cs.LGMay 22, 2025

SPAR: Placement-aware Self-Supervised Representation Learning for Distributed Sensing

Authors: Yizhuo Chen, Tianchen Wang, You Lyu, Yanlan Hu, Jinyang Li, Tomoyoshi Kimura, Hongjue Zhao, Yigong Hu, 

Organizations: University of Illinois Urbana-Champaign, Urbana, IL, USA · Stanford University, Stanford, CA, USA

Abstract

To support distributed sensing applications that require understanding of entities in the environment and their locations, this paper develops a self-supervised approach for placement-aware representation learning in spatially-distributed multi-sensor systems. The goal is to produce a latent representation of the physical environment that decouples upstream sensor placement from downstream analytics (in the sense that the pretrained encoder is reused without retraining when sensor placement changes), while supporting spatial reasoning tasks, such as joint target localization and classification, that are sensitive to sensor placement. A challenge lies in the multiplicity of sensing vantage points that collectively offer information on the shared physical environment. The dependence of measurements on sensor locations implies that the latent representation must jointly encode both the observed signals and observer vantage points, doing so in a way that supports the downstream tasks. Moreover, since the exact deployment-time sensor placement is generally not known at training time, the representation learning framework must be pre-trained to produce such a representation without (training-time) knowledge of deployment-time placement. This challenge distinguishes our goal from approaches, such as NeRF and Gaussian Splatting, that are trained to represent each particular environment given a specific placement of observation points in that environment. In contrast, our representation learning solution, called SPAR, pretrains an encoder independently of future placement by utilizing a novel training objective that emphasizes the duality between signals and vantage points...

Figures & tables

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Jun 15, 2026

Sensor-Conditioned Representation Learning via Scene-Relevant Observation Quotients

Learned representations in intelligent sensing systems are often evaluated by reconstruction fidelity or downstream prediction accuracy, but these criteria do not specify which latent distinctions are justified by the sensing process. In sensor-conditioned environments, nuisance factors can change measurements without changing the scene, while distinct scenes may be indistinguishable under limited sensing capability. This paper formulates sensor-conditioned representation correctness as preserving sensing-supported scene distinctions while suppressing nuisance-induced and sensor-unsupported variation. We introduce the scene-relevant observation quotient, a representation target induced by sensing-supported distinguishability after nuisance canonicalization, and develop Observation-Quotient Tucker-Structured Autoencoding (OQ-TSAE), a scene-nuisance factorized framework with diagnostics for false distinction, false merge, nuisance sensitivity, and latent ordering consistency. Experiments on a controlled benchmark show that quotient-consistent supervision improves representation-correctness diagnostics over reconstruction-oriented, metric-learning, and contrastive-learning baselines. Sensitivity, perturbation, and ablation studies show the importance of quotient-aligned supervision, reliable quotient relations, and quotient geometry. Complementary real-radar experiments show that a reconstruction-only OQ-TSAE variant retains competitive downstream utility, robustness under observation degradation, and low seed-to-seed variability. These results suggest that sensor-conditioned representations should be evaluated not only by predictive utility, but also by whether their latent geometry preserves sensing-justified scene distinctions.
Mar 13, 2026

Representation Learning for Spatiotemporal Physical Systems

Machine learning approaches to spatiotemporal physical systems have primarily focused on next-frame prediction, with the goal of learning an accurate emulator for the system's evolution in time. However, these emulators are computationally expensive to train and are subject to performance pitfalls, such as compounding errors during autoregressive rollout. In this work, we take a different perspective and look at scientific tasks further downstream of predicting the next frame, such as estimation of a system's governing physical parameters. Accuracy on these tasks offers a uniquely quantifiable glimpse into the physical relevance of the representations of these models. We evaluate the effectiveness of general-purpose self-supervised methods in learning physics-grounded representations that are useful for downstream scientific tasks. Surprisingly, we find that not all methods designed for physical modeling outperform generic self-supervised learning methods on these tasks, and methods that learn in the latent space (e.g., joint embedding predictive architectures, or JEPAs) outperform those optimizing pixel-level prediction objectives. Code is available at https://github.com/helenqu/physical-representation-learning.
Sep 21, 2026

Tactile-JEPA: Topology-Aware Self-Supervised Representation Learning for Distributed Tactile Sensors

Tactile sensing is an essential modality for robots performing contact-rich, dexterous manipulation, particularly under visual occlusion. While pre-trained image encoders are standard in robot learning pipelines, tactile encoders are still commonly trained from scratch from raw, noisy signals, which might limit their expressivity. Existing self-supervised learning (SSL) approaches focus predominantly on vision-based tactile sensors, leaving distributed electronic skins largely unaddressed. These sensors, however, have a distinctive property: their sensing elements are sparse and irregularly arranged over the surface they cover, which makes direct reuse of visual SSL methods suboptimal. We present Tactile-JEPA, an efficient self-supervised pre-training method that uses the spatial arrangement of tactile sensors to learn topology-aware representations. Specifically, it is trained to predict the embeddings of masked sensing elements from the unmasked remainder, using the sensor connectivity graph to guide spatial masking. Our analysis shows that effective tactile representations require capturing both local contact details and the global state of the tactile surface, which we achieve through dual-scale masking. Across three diverse datasets spanning magnetic and piezoresistive sensors, different robot embodiments, and single- and paired-sensor configurations, Tactile-JEPA reduces force estimation error by 6.3% and in-hand orientation error by 20.8% over the prior state-of-the-art, with consistent gains in other downstream applications, including policy learning. Overall, our results demonstrate that the benefit of tactile sensing depends critically on the quality of encoder pre-training, a problem which Tactile-JEPA addresses directly. Code is available at https://github.com/E-Kovtun/tactile.