New humanoid robots learn to see and feel like a human
Sophisticated sensor and projection devices will make humanoid robots a natural part of homes and work places.
New humanoid robots learn to see and feel like a human
The humanoid robot’s moment has arrived.
We know this, not because big technology companies are suddenly able to stage Tiktok video-friendly performances of robots dancing or mimicking a human partner’s movement. That’s hype: the reality of how humanoid robots are going to perform is actually far more interesting.
The reason for the worldwide wave of interest in humanoid robots is because of a new alignment between the robot’s technical capabilities – made possible for the first time by AI – and the huge unmet demand for automation in spaces and places that are designed for humans, not for machines.
To be humanoid, these new robots will need to share many of the characteristics of a human.
This means, of course, to have a human-like skeleton and the ability to walk on two legs. But it means more besides: successful humanoid robots must also behave like humans, to have a comparable sense of touch and sight, to move around our environment like a human does, and even to make emotional connections.
And it is here, in putting the sensitivity of a human into a humanoid robot, that digital photonics technology from ams OSRAM has a valuable part to play. According to a 2026 Market and Technology Trends report from Yole Group, this ‘sensory awakening’ is a key factor underpinning growth of the humanoid robotics market value from $0.6bn in 2025 to a forecast $51bn in 2035. According to the Yole report, industry and logistics are the first markets in which humanoid robots have been deployed. Humanoids are also beginning to find uses in professional services, such as shopping guidance, and in hazardous environments such as nuclear power plant inspections. Consumer applications will be the final domain for robots to master, as, says Yole, ‘in scenarios involving extensive human-computer interaction, the capability of humanoid robots to operate stably and predictably still needs to be improved.
The human edge: dexterity, mobility, sensitivity
The automation enabled by robotics has produced great advances in output and efficiency in factories, warehouses and farms, environments which are suitable for mobile machines such as automatic guided vehicles (AGVs). These environments’ flat surfaces are suitable for robot wheels, spaces are demarcated separately for humans and for robots, and automation is limited to a range of functions for which the robots are specifically designed.
But this current version of robotics has failed to touch large areas of human activity, because homes, hospitals and other spaces are made for use by humans, not by wheeled machines. To bring automation to these spaces, the robot must be able to:
- Climb up and down stairs and steps
- Move safely around spaces that are filled with people and objects such as furniture. A difficult challenge for a robot, for instance, is knowing how to stand next to a person without being so close as to make them feel uncomfortable, or how to touch a person without using too much or too little force.
- Manipulate tools and objects that are made for human hands. The simple act of opening a drawer, for instance, requires a remarkable combination of sensing and kinetic capabilities.
The vision for humanoid robots is that they can inhabit these inherently human spaces, and provide there the benefits of automation, in functions such as cleaning, personal care, and materials handling.
The technologies of AI and machine learning are enabling this vision to become reality. The humanoid can now for the first time apply the cognitive capability required to identify objects, understand and implement instructions about complex tasks, and respond to natural language.
So AI gives the humanoid robot human-like cognition. Optical technology enables it to operate effectively in the human environment, and in a way which is acceptable to people.
Digital photonics for advanced sensing and communication
The problem of humanoid robot motion has largely been solved, as the popular videos of dancing robots show. But humanoid robots do not just need to be able to move on an empty stage: to do useful work, they must operate in crowded spaces alongside humans, and perform functions which are intuitive for a human, such as picking up an egg from a kitchen table without breaking it, but which are technically challenging for a machine.
This means that the skills of a humanoid robot will need to include the ability to:
- Navigate
- Feel
- See
- Communicate
- Act
This is where intelligent optical chips – the digital photonics technology in which ams OSRAM specializes – can provide unique capabilities in some cases, and in others complement familiar components such as video cameras and display screens.
In humanoid robot navigation, direct time-of-flight (dToF) infrared (IR) ranging sensors provide a precise measurement of the distance from objects, so that the robot can avoid collisions with people or objects. The latest TMF8829 dToF sensor generates a ‘picture’ of the space around the robot made up of 48x32 pixels – detailed enough to distinguish familiar objects such as chairs, tables and people.
In new implementations of this ranging technology, the dToF sensor’s output is overlaid on a vision camera’s image and fed to AI, which can recognize visible objects. This means that the robot can create a 3D map of its environment, identify the objects in it, and steer a safe course around it while maintaining an appropriate distance from people and pets.
The use of dToF sensors also enables robots to do the real work of manipulating objects safely in the presence of people. Much less expensive than vision cameras, dToF sensors can be deployed at multiple points on the hands of the robot, providing real-time measurement of proximity, and ensuring that a limb driven by a powerful electric motor does not collide with people or objects.
Feeling is also a critical element of the human experience. In a humanoid robot, tactile sensing can achieve awareness and responsiveness which goes further than the human sense.
Many automation tasks that a future humanoid robot might perform – food preparation is an example – require precisely calibrated force to be applied to objects. Here, new optical force sensors provide a compact and robust solution. A miniature, fully integrated sensor module, positioned directly behind a surface such as robot ‘skin’, detects deformations as small as 1 micron, enabling precise measurement of force. The optical force sensor can work in any surface material, including metal, as long as it allows for sufficient surface deformation and reflects near infrared (NIR) light.
Mounted behind semi-transparent surface materials such as some plastics, the sensor can even recognize approaching objects or basic gestures over distances up to 30cm. Compared to other resistive, capacitive or piezo-electric methods for measuring force, the optical method is more robust, more reliable, and much simpler to integrate.
The lack of any electrical connection to the surface material makes it safer – it is immune to EMI or ESD disturbance. The small form factor of the optical force sensor and its scalability allow integration for example into every finger of a humanoid robot’s hand, enhancing its tactile sensing.
Robot manufacturers can also use these sensors to implement innovative and stylish designs for control elements such as buttons: their optical measurement principle means that they are easy to sanitize, and they function in all conditions, including in harsh environments, and when operated with gloves.
Seeing is the most important way in which sighted humans sense the world. For the humanoid robot, this requires a capability both to see and to be seen.
The robot’s visual perception of the world, provided by cameras, may be supported by light sources to illuminate the scene. As a photonics powerhouse, ams OSRAM offers the industry’s widest range of LED options at visible light wavelengths, including white and color LED emitters with many wavelength and package options. In addition, ambient light sensors enable the camera’s operation to be tuned to match the scene’s lighting conditions.
In this way, a humanoid robot’s visual system operates similarly to that of a human. But the visual acuity of a humanoid robot can be enhanced beyond the ability of human eyes with specialized infrared (non-visible) LEDs. Structured light emitters, which project a pattern of dots on to a surface to reveal its contours, enable a humanoid robot to ‘see’ even minute deformations in a surface such as a floor. This will, for instance, help a bipedal robot to maintain its balance as it walks on uneven surfaces.
A humanoid robot can be designed to see what humans cannot in another way as well. Active spectroscopy solutions from ams OSRAM configure low-cost combinations of emitters and photodetectors to provide valuable application-specific spectral sensing capabilities.
Spectral sensing is a technique for detecting and recognizing the distinctive spectral signature of materials when illuminated by visible or IR light. A humanoid robot which performs active spectroscopy can, for instance, analyze the material composition of objects such as fabrics, or detect invisible gaseous pollutants in the air.
The power of vision applies to people who are adjacent to a robot, as much as to the robot itself. This is why the broad range of ams OSRAM LEDs is important to this market: the portfolio of white and color emitters enables the humanoid robot to produce a multitude of visual signals, including task-related information signals, and warning signs which mark a danger area around the robot which people should not enter. The co-ordination of optical signals is supported by the Open System Protocol (OSP), a standard interface for connecting and controlling multiple LEDs and other devices, backed by OSP-compatible LED drivers.
Optical signaling can even be built into the ‘skin’ of a humanoid robot using the new ALIYOS™ technology for ‘light out of nowhere’. This LED-on-foil technology enables the manufacturer to make an emitter an integral part of the surface of a robot, so that the skin itself functions as a light emitter.
The visible light from emitters is not only a means to see and be seen: it also provides an important mode of communication for humanoid robots. At its simplest, this can be a pulsing band of colored light projected on to the floor to indicate to people the path which a humanoid robot intends to follow through a space.
New types of microLED emitters can communicate more with light, by projecting rich patterns on to the surface in front of a robot. The EVIYOS™ platform is a system for controlling arrays of thousands of microLEDs individually. These microLED ‘pixels’ can project complex patterns on to nearby surfaces: words, icons, pictures, or other symbols to convey messages to people near to a robot.
Color LEDs can also be used for non-verbal but human communication: for instance, a red glow in a humanoid’s face to indicate embarrassment when it has made a mistake. An important challenge for humanoid robots is to be accepted by people in their homes or places of work. The ability to form a kind of emotional bond with a robot could turn out to be as important in this as the practical and economic advantages of robotic automation.
As human as a machine can be
The automation capabilities of humanoid robots – their ability to replace or supplement human effort in factories, offices and homes – will affect how far and fast their adoption goes. But beyond the purely functional case for humanoid robots, the ability to move, act and communicate in a way which aligns with human habits and preferences will dictate whether this technology is accepted into the mainstream of human society.
Optical sensing and signaling technology has a huge role to play in the social acceptance of humanoid robots. Early implementations of robotic automation have given rise to feelings of hostility on the part of human workers, with instances reported of people refusing to work alongside robots, or disrupting their operation: this is due to workers’ resentment at the prospect of being replaced by faceless machines.
Humanoid robots can avoid this fate if their design can incorporate elements of humanity: the signals generated and the emotions evoked by light can play a large part in bridging the divide between people and the machines that they are asked to work with.
It is with digital photonics technology, which provides for intelligent control of light, that ams OSRAM can help manufacturers of humanoid robots to make their products both operationally effective and socially acceptable.