BARN Challenge 2026
Annual navigation challenge using hidden BARN simulations and physical finals; the underlying BARN dataset is a related but distinct entity.
- Entity
- challenge
- Metric
- Simulation trial score
- Evidence
- verified
Registry
Robotics benchmarks, described on common ground
Search by name or scope, then narrow by domain and evaluation setting. Every record is source-backed; none is a ranking.
10 benchmarks
CALVIN
Simulated benchmark, environment, dataset, and challenge for long-horizon language-conditioned control.
- Entity
- benchmark suite
- Metric
- No universal metric
- Evidence
- potentially stale
Habitat
Simulation ecosystem supporting navigation, rearrangement, instruction following, embodied question answering, and related challenge families.
- Entity
- platform framework
- Metric
- No universal metric
- Evidence
- verified
LIBERO
Lifelong robot-learning benchmark with four suites and 130 tasks.
- Entity
- benchmark suite
- Metric
- Forward Transfer
- Evidence
- verified
ManiSkill 3
Current simulation and training framework and benchmark ecosystem; ManiSkill generations are materially distinct and must be versioned.
- Entity
- platform framework
- Metric
- No universal metric
- Evidence
- verified
Meta-World
Multi-task and meta-reinforcement-learning benchmark suite with versioned MT and ML protocols.
- Entity
- benchmark suite
- Metric
- Average task success rate
- Evidence
- verified
Open X-Embodiment
Open dataset collection and RT-X cross-embodiment research ecosystem; it is not one fixed benchmark.
- Entity
- dataset resource
- Metric
- No universal metric
- Evidence
- verified
RLBench
Robot-learning benchmark and environment with 100 hand-designed tasks, variations, demonstrations, and multimodal observations.
- Entity
- benchmark suite
- Metric
- Task success rate
- Evidence
- verified
RoboChallenge Table30
Real-robot tabletop benchmark suite on the RoboChallenge.ai evaluation platform; Table30v2 is modeled as a versioned child.
- Entity
- challenge
- Metric
- End-to-end task success rate
- Evidence
- verified
RobotPerf
Open reference robotics-computing benchmark suite using ROS 2 as a common baseline across multiple workload and hardware categories.
- Entity
- benchmark suite
- Metric
- No universal metric
- Evidence
- verified