Benchmark RadarHumanoidVLN
Evaluates vision-language navigation for humanoid robots across four embodiments in physics-grounded simulator scenarios. Includes 933 episodes with instructions and multiple stylistic variants, assessing success rate and normalized Dynamic Time Warping.
- Released
- 2026-08-13
- Readiness
- Inspectable
- Primary field
- Robotics & Autonomous Systems
Why it matters
Addresses the gap in VLN benchmarks for bipedal locomotion and diverse morphologies, providing a platform to compare navigation models under physical constraints and supporting sim-to-real transfer studies.
Motivation
Vision-Language Navigation (VLN) for humanoid robots poses challenges existing benchmarks fail to address: bipedal locomotion imposes physical constraints absent from wheeled agents, humanoid morphologies vary across platforms, and egocentric observations are distorted by locomotion-induced camera dynamics.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.