ProTac uses internal cameras inside soft robot arms to detect both proximity and touch simultaneously, replacing traditional force-torque sensors with a vision-based approach.
ProTac places cameras inside a semi-transparent robot arm. The cameras detect light changes to sense nearby objects and physical contact without external sensors.
The ProTac system, developed at the Japan Advanced Institute of Science and Technology (JAIST), embeds cameras inside the structure of a robot arm. The arm material is semi-transparent, which lets external light pass through. When an object approaches the arm, it changes the light pattern that the internal cameras see. When the arm is physically touched, the deformation of the surface creates a different visual signal. One sensing system handles both jobs. From what I can find, this is a meaningful departure from the conventional approach, where proximity sensing and contact sensing are handled by separate hardware layers. According to New Atlas, the goal is to allow robots to operate more safely around humans by detecting both approach and contact through this single unified method.
Proximity sensing typically uses infrared, ultrasound, or capacitive methods to detect objects before contact. Contact or force sensing usually relies on force-torque sensors, strain gauges, or pressure-sensitive skins applied to the robot surface. These two problems have historically needed different physics to solve, which is why they have lived in different hardware. ProTac's approach is notable because it uses a single optical principle to bridge both.
ProTac has the potential to replace dedicated wrist-mounted force-torque sensors and separate surface proximity hardware with a single internal camera system.
In conventional humanoid robot arms, the sensing stack typically includes a six-axis force-torque sensor mounted at the wrist, plus some form of proximity detection on the outer surface. These components add cost, weight, and complexity. They also introduce failure points. A wrist-mounted force-torque sensor sits in a mechanically stressed location and can be damaged during unexpected collisions. The ProTac approach moves the sensing element inside the structural arm, away from the contact surface. As far as I understand it, this could reduce exposure to mechanical damage while preserving the ability to detect both pre-contact and contact events. The trade-off is that you are now relying on computer vision and light physics rather than direct mechanical measurement.
ProTac trades direct mechanical measurement for optical inference, gaining flexibility and lower cost but introducing new dependencies on lighting, materials, and processing.
Every sensing architecture involves trade-offs. Here is what the data shows about the trade-offs in ProTac's design. The system depends on the semi-transparent arm material transmitting consistent light. That means material selection is now a sensing constraint, not just a structural one. The cameras need to process visual information in real time, which adds computation load. Traditional force-torque sensors output a direct electrical signal proportional to force, which is fast and unambiguous. An optical system has to interpret visual patterns, which introduces latency and potential ambiguity. On the other side, the design gains spatial resolution. A camera can potentially detect where along the arm something is approaching or touching, not just that contact happened. A single wrist force-torque sensor cannot tell you where on the arm the contact occurred.
In standard robot arm design, the outer shell material is chosen for strength, weight, and surface finish. ProTac adds a new constraint: optical transmission. The material must let enough light through for the internal cameras to see external events. This couples the mechanical design and sensing design together in a way that is unusual. It is a smart integration, but it also means you cannot optimize these properties independently.
A conventional force-torque sensor gives you a precise scalar reading of force and torque at one point. ProTac's camera-based approach theoretically gives you spatial information across the arm surface. For tasks like human-robot collaboration where you want to know not just that contact happened but where it happened, that spatial awareness could be genuinely useful. I am still working out the implications of this for control systems, but it seems like a meaningful capability difference.
Safe human-robot collaboration requires detecting humans before and during contact. ProTac addresses both phases, which is the core technical requirement for collaborative robot safety.
According to New Atlas, the stated motivation for ProTac is to allow robots to operate more safely around humans. This frames the design as a safety system, not just a sensing upgrade. The two-phase sensing capability matters here. Detecting proximity allows a robot to slow down or change trajectory before contact happens. Detecting contact allows it to respond appropriately when contact does occur. Standard industrial robot safety uses external cameras or lidar for proximity and relies on joint torque sensing or skin-based sensors for contact. ProTac proposes to handle both with the arm structure itself. The sources suggest this originated at JAIST in Japan, a country with both strong robotics research infrastructure and significant social motivation to develop robots that can work alongside aging populations.
ProTac reflects a broader trend toward integrating sensing into robot structure rather than adding discrete sensor modules to existing designs.
Let me break down the components of what this design philosophy implies. There is a growing research direction around what some call structural sensing or morphological sensing, where the physical form of the robot does sensing work rather than relying on bolt-on sensor modules. ProTac is one data point in this direction. Others include soft robotic skins with distributed pressure sensing and joint designs with built-in compliance for force estimation. The ProTac approach is notable because it uses existing commodity hardware, cameras, rather than requiring new materials or custom electronics. The sources do not yet indicate commercial deployment or production plans, so I want to be careful not to overstate where this is in the development pipeline. This appears to be research-stage work from JAIST, not a product announcement.
Key unknowns include performance in variable lighting, durability of the semi-transparent material, latency compared to direct force sensing, and scalability to full humanoid arm designs.
I am still learning about this area, but several questions stand out after going through the available sources. First, how does the system perform in variable or low lighting? An optical system that depends on external light passing through the arm would be sensitive to the ambient environment in ways that a strain gauge or piezoelectric sensor is not. Second, how durable is the semi-transparent material over thousands of contact cycles? Robot arms take mechanical abuse. Third, what is the actual latency from contact event to control system response, and how does that compare to direct force sensing? For safety-critical applications, milliseconds matter. The New Atlas report describes the concept and the research origin but does not provide specific performance benchmarks on these points. Those numbers would be the deciding factor in whether this approach can compete with established force-torque sensor technology in real deployments.
ProTac is a sensing system developed at the Japan Advanced Institute of Science and Technology (JAIST). It uses cameras embedded inside semi-transparent robot arms to detect both proximity to objects and physical contact, using changes in light patterns as the sensing mechanism.
A standard force-torque sensor measures mechanical force directly at a point, typically the robot wrist. ProTac uses optical inference from internal cameras to detect contact across the arm surface. It also adds proximity sensing before contact, which a wrist-mounted force-torque sensor cannot provide on its own.
Based on what the sources describe, the main potential limitations are sensitivity to ambient lighting conditions, the durability of the semi-transparent arm material under repeated contact, and processing latency compared to direct mechanical sensing. Published benchmark data on these points is not yet available in the sources I found.
Proximity sensing allows a robot to detect that a human is approaching before contact occurs. This gives the control system time to slow down or change trajectory, reducing collision force. According to New Atlas, this pre-contact awareness is a core motivation behind the ProTac research at JAIST.
From what I can find, ProTac is currently a research-stage system from JAIST. The New Atlas report from August 2025 describes the concept and its research origins but does not indicate commercial production plans or deployment timelines. It is a technology to watch, not yet a product.