of wireless sensing, computation and communication (SC${}^2$) for specific edge AI tasks, has aroused paradigm shift by enabling (partial) sharing of the radio-frequency (RF) transceivers and information processing pipelines among these three fundamental functionalities of IoT. However, most existing design frameworks separate these designs incurring unnecessary signaling overhead and waste of energy, and it is therefore of paramount importance to advance fully integrated sensing, computation and communication (ISCC) to achieve ultra-reliable and low-latency edge intelligence acquisition. In this article, we provide an overview of principles of enabling ISCC technologies followed by two concrete use cases of edge AI tasks demonstrating the advantage of task-oriented ISCC, and pointed out some practical challenges in edge AI design with advanced ISCC solutions.

cs.IT↗

arXiv2022

MMLN: Leveraging Domain Knowledge for Multimodal Diagnosis

Recent studies show that deep learning models achieve good performance on medical imaging tasks such as diagnosis prediction. Among the models, multimodality has been an emerging trend, integrating different forms of data such as chest X-ray (CXR) images and electronic medical records (EMRs). However, most existing methods incorporate them in a model-free manner, which lacks theoretical support and ignores the intrinsic relations between different data sources. To address this problem, we propose a knowledge-driven and data-driven framework for lung disease diagnosis. By incorporating domain knowledge, machine learning models can reduce the dependence on labeled data and improve interpretability. We formulate diagnosis rules according to authoritative clinical medicine guidelines and learn the weights of rules from text data. Finally, a multimodal fusion consisting of text and image data is designed to infer the marginal probability of lung disease. We conduct experiments on a real-world dataset collected from a hospital. The results show that the proposed method outperforms the state-of-the-art multimodal baselines in terms of accuracy and interpretability.

cs.LG↗

arXiv2021

Energy-Efficient UAV Multicasting with Simultaneous FSO Backhaul and Power Transfer

This letter studies an unmanned aerial vehicle (UAV) aided multicasting (MC) system, which is enabled by simultaneous free space optics (FSO) backhaul and power transfer. The UAV applies the power-splitting technique to harvest wireless power and decode backhaul information simultaneously over the FSO link, while at the same time using the harvested power to multicast the backhauled information over the radio frequency (RF) links to multiple ground users (GUs). We derive the UAV's achievable MC rate under the Poisson point process (PPP) based GU distribution. By jointly designing the FSO and RF links and the UAV altitude, we maximize the system-level energy efficiency (EE), which can be equivalently expressed as the ratio of the UAV's MC rate over the optics base station (OBS) transmit power, subject to the UAV's sustainable operation and reliable backhauling constraints. Due to the non-convexity of this problem, we propose suboptimal solutions with low complexity. Numerical results show the close-to-optimal EE performance by properly balancing the power-rate tradeoff between the FSO power and the MC data transmissions.

cs.IT↗

arXiv2020

Artificial Intelligence-Enabled Intelligent 6G Networks

With the rapid development of smart terminals and infrastructures, as well as diversified applications (e.g., virtual and augmented reality, remote surgery and holographic projection) with colorful requirements, current networks (e.g., 4G and upcoming 5G networks) may not be able to completely meet quickly rising traffic demands. Accordingly, efforts from both industry and academia have already been put to the research on 6G networks. Recently, artificial intelligence (AI) has been utilized as a new paradigm for the design and optimization of 6G networks with a high level of intelligence. Therefore, this article proposes an AI-enabled intelligent architecture for 6G networks to realize knowledge discovery, smart resource management, automatic network adjustment and intelligent service provisioning, where the architecture is divided into four layers: intelligent sensing layer, data mining and analytics layer, intelligent control layer and smart application layer. We then review and discuss the applications of AI techniques for 6G networks and elaborate how to employ the AI techniques to efficiently and effectively optimize the network performance, including AI-empowered mobile edge computing, intelligent mobility and handover management, and smart spectrum management. Moreover, we highlight important future research directions and potential solutions for AI-enabled intelligent 6G networks, including computation efficiency, algorithms robustness, hardware development and energy management.

cs.NI↗

arXiv2019

Faster and Safer Training by Embedding High-Level Knowledge into Deep Reinforcement Learning

Deep reinforcement learning has been successfully used in many dynamic decision making domains, especially those with very large state spaces. However, it is also well-known that deep reinforcement learning can be very slow and resource intensive. The resulting system is often brittle and difficult to explain. In this paper, we attempt to address some of these problems by proposing a framework of Rule-interposing Learning (RIL) that embeds high level rules into the deep reinforcement learning. With some good rules, this framework not only can accelerate the learning process, but also keep it away from catastrophic explorations, thus making the system relatively stable even during the very early stage of training. Moreover, given the rules are high level and easy to interpret, they can be easily maintained, updated and shared with other similar tasks.

cs.AI↗

arXiv2019

Enhanced Energy-Efficient Downlink Resource Allocation in Green Non-Orthogonal Multiple Access Systems

Despite numerous advantages, non-orthogonal multiple access (NOMA) technique can bring additional interference for the neighboring ultra-dense networks if the power consumption of the system is not properly optimized. While targeting on the green communication concept, in this paper, we propose an energy-efficient downlink resource allocation scheme for a NOMA-equipped cellular network. The objective of this work is to allocate subchannels and power of the base station among the users so that the overall energy efficiency is maximized. Since this problem is NP-hard, we attempt to find an elegant solution with reasonable complexity that provides good performance for some realistic applications. To this end, we decompose the problem into a subchannel allocation subproblem followed by a power loading subproblem that allocates power to each user's data stream on each of its allocated subchannels. We first employ a many-to-many matching model under the assumption of uniform power loading in order to obtain the solution of the first subproblem with reasonable performance. Once the the subchannel-user mapping information is known from the first solution, we propose a geometric programming (GP)-based power loading scheme upon approximating the energy efficiency of the system by a ratio of two posynomials. The techniques adopted for these subproblems better exploit the available multi-user diversity compared to the techniques used in an earlier work. Having observed the computational overhead of the GP-based power loading scheme, we also propose a suboptimal computationally-efficient algorithm for the power loading subproblem with a polynomial time complexity that provides reasonably good performance. Extensive simulation has been conducted to verify that our proposed solution schemes always outperform the existing work while consuming much less power at the base station.

cs.NI↗

arXiv2019

Detecting and Diagnosing Energy Issues for Mobile Applications

Energy efficiency is an important criterion to judge the quality of mobile apps, but one third of our randomly sampled apps suffer from energy issues that can quickly drain battery power. To understand these issues, we conducted an empirical study on 27 well-maintained apps such as Chrome and Firefox, whose issue tracking systems are publicly accessible. Our study revealed that the main root causes of energy issues include unnecessary workload and excessively frequent operations. Surprisingly, these issues are beyond the application of present technology on energy issue detection. We also found that 20.6% of energy issues can only manifest themselves under specific contexts such as poor network performance, but such contexts are again neglected by present technology. Therefore, we proposed a novel testing framework for detecting energy issues in real-world apps. Our framework examines apps with well-designed input sequences and runtime contexts. To identify the root causes mentioned above, we employed a machine learning algorithm to cluster the workloads and further evaluate their necessity. For the issues concealed by the specific contexts, we carefully set up several execution contexts to pinpoint them. More importantly, we developed leading edge technology, e.g. pre-designing input sequences with potential energy overuse and tuning tests on-the-fly, to achieve high efficacy in detecting energy issues. A large-scale evaluation shows that 91.6% issues detected in our test were previously unknown to developers. On average, these issues double the energy costs of the apps. Furthermore, our test achieves a low number of false positives. Finally, we show how our test reports can help developers fix the issues.

cs.SE↗

arXiv2018

Throughput Maximization for Laser-Powered UAV Wireless Communication Systems

Laser power has become a viable solution to provide convenient and sustainable energy supply to unmanned aerial vehicles (UAVs). In this paper, we study a laser-powered UAV wireless communication system, where a laser transmitter sends laser beams to charge a fixed-wing UAV in flight, and the UAV uses the harvested laser energy to communicate with a ground station. To maintain the UAV's sustainable operation, its total energy consumption cannot exceed that harvested from the laser transmitter. Under such a laser energy harvesting constraint, we maximize the downlink communication throughput from the UAV to the ground station over a finite time duration, by jointly optimizing the UAV's trajectory and its transmit power allocation. However, due to the complicated UAV energy consumption model, this problem is non-convex and difficult to be solved. To tackle the problem, we first consider a special case with a double-circular UAV trajectory which balances the tradeoff between maximizing the performance of laser energy harvesting versus wireless communication at the UAV. Next, based on the obtained double-circular trajectory, we propose an efficient solution to the general problem, by applying the techniques of alternating optimization and sequential convex programming (SCP). Finally, numerical results are provided to validate the communication throughput performance of the proposed design.

cs.IT↗

arXiv2017

Enhanced Uplink Resource Allocation in Non-Orthogonal Multiple Access Systems

Non-orthogonal multiple access (NOMA) is envisioned to be one of the most beneficial technologies for next generation wireless networks due to its enhanced performance compared to other conventional radio access techniques. Although the principle of NOMA allows multiple users to use the same frequency resource, due to decoding complication, information of users in practical systems cannot be decoded successfully if many of them use the same channel. Consequently, assigned spectrum of a system needs to be split into multiple subchannels in order to multiplex that among many users. Uplink resource allocation for such systems is more complicated compared to the downlink ones due to the individual users' power constraints and discrete nature of subchannel assignment. In this paper, we propose an uplink subchannel and power allocation scheme for such systems. Due to the NP-hard and non-convex nature of the problem, the complete solution, that optimizes both subchannel assignment and power allocation jointly, is intractable. Consequently, we solve the problem in two steps. First, based on the assumption that the maximal power level of a user is subdivided equally among its allocated subchannels, we apply many-to-many matching model to solve the subchannel-user mapping problem. Then, in order to enhance the performance of the system further, we apply iterative water-filling and geometric programming two power allocation techniques to allocate power in each allocated subchannel-user slot optimally. Extensive simulation has been conducted to verify the effectiveness of the proposed scheme. The results demonstrate that the proposed scheme always outperforms all existing works in this context under all possible scenarios.

cs.IT↗