Creating robots ever closer to humansthis is one of the objectives of modern robotics. Some bipedal robots (like Atlas, from Boston Dynamics) or quadrupedal robots (the little Spot, from the same company) master locomotion with remarkable agility, they do not strictly speaking have a face. Engineered Arts created the humanoid robot Ameca in 2021, but its facial expressions and facial expressions remain disturbing.
And for good reason, robotics comes up against the famous “ uncanny valley » when it comes to recreating human faces. A term popularized by Japanese roboticist Masahiro Mori in the 1970s that describes a specific phenomenon: the more a humanoid robot resembles a human, the more likely we are to feel a sense of unease, or even repulsion, when its imperfections become apparent. A team of Japanese researchers today proposes a new approach to overcome this obstacle, detailed and published on December 20 in the magazine Journal of Robotics and Mechatronics. But, as you will see, it’s not yet won.
Waves and emotions: the magic formula for more expressive robots
The innovation developed at Osaka University is based on a fundamental principle: considering facial expressions no longer as pre-programmed sequences, but like an interweaving of dynamic waves. This method, called “ dynamic arousal expression » (which could be painfully translated as “ dynamic excitation expression system “) marks a departure from conventional techniques aimed at making a robot’s face more natural.
The researchers broke down each facial manifestation – from yawning to blinking to breathing – into individual waves. These elementary components are then associated with precise physical parameters : the opening of the lips, the movement of the eyebrows, the tilt of the head. The originality of this technique lies in the use of a continuous emotional spectrum, ranging from “ sleepy ” has “ excited », which modulates the amplitude of these waves.
A complex synthesis of facial movements
Let’s take the example of a state of drowsiness: rather than simply activating a “ sleepy expression “, the system generates several simultaneous waves. A first wave controls the breathing rate, which becomes slower and deeper. A second modulates the opening of the eyelids, creating longer and more frequent blinks. A third influences the position of the head, which tilts slightly. These different waves combine and influence each other, creating a natural and evolving expression.
This mathematical approach, inspired by wave physics, allows the robot to navigate smoothly between different emotional states. If the android goes from drowsiness to excitement, the waves do not change suddenly, but gradually transform.
By considering facial expressions as dynamic phenomena rather than static sequences, the result, as you can see in the video below, is generally successful. However, it is far from perfectbut we will come back to that shortly.
As Koichi Osuka, lead author of the study, explains: “ This advance in the synthesis of dynamic facial expressions paves the way for robots capable of complex facial movements, exhibiting more vivid expressions and mood changes in response to their environment “. The results demonstrate almost instantaneous responsiveness of the system, freeing programmers from the need to individually choreograph each emotional state.
The end of the uncanny valley?
However, let’s not kid ourselves, this baby robot still a bit disturbing to watch. If the facial movements gain in naturalness and fluidity, certain elements persist in reminding us of the artificiality of the whole. Rigid movements that are still too mechanical, strange transitions between sequences of expression and above all an abysmally empty look. The eyes, in particular, with their sometimes erratic movements and their synthetic appearance, keep this robot well and truly in the valley of the uncanny.
The wave approach to facial expressions nevertheless represents notable progress compared to traditional pre-programmed reactions. Modeling emotions as a continuous spectrum of interacting waves will ensure a promising theoretical and technical basis for future, more successful robots.
Why is a robot’s expressiveness so important? Because an android’s ability to manifest emotional states in a fluid and natural way is a key element for its social acceptance. If we are ever to move towards a society where humanoid robots are part of the landscape, this factor is essential. Another path is also possible: stay away from any attempt to imitate the human face. We could have robots that could communicate their emotional states through other meanssuch as color changes, sounds or movements.
- Researchers at Osaka University have designed a robot whose facial expressions use dynamic waves to simulate emotions.
- This robot, despite notable progress in fluidity and expressiveness, remains disturbing because of certain artificial elements such as the eyes.
- Robotic expressiveness is essential for social acceptance, although non-human alternatives, such as light or sound signals, remain feasible.
