Embedded Files
Abstract
Abstract
We propose a Tactile perception-based approach to address the Needle Threading task and name it T-NT. This task is divided into two main stages: Tail-end Finding and Tail-end Insertion. In the first stage, the agent traces the contour of the thread twice using vision-based tactile sensors mounted on the gripper fingers, to locate the tail-end of the thread. In the second stage, it employs a tactile-guided RL model to drive the robot to insert the thread into the target needle eyelet. Extensive experiments on real robots are conducted to demonstrate the efficacy of our method.
Overview
Overview
Tail-end Finding: We first find the thread's rest length by gliding it from the beginning point to the thread tip. Then, we trace the thread from the same beginning point and stop the movement early, to let the tail-end has a certain length. The tip position is estimated in consideration of the gravity force with a neural network, with the input of the tail-end length, the orientation estimated by the tactile image, and the gripper pose of the arm.
Tail-end Finding: We first find the thread's rest length by gliding it from the beginning point to the thread tip. Then, we trace the thread from the same beginning point and stop the movement early, to let the tail-end has a certain length. The tip position is estimated in consideration of the gravity force with a neural network, with the input of the tail-end length, the orientation estimated by the tactile image, and the gripper pose of the arm.
Tail-end Insertion: We first drive the tail-end in front of the eyelet, and poke on the tactile sensor. We formulate the problem as learning a policy that sequences move actions with a robot from tactile observations, including the needle eyelet masks and poking spot masks on the tactile images. We train the modle in simulation with one set of thread and eyelet, and directly transfer the agent to real-world experiments.
Tail-end Insertion: We first drive the tail-end in front of the eyelet, and poke on the tactile sensor. We formulate the problem as learning a policy that sequences move actions with a robot from tactile observations, including the needle eyelet masks and poking spot masks on the tactile images. We train the modle in simulation with one set of thread and eyelet, and directly transfer the agent to real-world experiments.
Simulation
Simulation
Needle #2 & Thread #2
Needle #2 & Thread #2
Needle #1 & Thread #1
Needle #1 & Thread #1
Needle #3 & Thread #3
Needle #3 & Thread #3
RL Training Curve with PPO
RL Training Curve with PPO
These curves demonstrate 5 of our RL training using PPO. We record the average reward of every 100 steps for our PPO training. The average reward have reached around 30, which indicates that the agent can finish the insertion with around 100/30 steps, and this number can help us to decide the threshold of steps to determine the success or failure of the insertion stage.
Experiments and Results
Experiments and Results
Eyelets and Threads: We use threads with thicknesses of 0.2mm, 0.5mm, 1mm, and 2mm. The sizes of the eyelets clearance are 0.6mm * 7.5mm, 1.6mm * 15mm, and 2.4mm * 9mm.
Eyelets and Threads: We use threads with thicknesses of 0.2mm, 0.5mm, 1mm, and 2mm. The sizes of the eyelets clearance are 0.6mm * 7.5mm, 1.6mm * 15mm, and 2.4mm * 9mm.
Qualitative Results: We can see in the figure that the tip pose estimation error is significantly small, within 5mm. And for different kinds of threads and eyelets, our method is able to perform great success rate. Quantitative results are shown in the paper, and in our video we demonstrate more needle-threading examples.
Qualitative Results: We can see in the figure that the tip pose estimation error is significantly small, within 5mm. And for different kinds of threads and eyelets, our method is able to perform great success rate. Quantitative results are shown in the paper, and in our video we demonstrate more needle-threading examples.
Tail-end Finding
Tail-end Finding
Tail-end Insertion
Tail-end Insertion
Page updated
Google Sites
Report abuse