Driving · 2026-10-14
LynxScenes: A Large-Scale Real-World Benchmark for Outdoor Wheel-Legged Robot Perception, Navigation, Imagination and Reconstruction
Jun Tian, Hao Wu, Ruo Chen, Yili Liao, Hongyi Lin, Jincheng Yu, Chao Li, Yingjie Niu
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
The authors introduce LynxScenes, a large-scale real-world dataset and benchmark collected with a LYNX M20 wheel-legged robot for outdoor, socially compliant navigation and scene understanding. It provides ~100K non-overlapping 30-s clips (833.6 h, 3,294.7 km, 3.0M annotated frames) with synchronized multimodal sensing and dense annotations, supporting instruction-conditioned trajectory prediction, BEV perception, future scene prediction, and multi-view reconstruction.
Why it matters
Wheel-legged robots in human-shared outdoor spaces need perception, future-scene reasoning, instruction-aware planning, and reconstruction under the same real-world embodiment and interaction distribution. LynxScenes jointly provides synchronized sensing, dense 3D object/road-structure supervision, behavior-level multi-layer navigation instructions, and aligned continuous future trajectories, enabling evaluation beyond trajectory imitation via a log-replay pseudo-closed-loop score.
Method
- Collect ~100K non-overlapping 30-s teleoperated navigation clips with a LYNX M20 robot, synchronized four fisheye cameras, dual LiDARs, IMU, odometry, and robot states; sample annotated frames at 1 Hz.
- Provide dense spatial annotations: 20.82M oriented 3D boxes (27 categories) with persistent identities, 13.89M road-structure annotations (4 classes), and 20 navigation instructions in three layers aligned to continuous future trajectories.
- Benchmark four tasks (instruction-conditioned trajectory prediction, BEV perception, future scene prediction, multi-view reconstruction) and add a log-replay-based pseudo-closed-loop evaluator for safety/progress/instruction compliance (QNS).
Limitation
The reconstruction task evaluates observation-grounded novel-view rendering on real robot navigation sequences rather than claiming full closed-loop simulation capability, and its benchmark scope does not directly measure geometric reconstruction accuracy, future scene dynamics, or closed-loop navigation safety.
Abstract (from arXiv)
Wheel-legged robots operating in outdoor spaces shared with humans require accurate perception, socially compliant navigation, future-scene reasoning, and scene reconstruction. However, existing datasets typically support these capabilities in isolation and rarely provide dense spatial and behavior-level supervision on real wheel-legged platforms. We introduce LynxScenes, a large-scale real-world dataset and benchmark collected with a LYNX M20 wheel-legged robot. It contains approximately 100K 30-s navigation clips spanning 833.6 h and 3,294.7 km, with 3.0M annotated frames. The synchronized recordings include four fisheye cameras, dual LiDARs, IMU, odometry, and robot states. Annotations comprise 20.82M oriented 3D boxes across 27 categories, 13.89M road-structure annotations, and 20 navigation instructions organized into three layers and aligned with continuous future trajectories. LynxScenes supports four complementary tasks: instruction-conditioned trajectory prediction, BEV-based perception of 3D objects and road structures, future scene prediction, and multi-view scene reconstruction. For trajectory prediction, we further introduce a log-replay-based pseudo-closed-loop evaluator that measures safety, progress, and instruction compliance beyond trajectory imitation. Extensive evaluations across representative navigation, perception, world-model, and reconstruction methods reveal substantial remaining challenges in close-range outdoor robot navigation. The dataset and evaluation toolkit are planned for public release.
Related papers
- LiteNWM: Efficient Latent World Models for Onboard Visual Navigation in the Wild
- ScaleNav: Scalable Internet Video Pretraining for Language-Conditioned Local Navigation
- From Pixels to Executable Physics: Behavior-Grounded World Models for Embodied Action Consequence Prediction
- Acting from Belief, Looking When Needed: A Bayesian Spatial World Model for Navigation under Intermittent Perception
- DR-VLA: Learning Human-Like Quadruped Navigation from Egocentric Web Videos via Domain-Invariant Scene Representations
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-14 UTC
LynxScenes: A Large-Scale Real-World Benchmark for Outdoor Wheel-Legged Robot Perception, Navigation, Imagination and Reconstruction
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18093
Explicit source versions (1)
- v2 · 2026-10-17
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
- Explicit source version detected · v2
Source date: 2026-10-17 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:id explicit version (official metadata)
Abstract · arxiv
Wheel-legged robots operating in outdoor spaces shared with humans require accurate perception, socially compliant navigation, future-scene reasoning, and scene reconstruction. However, existing datasets typically support these capabilities in isolation and rarely provide dense spatial and behavior-level supervision on real wheel-legged platforms. We introduce LynxScenes, a large-scale real-world dataset and benchmark collected with a LYNX M20 wheel-legged robot. It contains approximately 100K 30-s navigation clips spanning 833.6 h and 3,294.7 km, with 3.0M annotated frames. The synchronized recordings include four fisheye cameras, dual LiDARs, IMU, odometry, and robot states. Annotations comprise 20.82M oriented 3D boxes across 27 categories, 13.89M road-structure annotations, and 20 navigation instructions organized into three layers and aligned with continuous future trajectories. LynxScenes supports four complementary tasks: instruction-conditioned trajectory prediction, BEV-based perception of 3D objects and road structures, future scene prediction, and multi-view scene reconstruction. For trajectory prediction, we further introduce a log-replay-based pseudo-closed-loop evaluator that measures safety, progress, and instruction compliance beyond trajectory imitation. Extensive evaluations across representative navigation, perception, world-model, and reconstruction methods reveal substantial remaining challenges in close-range outdoor robot navigation. The dataset and evaluation toolkit are planned for public release.