Gesture-based interaction is vital for human-robot communication in settings where verbal cues are constrained, such as aircraft ramps, industrial sites, and healthcare settings like eldercare facilities, rehabilitation centers, and operating rooms. Robots often rely on gestures due to factors like noise, communication barriers, or the need for sterile, hands-free interaction. These dynamic settings demand not only robust gesture recognition but also lifelong adaptability to new and evolving gesture vocabularies. A key challenge in such scenarios is catastrophic forgetting in Class-Incremental Learning (CIL), where the robot forgets previously learned signals when new ones are introduced. We present Momentary Experience Replay (MoER), a novel framework that enhances experience replay strategies to prioritize recently acquired knowledge, which is more vulnerable to degradation. MoER introduces a Momentary Buffer that dynamically increases replay frequency for recently learned gesture signals. Evaluated on a 12-class gesture recognition task using the public NATOPS dataset in a CIL experimental setting, MoER achieves 92% accuracy, outperforming baseline methods (ER and DER++) by up to 6%. MoER also shows improvements in backward transfer and retention stability, supporting its role in continual signal recognition in dynamic environments. This work lays the foundation for lifelong learning in healthcare robotics, where continual adaptation to new tasks and user-specific gestures is critical for effective and trustworthy assistive interaction.

In dense RFID systems, where numerous readers and tags operate simultaneously in close proximity, securing reader-to-reader communication is essential to prevent attacks such as eavesdropping and spoofing. Existing protocols focus primarily on reader-to-tag communication and use static security mechanisms that may be inadequate in dynamic conditions. We propose DCA-KEAE, a Dynamic Context-Aware Key Exchange and Adaptive Encryption framework for RFID systems. DCAKEAE adapts security protocols in real-time based on factors such as reader proximity, system load, and threat levels: it employs lightweight symmetric keys for low-risk scenarios and escalates to stronger protocols like ECDH and AES-256 in highrisk environments. Evaluations with up to 10,000 readers show that DCA-KEAE reduces latency, optimizes encryption, and improves system throughput, offering a scalable and efficient solution for RFID networks, with applications extending to the Internet of Things (IoT), industrial automation, and smart grids.

Dense RFID environments pose significant challenges, such as reader collisions, tag interference, and scalability issues, which degrade system performance and reliability. This paper introduces RFIDNet, a novel protocol designed to address these challenges by dynamically coordinating reader activities and optimizing network resource utilization. The proposed RFIDNet is an innovative framework of advanced mechanisms that include a Carrier Sense Multiple Access with Reader Arbitration (CSMARA) scheme for efficient reader coordination, Dynamic Frequency Hopping (DFH) for interference mitigation, and merging Frequency and Time Division Multiple Access (F/TDMA) with Reduce Coverage Control (RCC) to handle unresolved contention. Experimental validations using a Universal Software Radio Peripheral (USRP) testbed and MATLAB simulations demonstrate that RFIDNet improves the overall system performance compared to the baseline. This confirms RFIDNet's robustness and scalability, making it a viable solution for realworld, dense RFID deployments.

In this work, we present a novel approach to enhance the robustness of autonomous robotic Radio Frequency Identification (RFID) inventory systems using Conformal Prediction (CP). Recent AI-driven approaches, especially deep-learning models, have made significant advances in performing inventory strategies and action planning. However, these models lack the capability to measure uncertainty during the prediction process, which can result in accumulated errors and lead to catastrophic failures. To address the above challenge, we propose a confidenceguaranteed policy using CP to ensure reliable predictions in RFID inventory tasks. Our method focuses on managing the uncertainty in sub-goal estimation for a trained model, ensuring that predictions can meet or exceed a user-specific confidence level. We conduct extensive experiments to assess the proposed method by regulating an existing model and evaluate its effectiveness in identifying uncertain predictions. The experimental results demonstrate the effectiveness of our approach in improving both the reliability and efficiency of RFID inventory tasks, ensuring consistent and trustworthy operation.

ABSTRACT Dense radio frequency identification (RFID) networks suffer from severe reader collisions, redundant reads, and inefficient resource utilization, particularly in large‐scale deployments. Existing approaches, including centralized and hybrid scheduling schemes, fail to scale effectively due to their reliance on global coordination and static allocation mechanisms. This paper presents NEMO (neighbourhood‐aware efficient management and optimization), a fully decentralized neighbourhood‐aware RFID network framework that dynamically optimizes scheduling, power control, and frequency allocation without requiring global coordination. NEMO leverages a novel adaptive scheduling mechanism to mitigate collisions while ensuring fair and efficient tag interrogation. Extensive universal software radio peripheral‐based hardware experiments and large‐scale simulations with up to 5000 readers and 1,000,000 tags demonstrate that NEMO outperforms state‐of‐the‐art protocols by achieving 25% higher throughput, 30% fewer collisions, 40% reduction in redundant reads, and improved energy efficiency by 18%. Additionally, NEMO exhibits scalability and robustness under extreme network congestion by maintaining high performance even as the numbers of readers and tags increase. The proposed framework is highly applicable to real‐world RFID deployments in warehouses, logistics, smart retail, Internet of Things, and industrial automation, where dense RFID environments demand efficient, adaptive, and decentralized resource management.

The fast development in Deep Learning (DL) has made it a promising technique for various autonomous robotic systems. Recently, researchers have explored deploying DL models, such as Reinforcement Learning and Imitation Learning, to enable robots for Radio-frequency Identification (RFID) based inventory tasks. However, the existing methods are either focused on a single field or need tremendous data and time to train. To address these problems, this paper presents a Cross-Modal Reasoning Model (CMRM), which is designed to extract high-dimension information from multiple sensors and learn to reason from spatial and historical features for latent cross-modal relations. Furthermore, CMRM aligns the learned tasking policy to high-level features to offer zero-shot generalization to unseen environments. We conduct extensive experiments in several virtual environments as well as in indoor settings with robots for RFID inventory. The experimental results demonstrate that the proposed CMRM can significantly improve learning efficiency by around 20 times. It also demonstrates a robust zero-shot generalization for deploying a learned policy in unseen environments to perform RFID inventory tasks successfully.

With the growing demand for location-based services, fingerprint has become a hot topic in the area of Internet of Things (IoT). However, the performance of fingerprinting-based indoor localization systems is usually affected by the quality and granularity of fingerprints. In this article, we present MapLoc, a long short-term memory (LSTM)-based indoor localization system that takes advantage of the continuous indoor uncertainty maps created using both earth magnetic field readings and WiFi received signal strengths (RSSs). A deep Gaussian process (DGP) model is trained to create indoor radio maps with confidence intervals, which are referred as uncertainty maps. Utilizing the uncertainty maps, an LSTM-based location prediction model is pretrained with artificial trajectory data sampled from the uncertainty maps, and then fine-tuned with the signal measurements collected in the field. In the training process, auxiliary outputs are implemented to overcome overfitting and improve the robustness of the system. Our extensive experiments demonstrate the outstanding performance of the proposed MapLoc system.

The developments in reinforcement learning provide a powerful and efficient learning framework for autonomous robotic systems. However, prior works rarely embed historical observations due to the exponentially increasing complexity, which may not perform well for large-scale long-horizon tasks that might require hundreds and thousands of steps to complete. In this paper, we propose Recurrent Imitation and Reinforcement Learning (RIRL) to address the challenges and enable robots for such tasks. The proposed RIRL incorporates a long short-term memory (LSTM) network to retain long-term memories, which could be an effective and efficient method to tackle the long dependency problem raised in long-horizon robotic tasks. To assess the performance of the RIRL, we test it with an optimized path planning problem for a robot to perform a Radiofrequency identification (RFID) inventory in dynamic and previously unknown environments. We experimentally validate RIRL’s feasibility and effectiveness in a visual game-based simulation platform, where the proposed RIRL model outperforms three baseline schemes with considerable gains.

With outbreak of the COVID-19 pandemic, contact tracing has become an important problem. It has been proven that maintaining social distance and isolating affected people are highly beneficial for curbing the spread of COVID-19, which all depend on identifying people’s trajectories. However, the current interview-based approach is costly, and the existing mobile app-based schemes rely on complete and accurate data. In this paper, we propose a transformer encoder-based approach with spatial position embedding extracted using a graph Combinatorial Laplacian matrix to interpolate incomplete human trajectories. To model human trajectory, we propose a graphical embedded module to extract spatial features based on predefined location clusters. The incomplete trajectory sequences are first preprocessed into matrices and then used to train a deep transformer encoder network for trajectory completion. Our experiments using a real world Bluetooth Low Energy (BLE) dataset validate the efficacy of our proposed approach, which outperforms several baseline methods.

In this paper, we present a Swin Transformer based indoor localization framework that employs RF hologram tensors to locate multiple ultra-high frequency (UHF) passive Radiofrequency identification (RFID) tags. The RF hologram tensor captures the strong relationship between RFID measurements and spatial location, and helps to improve the robustness of the system in dynamic environments. We develop a Swin Transformer-based hologram filter network to clean the fake peaks in hologram tensors caused by multipath propagation and phase wrapping, exploring the spatial relationship between tags. In contrast to fingerprinting-based localization systems that use deep networks as classifier, the proposed network treats localization as a regression problem. An intuitive peak finding algorithm is introduced for location estimation using the sanitized hologram tensors. We prototype the proposed system using commodity RFID devices and conduct extensive experiments to evaluate its performance.

Fingerprinting-based indoor localization has been a research focus for GPS denied areas. The development of neural networks has greatly promoted its application in indoor localization systems. However, recent studies showed that the machine learning models, including state-of-the-art neural networks, are vulnerable to adversarial examples, and thus, neural network-based indoor localization systems are also under the threat of adversarial attacks. To investigate the effect of adversarial attacks on indoor localization systems and to make such systems resilient to adversarial attacks, we propose AdvLoc, an adversarial deep learning for indoor localization system. With the proposed AdvLoc system, the effect of adversarial attacks on indoor localization is studied under six types of adversarial attack methods in both black-box attack and white-box attack scenarios. Furthermore, adversarial training is utilized in offline training of the proposed AdvLoc system, which is effective against first-order adversarial attacks. The proposed AdvLoc system is implemented with commodity WiFi devices and evaluated with extensive experiments in two representative indoor environments. The experimental results verify the robustness of the proposed system against first-order adversarial attacks in representative indoor environments.

In this paper, we present MulTLoc, a deep learning based indoor localization system for localizing multiple ultra-high frequency (UHF) passive RFID tags with RF hologram tensor filtering and upscaling. The proposed system leverages the RF hologram tensor as the input of the deep convolutional networks. The RF hologram tensor exhibits a strong relationship between the observation and the spatial location, which enhances the robustness of the system to the dynamic environment and equipment. To sanitize the RF hologram tensor, two architectures of deep networks are newly proposed. The hologram filter network suppresses the fake peaks resulting from the multipath and phase wrapping by leveraging the spatial relationship between tags. The tensor upscaling network recovers the high resolution hologram tensor from the output of the previous network, which enhances the localization accuracy of the system further. Comparing with the fingerprinting based localization systems using deep networks as the classifier, the networks in the MulTLoc system treat the localization problem as the regression problem, in which the ambiguity between fingerprints is reserved. To avoid the inherent errors in the fingerprinting based localization systems, the location estimation is given by intuitive peak finding algorithms using the recovered RF hologram tensor. We implement the proposed MulTLoc system with commodity RFID devices and verify its performance with extensive experiments.

With the increasing demand for location-based service, WiFi-based localization has become one of the most popular methods due to the wide deployment of WiFi and its low cost. To improve this technology, we propose DeepMap, a deep Gaussian process for indoor radio map construction and location estimation. Received signal strength (RSS) samples are used in DeepMap to generate accurate and fine-grained radio maps. A two-layer deep Gaussian process model is designed to determine the relationship between the location and RSS samples, while the model parameters are optimized with an offline Bayesian training method. To identify the location of a mobile device, a Bayesian fusion method is proposed, which leverages RSS samples from multiple access points (APs) to achieve high location estimation accuracy. We conduct comprehensive experiments to verify the performance of DeepMap in two indoor settings. DeepMap's robustness is validated using limited training data.

RFID-based sensors have attracted great interest due to the wide employment of RFID tags, ease of deployment, and low cost. However, most RFID-based sensors rely on hardware modification, which increases the cost and hampers the deployment. In this paper, we propose RFThermometer, a temperature estimation system with commercial UHF RFID tags. We first study the effect of temperature on RFID phase. To mitigate the problem caused by missing phase measurements, we propose a tensor completion method to restore missing phases and a Gaussian Process (GP) model for building phase-temperature maps in the offline stage. A dynamic time warping based greedy method is proposed to estimate the unknown temperature in the online stage. Experimental results verify the performance of RFThermometer.

With the fast-growing adoption of the radio-frequency identification (RFID) technology, RFID-based sensors have attracted great interest. Due to the limitation of RFID tags, most existing RFID-based temperature sensing works rely on hardware modification, which increases the cost and hampers its deployment. In this article, we propose RFThermometer, a remote temperature sensing system with commercial ultra high frequency (UHF) RFID tags. We first investigate the effect of temperature on RFID phases. To alleviate the precision deterioration caused by missing phase measurements, a tensor completion method is proposed to restore missing phases and a Gaussian process model is leveraged to construct a phase-temperature map in the offline stage. In the online stage, the unknown temperature is estimated by a dynamic time warping (DTW)-based greedy method. Extensive experimental results are presented to validate the performance of RFThermometer with off-the-shelf RFID devices.

In this paper, we present DeepMap, a deep Gaussian process for indoor radio map construction and location estimation. To address the shortcomings of existing Gaussian process based approaches, we present a DeepMap system, which employs deep Gaussian process for constructing received signal strength (RSS) radio maps and a Bayesian algorithm for online localization. We design a two-layer deep Gaussian process model to capture the relationship between the RSS space and the location space and provide an offline Bayesian training method to determine model parameters. A Bayesian fusion method using multiple APs is proposed for accurate location estimation. Experimental results verify the performances of DeepMap in a large indoor environment and validate its robustness with moderate training data.

With the increasing commercial prospect of personal Unmanned Aerial Vehicle (UAV), human and UAV interaction has been a compelling and challenging task. In this paper, we present the RFHUI, a human and UAV interaction system based on passive radio-frequency identification (RFID) technology which provides a remote control function. Three or more Ultra high frequency (UHF) RFID tags are attached on a board to create a hand-held controller. A COTS (Commercial Off-The-Shelf) RFID reader with multiple antennas is deployed to collect the observations of the tags. According to the phase measurement from the RFID reader, we leverage a Bayesian filter based method to localize the position of all tags in a global coordinate. From the estimated position of the attached tags, a 6 DOF (Degrees of Freedom) pose of the controller can be obtained. Therefore, when the user moves the controller, its pose will be precisely tracked in a real-time manner. Then, the flying commands, which are generated from the estimated pose of the controller, are sent to the UAV for navigation. We implemented a prototype of the RFHUI, and the experiment results show that it provides precise poses with 0.045 m error in position and 2.5° error in orientation for the controller. It therefore enables the controller to precisely and intuitively instruct the UAV's navigation in an indoor environment.