ChicGrasp: Imitation‐Learning‐Based Customized Dual‐Jaw Gripper Control for Manipulation of Delicate, Irregular Bio‐Products

Automated poultry processing lines still rely on humans to lift slippery, easily bruised carcasses onto a shackle conveyor. Deformability, anatomical variance, and hygiene rules make conventional suction and scripted motions unreliable. We present ChicGrasp, an end‐to‐end hardware–software codesign for this task. An independently actuated dual‐jaw pneumatic gripper clamps both chicken legs. A conditional diffusion‐policy controller is trained on 100 multiview teleoperation demonstrations with red–green–blue images and robot proprioception. It predicts a five‐dimensional action vector that combines three Cartesian translations with two binary jaw commands in a single step. On individually presented raw broiler carcasses, our system achieves an 80.71% grasp‐and‐lift success rate. The policy‐controlled grasp time phase takes 28 s, and the total cycle time, including the scripted rehang, requires 38 s, whereas state‐of‐the‐art implicit behavioral cloning and long short–term memory – Gaussian mixture model baselines fail entirely. All CAD, code, and datasets will be released as open source. ChicGrasp shows that imitation learning can begin to bridge the gap between rigid hardware and variable bio‐products, offering a reproducible benchmark and a public dataset for researchers in agricultural engineering and robot learning. Resources are available at github .