Precision agriculture demands continuous and accurate monitoring of soil moisture M and key macronutrients, including nitrogen N, phosphorus P, and potassium K, to optimize yields and conserve resources. Wireless soil sensing has been explored to measure these four components; however, current solutions require recalibration (i.e., re-train the data processing model) to handle variations in soil texture (characterized by aluminosilicates Al and organic carbon C, limiting their practicality. To address this, we introduce SoilX, a calibration-free soil sensing system that jointly measures six key components: {M, N, P, K, C, Al}. By explicitly modeling C and Al, SoilX eliminates texture- and carbon-dependent recalibration. SoilX incorporates Contrastive Cross-Component Learning (3CL), with two customized terms: the Orthogonality Regularizer and Separation Loss, to effectively disentangle cross-component interference. Additionally, we design a novel Tetrahedral Antenna Array with an antenna-switching mechanism, which can robustly measure soil dielectric permittivity independent of device placement. Extensive experiments demonstrate that SoilX reduces estimation errors by 23.8% to 31.5% over baselines and generalizes well to unseen fields.

Integrating large pre-trained models into federated learning (FL) can significantly improve generalization and convergence efficiency. A widely adopted strategy freezes the pre-trained backbone and fine-tunes a lightweight task head, thereby reducing computational and communication costs. However, this partial fine-tuning paradigm introduces new security risks, making the system vulnerable to poisoned updates and backdoor attacks. To address these challenges, we propose FedRACE, a unified framework for robust FL with partially frozen models. FedRACE comprises two core components: HStat-Net, a hierarchical network that refines frozen features into compact, linearly separable representations; and DevGuard, a server-side mechanism that detects malicious clients by evaluating statistical deviance in class-level predictions modeling generalized linear models (GLMs). DevGuard further incorporates adaptive thresholding based on theoretical misclassification bounds and employs randomized majority voting to enhance detection reliability. We implement FEDRACE on the FedScale platform and evaluate it on CIFAR-100, Food-101, and Tiny ImageNet under diverse attack scenarios. FedRACE achieves a true positive rate of up to 99.3% with a false positive rate below 1.2%, while preserving model accuracy and improving generalization.

Soil composition sensing is essential for precision agriculture, sustainable land management, and optimizing crop yields. However, existing sensing systems face major limitations, including extensive calibration needs to account for soil variability, a narrow focus on measurements for specific properties, and sensitivity to the placement of the device. These challenges hinder practical deployment. This demo shows SoilX, a comprehensive wireless soil sensing system that quantifies all major soil components---including aluminosilicates, water, organic carbon, and micronutrients---using RF and VNIR sensing technologies. To enable the generalizability, SoilX employs contrastive pretraining to mitigate cross-component interference. Additionally, a tetrahedron-based antenna geometry ensures robustness to device placements. Extensive evaluations in both lab and field settings demonstrate that SoilX achieves state-of-the-art accuracy in soil composition analysis with low costs.

Arm posture tracking is essential for many applications, such as gesture recognition, fitness training, and motion-based controls. Smartwatches with Inertial Measurement Unit (IMU) sensors (i.e., accelerometer, gyroscope, and magnetometer) provide a convenient way to track the orientation and location of the wrist. Existing orientation estimations are based on predefined data fusion methods that do not consider the variations in the data quality of different IMU sensors. Existing location estimations rely on the estimated orientation results. A small orientation estimation error may cause high inaccuracy in location estimation. Moreover, these location estimation algorithms, e.g., Hidden Markov Model and Particle Filters, cannot provide real-time tracking on commercial mobile devices due to high computation overhead. This paper presents RTAT, a Real-Time Arm Tracking system that tackles the above limitations in a data-driven way. RTAT estimates both orientation and location simultaneously using a multitask learning neural network. It also incorporates a unique attention layer and a dedicated loss function to learn the dynamic relationship among IMU sensors. RTAT supports real-time tracking by performing model inference on smartphones. Finally, to train RTAT's neural network, we develop an easy-to-use labeled data collection system that uses a low-cost virtual reality system to provide orientation and location labels for the smartwatch. Extensive experiments show RTAT significantly outperforms existing state-of-the-art solutions in both accuracy and latency.

In this demo, we present RTAT, a real-time arm tracking system that tracks both orientation and location of a smartwatch simultaneously by a multitask learning neural network. We incorporate an attention layer and design a dedicated loss for the multitask neural network to learn the dynamic relationships among Inertial Measurement Unit (IMU) sensors. RTAT supports real-time tracking by performing deep learning inference on a smartphone. Finally, to train RTAT, we develop an easy-to-use labeled data collection system that uses a low-cost virtual reality system to measure the ground truth orientation and location of the smartwatch. Extensive experiments show RTAT outperforms significantly the state-of-the-art solutions in inference accuracy and latency.