Benchmark RadarSoftVTBench
SoftVTBench is a safety-aware visuo-tactile benchmark for deformable object manipulation in Isaac Sim, with FEM-simulated objects, multi-view RGB, tactile sensing, proprioception, language instructions, and separate Goal Success and Safety Success metrics.
- Released
- 2026-07-05
- Readiness
- Paper only
- Primary field
- Robotics & Autonomous Systems
Why it matters
Existing manipulation benchmarks focus on task success and overlook physical safety, such as avoiding drops or excessive deformation. SoftVTBench addresses this gap by providing a protocol that evaluates both goal achievement and safety, offering decision value for developing safer robotic manipulation policies.
Motivation
Deformable object manipulation poses challenges beyond task completion: successful execution must also maintain safe physical interaction, holding the object stably without slip or drop while avoiding excessive deformation.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.