The increasing complexity of Beyond 5G and 6G networks necessitates new paradigms for autonomy and assur- ance. Traditional O-RAN control loops rely heavily on RIC- based orchestration, which centralizes intelligence and exposes the system to risks such as policy conflicts, data drift, and unsafe actions under unforeseen conditions. In this work, we argue that the future of autonomous networks lies in a multi-agentic architecture, where specialized agents collaborate to perform data collection, model training, prediction, policy generation, verification, deployment, and assurance. By replacing tightly- coupled centralized RIC-based workflows with distributed agents, the framework achieves autonomy, resilience, explainability, and system-wide safety. To substantiate this vision, we design and evaluate a traffic steering use case under surge and drift conditions. Results across four KPIs: RRC connected users, IP throughput, PRB utilization, and SINR, demonstrate that a naive predictor-driven deployment improves local KPIs but destabilizes neighbors, whereas the agentic system blocks unsafe policies, preserving global network health. This study highlights multi- agent architectures as a credible foundation for trustworthy AI- driven autonomy in next-generation RANs.

The rapidly expanding landscape of Internet of Things (IoT) applications, particularly immersive and real-time services, demands a new paradigm for network management. Traditional manual network configuration processes are inadequate to meet the ultra-low latency and high-throughput requirements of these applications. This paper presents an AI-driven Intent-based Networking (IBN) solution that operates directly on the client device, reframing a novel pluggable L4S (pL4S) architecture as a practical IBN system to achieve low-latency objectives. The proposed system, implemented on Samsung smartphones, comprises three key components: a Real-time Traffic Detector (RTTD) that translates implicit user intent into network policies; an Extended Berkeley Packet Filter (eBPF)-based pL4S module for dynamic kernel-agnostic policy enforcement; and a proactive L4S module that uses machine learning to ensure intent is met by predicting congestion and performing early marking. Through comprehensive evaluations in both simulated and live network environments, the system demonstrates significant improvements in latency and queue management over traditional methods. Our results confirm that this client-side IBN approach effectively addresses the challenges of slow kernel adoption and reactive congestion control, providing a viable and immediately deployable solution for pioneering low-latency technologies in the IoT ecosystem.

The escalating demand for real-time applications such as cloud gaming and virtual reality presents unique challenges in managing network latency and throughput simultaneously. Traditional congestion control mechanisms, which primarily rely on packet loss as a signal, are inadequate for modern applications requiring high throughput and ultra-low latency. This paper introduces a pluggable implementation of the Low Latency, Low Loss, Scalable throughput (L4S) architecture using eBPF (extended Berkeley Packet Filter) programs to enhance the Explicit Congestion Notification (ECN) mechanism. We specifically focus on replicating the L4S client functionality at the receiver end, which is crucial for the Accurate ECN (AccECN) protocol. By leveraging sched_cls eBPF programs, our design captures and manipulates packet flows at both ingress and egress points, enabling precise congestion feedback and efficient protocol negotiation directly from the data plane. Our approach ensures compatibility with existing network infrastructures and extends L4S benefits to devices operating on legacy kernels. The paper evaluates the implications of this design on network performance, particularly its ability to reduce latency and handle high throughput demands seamlessly. Through this innovation, we aim to accelerate the adoption of L4S across varied network environments, enhancing the quality of experience for latency-sensitive applications without extensive system overhauls.

In this paper, we propose SurroundSense, a novel event-driven framework that generates contextual information based on the user's surroundings using the Impulse Radio Ultra-Wideband (IR-UWB) Radar integrated into smartphones. By leveraging the channel impulse response (CIR) data obtained from the IR-UWB radar, SurroundSense constructs a detailed contextual understanding of the user's environment, allowing for tailored recommendations based on the surrounding conditions. The system processes the CIR data and combines it with acoustic sensing information, specifically noise levels and ambient light information to generate the surrounding information. Further-more, utilizing the IMU sensors enables an expanded field-of-view (FOV), providing a comprehensive$360^o$environmental awareness. This enhanced surrounding information enhances the accuracy and pertinence of personalized recommendations. Our approach showcases the potential of incorporating IR-UWB technology into personalization algorithms, ultimately improving user experiences and facilitating adaptive, context-aware applications.

Rapid advancement of radar technology has led to the emergence of cognitive radar systems, which utilize adaptive mechanisms to optimize performance in dynamic environments. This paper explores the integration of cognitive adaptability into smartphone-based Impulse Radio Ultra-Wideband (IRUWB) radar systems. By dynamically modifying the radar’s operational parameters based on real-time output analysis, we aim to address the limitations of current smartphone radar implementations, including high power consumption, static radar configurations, and the inherent mobility of smartphones. Our proposed Cognitive-Adaptive IR-UWB Radar (CAIR) system improves accuracy, power efficiency, and responsiveness, enabling effective target detection, target classification, gesture recognition, distance estimation, and vital sign monitoring in diverse scenarios. By incorporating cognitive radar principles, we present a novel approach to overcoming the challenges of varying environmental conditions and user contexts, ultimately delivering a more robust and versatile user experience. This paper outlines the CAIR architecture, algorithmic design, and adaptive control mechanisms, showcasing its potential to enhance smartphone radar sensing. When tested against the major smartphone use cases, our system improves accuracy by up to 11.5%, while achieving cognitive adaptability of up to 90%. Additionally, the Artificial Neural Network (ANN)-based cognitive model achieves an accuracy of 95% and an F1-score of 94%.

Call muting-unexpected silences during voice calls due to extended RTP packet loss is a major challenge in high-mobility 5G environments, severely degrading Mean Opinion Score (MOS) and user experience. We propose DRLCQ, a Deep Reinforcement Learning-based framework that dynamically tunes Cell Individual Offset (CIO) in real time to reduce mute events and enhance voice quality. Integrated as an xApp within the O-RAN Near-RT RIC, DRLCQ leverages live network KPIs (e.g., SINR, jitter, packet loss) to learn optimal handover decisions. Evaluated against static and heuristic baselines, DRLCQ achieves over 20% fewer call mute incidents and up to 85% higher MOS, demonstrating a scalable and intelligent solution for AI-native RAN control.

This paper introduces the Next-Generation Mobile Hotspot (NGMHS), a breakthrough solution for optimizing 5G mobile hotspot performance through intelligent, AI-powered traffic management and user-focused controls. Leveraging extended Berkeley Packet Filter (eBPF) technology, NGMHS efficiently monitors per-client traffic, reducing processing overhead while providing precise real-time control. Central to NGMHS is the AI-based Network Service Detector (NSD+), which dynamically prioritizes real-time traffic, significantly enhancing video call bitrates and reducing latency for gaming applications. Our evaluations on the Samsung A54 and Galaxy S24 devices demonstrate marked improvements in user experience, with innovations such as privacy-focused OTP user-profiles and granular traffic insights. NGMHS represents a pioneering step forward in 5G connectivity, offering a seamless, secure, and highly customizable mobile hotspot experience. NGMHS is successfully deployed across Samsung's A, M, S, Fold, and Flip series models, where increased user engagement and consistent performance improvements have been observed post-deployment.

This paper proposes a Generative AI-based Digital Twin (GDT) platform for automated network feature performance evaluation, designed for Beyond 5G (B5G) networks. The platform addresses the inefficiencies of manual evaluation by utilizing a conditional Generative Adversarial Network (cGAN) to simulate network performance based on historical data and new AI/ML features. The Network GDT integrates a novel Digital Twin Augmenting Condition (DTAC) framework, allowing for real-time simulation and performance evaluation of network features. This system significantly reduces the time and cost associated with manual evaluations, improves decision-making, and optimizes Quality of Service (QoS) and Quality of Experience (QoE). The cGAN-based model dynamically generates synthetic data, enabling comprehensive performance insights and proactive AI solution testing under various network scenarios. Experimental results demonstrate high prediction accuracy for congestion use case, validating the robustness of the proposed system. The platform's dual-phase strategy ensures that AI-based solutions are rigorously tested in simulated environments before deployment in real networks, minimizing risks and enhancing stability. This approach provides a scalable and efficient solution for future B5G networks, paving the way for more reliable and optimized wireless communication systems.

The surging popularity of mobile gaming has underscored the need for optimal performance in Real-Time Online Mobile Gaming (RT OMG). However, the concurrent Non-Real-Time (NRT) activities, such as downloads, can often compete with the RT OMG traffic, depriving the users of a seamless gaming experience. This paper presents a study on the drawbacks of using the app bitrate to quantify the RT OMG Quality of Experience (QoE). We propose the Game Stabilizer, a novel machine learning-driven solution to enhance the RT OMG experience. The Game Stabilizer effectively utilizes Wi-Fi network condition information and an End-to-End bandwidth estimate of the link to intelligently manage the NRT bandwidth, giving the users a consistent gameplay experience. Our proposed solution demonstrates promising results, notably reducing latency in the Smartphone gaming experience by up to 60%. This advancement holds significant promise for elevating the overall mobile gaming landscape.

In the rapidly advancing domain of 5G communication systems, channel decoding, particularly turbo decoding, has emerged as a significantly complex challenge. Turbo decoding is an essential element within communication frameworks, necessitating both efficiency and rapid processing to cater to the demanding data rates and stringent low latency requirements of 5G networks. This paper focuses on the unique contributions of employing a Sparse Recurrent Neural Network (SRNN) architecture, leveraging sparsity to reduce computational load while maintaining high performance significantly. Unlike existing approaches, our method introduces a novel piece-wise linear approximation of the activation function, enhancing efficiency and scalability for NextGen communication systems. Our approach leverages the principles of sparsity and employs a piece-wise linear approximation of the activation function to markedly reduce the computational load of the turbo-decoding process.Comprehensive evaluations demonstrate that our RNN architecture outperforms existing deep learning models in the context of turbo decoding and with a significantly lower computational footprint. This research contributes to the field by providing a scalable, efficient, and less computationally intensive turbo-decoding method, particularly suited for the next-generation cloud systems underlying 5G and beyond communication technologies.

In IEEE 802.11-based access networks (Wi-Fi), the client remains connected to a far-poor Access Point (AP) rather than switching to a near-better AP. This scenario is termed a sticky client problem. This scenario can severely impact the performance of real-time applications. Several standards, such as 802.11k1v/r, are being developed to enhance Wi-Fi roaming capabilities. However, the sticky client problem is yet to be solved completely. This paper proposes Sticky less, a novel method that leverages machine learning to learn the home Wi-Fi network behavior to address the sticky client problem. Initially, Stickyless divides the deployment area of APs into distinct zones, generates training data, and subsequently trains the machine learning module. The Stickyless employs clustering models to recommend selecting the optimal AP within a specific zone by considering the application performance and quality metrics. To conclude, Stickyless assesses performance using a proposed scoring and cascading module. We also developed a prototype to evaluate the Stickyless performance, outperforming the existing methods. The proposed method improves the Wi-Fi roaming experience by reducing stickiness up to 40 %. Thereby, it improves the link quality of the client by an average of 19 % and decreases the packet error rate by up to 3.5 % compared to the existing approaches. We also experimented with popular gaming apps, and Stickyless reduced the latency by up to 7 -fold.

Smartphone consumers use various applications (apps), including online gaming, chat, streaming, video calling, and social networking. The smartphone relies on the network backhaul, such as a Wi-Fi access point, to provide the required Quality of Service (QoS). In its send-and-receive queues, the smartphone processes the packets in a first-in-first-out (FIFO) fashion. Many people worldwide started using video calling apps daily during the pandemic and post-pandemic periods. On the other hand, online gaming apps skyrocketed and continue to engage people. When real-time (RT) video calling and gaming apps race with non-real-time (NRT) traffic, we found a severe degradation in the Quality of Experience (QoE). In this work, we propose an Application Prioritization Engine (APE) framework that will improve user experience by dynamically allocating bandwidth to the different apps in the smartphone. APE helps improve the end-user experience by detecting and prioritizing real-time traffic over concurrent best-effort traffic. We introduced an eBPF (extended Berkeley Packet filter) that can control the NRT traffic to the extent that it does not affect the RT traffic. We evaluate the performance of APE with the top-chart video calling and gaming apps in a live-air scenario. APE enhances video calling performance in poor network conditions by improving the bit rate to 110%. Furthermore, it provides a four-fold gaming latency reduction despite NRT traffic. APE is a tech-transferred, app-agnostic, server-independent solution enabled in the latest Samsung flagship Smartphones with Android 13 OS.

The Next Generation Networks (NGN) set its standard to provide very high data rates, Ultra-Reliable Low Latency Communications (URLLC), increased network capacity and significantly improved Quality of Service (QoS). Thus, it provides an infrastructure for the Internet of Things (IoT) to power billions of connected devices. With this upsurge in IoT devices, the need for adopting IPv6 over exhausting IPv4 addresses becomes more unavoidable. Hence, Internet Service Providers (ISP's) are adopting IPv6 along with IPv4 addresses using an address transition method called dual-stack. However, the Dual-Stack network causes relatively more connectivity overhead than the single stack network. For example, DNS lookup and TCP connection time are comparatively high in dual-stack mobile devices. This affects the page loading time of the application, thereby impacting the user experience significantly. In this paper, we analyzed all these connectivity overheads and propose a novel solution called NexGen Connectivity optimizer (NexGenCO), which provides better connectivity for the applications using network-aware concurrency and intelligent DNS caching. NexGenCO is prototyped and evaluated in Samsung devices with Android Pie. It significantly reduces connectivity overhead and improves page loading time up to 18% consistently.

To improve the performance of the Transmission Control Protocol (TCP), the Quick UDP Internet Connections (QUIC) was introduced. However, from the recent literature, in the wireless networks, QUIC does not fully utilize the link capacity, because of varying network wireless medium characteristics for an application. To the best of our knowledge, for the first time, this paper has proposed two novel protocols to overcome the wireless network's challenges, by MultiPipe (multi sessions) QUIC protocol (MP-QUIC). The two MP-QUIC's novel protocols are (i) Round Robin MP-QUIC (RR-MP-QUIC) and (ii) Cross-Layer Burst Aware MP-QUIC (CBA-MP-QUIC). In RR-MP-QUIC, multiple pipes are created between source as well as destination and assign a pipe to each object in a round-robin manner. We set up the testbed over live air commercial network. The experimental results reveal that the RR-MP-QUIC Page Load Time (PLT) of a web page decreases by 76% with respect to QUIC. Furthermore, to improve the performance of RR-MPQUIC, we proposed CBA-MP-QUIC, which optimally creates pipes, efficiently schedules the objects and understands with respect to cross-layer (wireless channel characteristics and load in data link layer) parameters. Hence this adapts dynamically to the network conditions and maintains fairness between the applications in User Equipment (UE) and within the same application. Experimental result of CBA-MP-QUIC on live air improves the web page PLT by 143% with respect to QUIC.

Multipath TCP (MPTCP) is an enhancement of TCP, capable of using multiple network paths to enhance the throughput and reliability. The current implementation of MPTCP does not consider wireless network characteristics. Unlike the wired networks, the path characteristic and mode of operations may vary among various wireless network interfaces dynamically. Hence we propose Smart Multipath Switch (SMS), dynamic MPTCP subflow management for the wireless network. SMS uses a learning-based approach, adapts to ad-hoc wireless conditions, thereby dynamically controls and manages the subflow in MPTCP for better user experience and network utilization. To demonstrate the effectiveness of our proposal, we performed live air experiments with the help of Samsung Galaxy S8 in different locations (Korea, Thailand, and India) and also performed simulations in our lab at Samsung Electronics, Headquarters. Our experiments show that the proposed solution provides the aggregation ratio consistently above 80%. Furthermore, the SMS, using auto-tuning logic, improves the throughput by up to 51.5% compared with legacy.

Requirements for Next Generation Mobile Networks (NGMN) include low latency, higher throughput, scalability, and energy efficiency. As 5G millimeter wave (mmWave) band is short-range, the handover is inevitable. Google proposed QUIC (Quick UDP Internet Connection), which aims to address these challenges. However, Google QUIC (GQUIC), follows “WiFi-First” policy causing frequent network switching, which can lead to a throughput reduction and fast battery degradation. In this paper, we propose Cross-layer QUIC (CQUIC) framework, that follows “WiFi-if-best” policy to enhance the throughput and resilience by using a Cross-Layer approach. CQUIC proposes a novel migration scheme in QUIC which adapts to the dynamic network characteristics. GQUIC protocol with low bandwidth and high round-trip-time fail to migrate for seamless User Experience. CQUIC algorithm predicts Cross-Layer Score (CLS) which incorporates predicted Signal-to-Interference Noise Ratio (SINR), QUIC Bandwidth, round-triptime (RTT) stats from QUIC Session and models the handover decision pro-actively. Compared with state-of-the-art methods such as GQUIC and HTTP (using TCP) this paper reveals the significant benefits of the proposed method. A series of experimental results obtained in live air network over Samsung Galaxy S10 devices show CQUIC outperforms the GQUIC by 20%, TCP by 36% and MPTCP (Backup) by 17% in terms of throughput. Furthermore, CQUIC compared with MPTCP, reduces the data consumption over mobile network and operates green by reducing the power consumption by 25%.

Exponential growth in the number of subscribers intrigued Mobile Network Operators(MNOs) to invest substantial efforts towards the faster transition to IPv6 address. The current solution deployed by leading MNOs uses Dual IP 464XLAT to achieve address translation. There has been significant research on the dual IP system. However, an important area of study which is not investigated in detail is the correlation of 464XLAT in the multi-core architecture. We investigate the effects of this architecture on bandwidth utilization of mobile Smartphones with emphasis on multi-core scheduling algorithm. In this paper, we propose a novel networking packet scheduling scheme called CAA - CLAT Aware Affinity Scheduler for Next Generation Mobile Networks. CAA classifies the packets according to the characteristics and efficiently schedules among the CPU cores at its best effort for improved throughput in Dual Stack Smartphones. We also propose CAA-LITE, a lightweight version where the affinity scheduling is static with minimal steps. To illustrate the effectiveness of our proposed method, we conducted simulations in our lab at Samsung R&D India Bangalore and live air experiments in Samsung Electronics, South Korea. Our live air experiments show that the CAA outperforms the legacy by improving the throughput of around 90% under various operational conditions consistently. Moreover, our power consumption test shows that the CAA improves the power by 22% compared to original approaches.

Over the past few years, Voice/Video over Wi-Fi (VoWiFi) technology has made great strides and several researches predict its rapid growth. Voice over LTE (VoLTE) guarantees the Quality of Service (QoS) by creating dedicated channels. However, the QoS of VoWiFi service is still an area to be considered since the non-dedicated Wi-Fi does not provide any Guaranteed Bit Rate (GBR) mechanism. When multiple applications are in use, there will be sharing of bandwidth amongst these applications. Hence, QoS might not be guaranteed. VoWiFi being a high priority user preferred service, sharing of bandwidth should not affect its performance. In this paper, we propose Dedicated VoWiFi (D-VoWiFi), a paradigm to efficiently allocate the required bandwidth to VoWiFi service using GBR scheduler. The GBR scheduler prioritizes VoWiFi and assures better received average framerate per second (fps) irrespective of the other applications' activity. We provide User Equipment (UE) only solution which makes it completely agnostic of the server or any middle box (such as router, P-GW, etc). We demonstrate the effectiveness of our model with the help of live air experiments (performed in Samsung Electronics) using Samsung Galaxy S8 and S8+ devices. From experiment and analysis, we conclude that our model outperforms the current methodology in terms of QoS. We are able to achieve averaged freeze rate reduction by 23% while running single parallel download and up to 375% while multiple parallel downloads are in progress.

At present, most of the research work going around focuses on evolution of Next Generation Networks (NGN). The primary focus of this research work is towards improving the performance by reducing latency in network, increasing peak throughput and improving spectral efficiencies. Even though Fifth Generation (5G) network standards set its requirement to lower the latency, the Internet Protocol (IP) suite introduces significant delay in the network. For example, Domain Name System (DNS) resolution in a device takes at least one Round Trip Time (RTT) irrespective of the network infrastructure. Sometimes, slow responsiveness of DNS server triggers client to send multiple queries, which results in user perceived delay in the client application. Moreover, DNS resolution delays socket set-up time and creates connectivity overhead to the client application and also affects user experience significantly. We considered this limitation and propose a novel solution called Flare-DNS Resolver (FDR). It is a lightweight and client only solution which can be easily deployed across all mobile platforms. We successfully implemented FDR in Samsung flagship models having Android Oreo Operating System. The recent version of FDR is tested in both Samsung Galaxy S8 and S9 variants. FDR significantly improves application page loading time by 10% to 15% consistently.

Mobile Hotspots are very popular as they use 4G-LTE (via eNodeB to the network) to provide connectivity to multiple User Equipments (UEs). Additionally, In order to improve the performance of a session from a UE which is directly connected to the eNodeB, literature has proposed Multi-Pathing (MP) concepts (Multi Path TCP, Multi Path QUIC). However, these MP concepts do not work for a UE connected to eNodeB through the hotspot. Moreover, to improve the performance of hotspot, there can be MP connections from mobile hotspot itself. First one from hotspot to LTE eNodeB (Licensed spectrum) and second to WiFi (unlicensed) Access Point (AP). To the best of our knowledge, in available literature, a hotspot device does not provide MP aggregation of licensed and unlicensed spectrums to the connected UEs. Secondly, traffic from the hotspot device majorly goes over LTE hence congesting the licensed spectrum. With respect to the above two challenges, we proposed a novel hotspot architecture called Next-Gen Mobile HotSpot (NextGen-MHS), which is a unique approach to aggregate any combination of available licensed (4G/5G) with unlicensed (WiFi/5G) spectrum and provide the aggregated connection as a Mobile-Hotspot AP. We have successfully implemented this new architecture of NextGen-MHS and conducted evaluation experiments with 4G (licensed) and WiFi (unlicensed). NextGen-MHS significantly provides better results with 80% aggregated throughput and 45% reduction in data traffic over licensed spectrum.

In the past few decades, many Transmission Control Protocol (TCP) congestion control algorithms have been investigated to meet the growing network demands and to enhance the performance of TCP in lossy or high-bandwidth-delay-product (high-BDP) networks. However, it is still challenging to implement a dynamic congestion control algorithm for wide range of diverse mobile users, network conditions and applications. This paper explores avenues for enhancement of TCP congestion control algorithm for next generation mobile networks by dynamically learning the available bandwidth and deriving the Congestion Control Factor N. N is used to Adaptive Increase/Adaptive Decrease (AIAD) the Congestion Window (CWND) dynamically instead of using the traditional approach that is Additive Increase/ Multiplicative Decrease (AIMD) paradigm. Once there is congestion, our proposed algorithm will not allow the CWND to decrease multiplicatively or steeply. After dropping to a certain level (lower than legacy), we try to take the CWND to previous state adaptively with the help of calculated bandwidth (based on learning). This in turn helps to efficiently control the CWND for better network utilization especially in case of lossy and high-BDP conditions. As soon as it reaches the original state, it remains stable for longer time as compared to legacy until packet loss or time out. We demonstrate the effectiveness of our algorithm with the help of live air experiments (performed in Samsung R&D India, Bangalore) and NS3 based simulation experiments. Through our experiments, we show that our algorithm outperforms the legacy congestion control algorithms (like CUBIC, RENO, TCPW) and the existing CLTCP algorithm in terms of goodput, intra algorithm fairness and inter algorithm fairness maintaining the scalability and friendliness.

Voice over WiFi (VoWiFi) is a complementary technology to Voice over Long Term Evolution (VoLTE) which utilizes IP Multimedia Subsystem (IMS) to provide a packet-switched voice service over WiFi network. VoWiFi extends the benefit of VoLTE sans cellular coverage. In the existing VoWiFi architecture, IMS sessions can be seamlessly handed over from LTE to WiFi (L2W) and vice versa (W2L). A major limitation of this approach is that direct handover from one VoWiFi AP (Access Point) to another VoWiFi AP (W2W) is not feasible. This results in disruption of IMS services such as call drop at each AP transition in areas with limited or no cellular coverage. Through our current work, we address this issue by devising a novel IMS session handover for W2W scenarios based on cross-layer parameter learning. We have utilized Real Simultaneous Dual Band (RSDB) feature available with Broadcom chip-set BCM4359 to realize our idea. The solution is verified with results obtained from live air experiments performed on Samsung Note5 running on Android M. We are able to achieve an average improvement of 67% in terms of call drops and 54% in terms of mute observed during IMS call as compared to legacy solutions.