Sensor Configuration Matters: A Systematic Evaluation of Multimodal SLAM on Quadruped Robots

A systematic evaluation of simultaneous localization and mapping (SLAM) systems on quadruped robots finds that sensor hardware choices — camera type, shutter mechanism, and inertial sensor tier — substantially affect performance under the punishing dynamics of legged locomotion, according to a paper submitted June 17 [1]. The study, authored by Roberto Corlito and submitted to arXiv on 17 Jun 2026 at 13:41:07 UTC, isolates the impacts of camera modalities, shutter techniques, and inertial sensor tiers on localization accuracy, algorithmic robustness, and computational resource utilization [1]. The work addresses what researchers describe as a critical evaluation gap: while visual-inertial SLAM has matured on wheeled, handheld, and aerial platforms, quadruped robots introduce distinct sensory challenges including foot-impact shocks, high-frequency mechanical vibrations, and rapid angular rotations that degrade standard perception pipelines [2]. The evaluation uses the GrandTour dataset, a 4,372 KB collection recorded on an ANYmal D quadruped, to benchmark state-of-the-art visual, visual-inertial, and LiDAR-visual-inertial SLAM methods [1]. Among the empirical findings, stereo configurations consistently outperform monocular and RGB-D modalities, and global shutter cameras significantly mitigate motion-induced tracking failures compared to rolling shutter cameras [2]. A counterintuitive result emerged: standard inertial integration can degrade the performance of primarily vision-based frameworks under harsh legged locomotion, challenging assumptions that adding an inertial measurement unit always improves robustness [2]. The paper offers concrete design guidelines for tailoring custom sensor payloads to achieve dependable perception on agile legged systems [1]. The work appears on arXiv, a preprint repository that, through integrations with platforms such as Hugging Face, allows researchers to link models, datasets, and interactive demos directly to paper pages [4][5]. The Hugging Face Papers page surfaces trending preprints daily, with community submissions and discussion threads, though the SLAM evaluation had not yet appeared among the day's trending papers as of June 22 [6]. The broader robotics perception landscape continues to see rapid preprint activity. A separate scoping review of quantum circuit generation systems, for instance, found that while all reviewed systems addressed syntactic validity and most addressed semantic correctness, none reported end-to-end evaluation on quantum hardware, leaving a gap between generated circuits and practical deployment [3]. That review, like the SLAM evaluation, underscores a recurring theme in systems research: laboratory benchmarks do not always translate to real-world resilience without careful hardware-software co-design [3].