A Robotic Gripper with Versatility and Collision Robustness for Robot Learning

1ReNeu Robotics Lab, Department of Mechanical Engineering
2Robot Interactive Intelligence Lab, Department of Computer Science
The University of Texas at Austin
Under Review


We present a new approach to robot hand design specifically suited for successfully implementing robot learning methods to accomplish tasks in daily human environments. We introduce BaRiFlex, an innovative gripper design that alleviates the issues caused by unexpected contact and collisions during robot learning, offering robustness, grasping versatility, task versatility, and simplicity to the learning processes. This achievement is enabled by the incorporation of low-inertia actuators, providing high Back-drivability, and the strategic combination of Rigid and Flexible materials which enhances versatility and the gripper’s resilience against unpredicted collisions. Furthermore, the integration of flexible Fin-Ray linkages and rigid linkages allows the gripper to execute compliant grasping and precise pinching. We conducted rigorous performance tests to characterize the novel gripper’s compliance, durability, grasping and task versatility, and precision. We also integrated the BaRiFlex with a 7 Degree of Freedom (DoF) Franka Emika’s Panda robotic arm to evaluate its capacity to support a trial-and-error (reinforcement learning) training procedure. The results of our experimental study are then compared to those obtained using the original rigid Franka Hand and a reference Fin-Ray soft gripper, demonstrating the superior capabilities and advantages of our developed gripper system.

Performance Evaluation

BaRiFlex has the ability to withstand contacts and collisions, perform precise pinch grasps as well as compliant grasps, to be robust to noises in poses of objects and finally, to perform a wide variety of tasks!

Robustness to Collisions

Grasping Versatility

Task Versatility

Real World Reinforcement Learning

Experimental evaluation for BaRiFlex supporting real-world reinforcement learning. The hand is used to learn to pick up a cube from trial and error. The gripper collides with the surface a total of 49 times without any damage eventually achieving a 100% success rate at the task. The video shows an example of collision and an example of a successful grasp. This experiment shows that BaRiFlex can withstand collisions that are typical in a real-world reinforcement learning training.

RL Test

RL Test Result


Design Concept

(a) The rigid 4-bar linkage mechanism transmits the torques from the direct drive actuator and the collision forces back to it, enabling (b) smooth back-drive motion, and facilitating (c) parallel precise grasping when combined with the underactuated fingertips. The torsional springs at the fingertips further enhance robustness by absorbing collision forces. The inner linkage is constructed with a Fin-Ray structure of soft 3D printed material that yields (d) compliant grasping with adaption to the objects’ shape, increasing BaRiFlex’s grasp versatility. The design is simple, with only one actuated DoF and no gearbox, cost is under 500 USD, and is manufacturable in one day with two 3D printers.

CAD design for BaRiFlex


Design Specification

Specification Value Unit
Weight 750 g
Size 125 x 255 x 83 (W x H x T) mm
Max Stroke 200 mm
Continuous Force
11 N
Rnage of Motion 86.5 Deg
Speed 1440 Deg/s
Closing time 0.18 Seconds
Rated Torque 0.6 Nm
Gear-Pinion Ratio 1.54 (37T-24T) -

Grasping Versatility

The target objects are placed at five locations (up, down, left, right, and center with 5mm distance) and with four orientations (15, 30, 45 degrees) to evaluate the tolerance of the grippers to inaccuracies in poses. We test multiple YCB objects with different shapes, sizes, surfaces, and weights. BaRiFlex demonstrates a superior grasp versatility in all cases, especially with objects that require pinch grasps or conformity, thanks to its rigid-flexible design.


Fin-Ray Soft


GV Result


Task Versatility

BaRiFlex is mounted on a portable device and used by a human to perform multiple tasks such as opening heavy door, opening fridge door, catching a fast moving object. Through this human-controlled manipulation, we empirically evidence the high task-variability enabled by BaRiFlex, which covers a significant fraction of possible tasks in household domains.

Open Door-1

Open Door-2

Open Fridge

Catch the ball

Robustness Test

An accurate linear stepper motor presses on the tested hand (Rigid, 95A Fin-Ray Soft, 87A Fin-Ray Soft, and BaRiFlex). The reactive forces corresponding to different collision distances are recorded (Right). BaRiFlex exhibits the highest compliance, being able to absorb more impact forces, which facilitates interactions and learning in unstructured environments.


Soft 95A

Soft 87A


Robustness Test Result

Robustness/Compliance Result

Durability & Precision Test

The test involves the gripper’s fingertip pressing a dial indicator with a resolution of 0.001 inches, 25 times. We measure the actual displacement of the dial indicator. As can be seen in the graph (rightmost figure), the variance in the actual displacement is very low (maximum deviation is 0.0889 mm), demonstrating high precision of BaRiFlex.



Precision Result

Precsion Result


	title = {BaRiFlex: A Robotic Gripper with Versatility and Collision Robustness for Robot Learning},
	url = {},
	doi = {10.48550/arXiv.2312.05323},
	publisher = {arXiv},
	author = {Jeong, Gu-Cheol and Bahety, Arpit and Pedraza, Gabriel and Deshpande, Ashish D. and Martín-Martín, Roberto},
	month = dec,
	year = {2023}}