Robot Hands with Built-In Vision-Based Tactile Sensors

FingerVision Inc.

Robot Hands with Built-In Vision-Based Tactile Sensors
FingerVision's robot hands

FingerVision's robot hands

Just like a human—robot hands that gently hold workpieces

While conventional tactile sensors have developed using expensive materials and advanced technology, they have not achieved the multitasking capabilities comparable to human hands or durability. With its simple design of cameras embedded in the hand and transparent silicone skin, FingerVision’s robot hands analyze real-time images to detect the hardness, shape, and slipperiness of workpieces being touched. This allows to adjust the hand movements based on the state of the workpiece it is holding. Consequently, the FingerVision robot hands have successfully automated tasks that were previously difficult, such as gently handling and assembling different types of food into containers and gripping cables, and are rapidly expanding the range of applications.

Slip distribution and force distribution

Slip distribution and force distribution

Handles a wide range of workpieces, including soft and irregularly shaped items

The touch-based force sensing of conventional tactile sensors requires specific instructions for grip force for each workpiece, presenting challenges in adapting to handle diverse items, gripping positions, irregular shapes, and deformable workpieces.

Incorporating both tactile and vision sensors, FingerVision's robot hands provide multiple modalities, including not only force distribution (in x, y, and z directions) but also slip distribution, weight of the object, gripping position and posture, shape, texture, and presence of foreign objects. This allows it to handle a wide range of workpieces, including soft and irregularly shaped items, making it suitable for various applications.

The camera-based nature of the sensor results in an extremely simple hardware mechanism consisting of a camera and silicone hand, making it cost-effective and economical. External forces are absorbed by the skin (silicone), physically isolating the electronic components (camera), which fundamentally enhances durability.

FingerVision is the world's first tactile sensor based on a camera.
Picking up strawberries

Picking up strawberries

Robotic system applications utilizing vision-based tactile sensors

Not only do we sell hands with built-in vision-based tactile sensors, but also build advanced robot systems utilizing our cutting-edge vision-based tactile sensors. We have successfully automated the process of gently handling and assembling different types of food into containers at factories, and these systems are now operational in multiple facilities. There is a public growing interest in automating processes that handle a variety of soft items. We are capable of designing and producing automation systems for various tasks, including preceding and subsequent processes. These tasks include gripping workpieces like wire harnesses and flexible printed circuits, or FPCs, (e.g., cable insertion), as well as picking and placing delicate items like fragile strawberries.
Gently handling and assembling food

Gently handling and assembling food

Examples of application

By commercializing visual-tactile sensors, we strive to expand the application areas of robots and machines in real-world settings, aiming to solve various societal challenges.

An example of application is in the food processing industry, where we have automated the assembly of bento boxes, which frequently undergo product renewals and require gentle and flexible handling, using general-purpose robot systems equipped with tactile sensors. Using the FingerVision robot hand's feedback control and label information read via the hand's camera, it is possible to sort the items automatically when the hand grab them.