Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• AdaClearGrasp is a closed-loop decision-execution framework for adaptive clearing and zero-shot dexterous grasping in densely cluttered environments.method 95%
• A pretrained vision-language model (VLM) is used to interpret visual observations and language task descriptions for grasp interference reasoning and planning.method 95%
• A reinforcement learning policy with a relative hand-object distance representation is trained for zero-shot generalization in dexterous grasping.method 95%
• Visual feedback is used during execution to monitor outcomes and trigger replanning upon failures, creating a closed-loop correction mechanism.method 95%
• Clutter-Bench is introduced as the first simulation benchmark for evaluating language-conditioned dexterous grasping in clutter, with graded clutter complexity.method 95%
• Clutter-Bench includes seven target objects across three clutter levels, resulting in 210 task scenarios.method 95%
• AdaClearGrasp significantly improves grasp success rates in densely cluttered environments.result 95%
• The framework enables robots to adaptively decide whether to clear surrounding objects or directly grasp the target.technical 95%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card