Humanoid robot drives car autonomously

Humanoid robots could serve as chauffeurs as an alternative to autonomous cars. Robot Musashi from the University of Tokyo is pursuing this approach.

listen Print view
Musashi robot in an electric car

The humanoid robot Musashi drives an electric car autonomously.

(Image: University of Tokyo)

4 min. read

Robotics researchers at the University of Tokyo have developed a humanoid robot called "Musashi" that could, in principle, drive any conventional car as a chauffeur. With their approach, the researchers want to show that it is not absolutely necessary to automate a complete car, but that a relatively inexpensive humanoid robot is sufficient.

Autonomous driving is an expensive business, as the vehicle has to be equipped with a certain amount of intelligence and all kinds of sensors. In the study "Toward Autonomous Driving by Musculoskeletal Humanoids: A Study of Developed Hardware and Learning-Based Software" (PDF), which has been published as a preprint on Arxiv, the Japanese scientists therefore asked themselves whether it would not be simpler and cheaper to use a humanoid robot as a driver in a conventional car.

To this end, they designed the Musashi robot, which resembles a human driver. As a humanoid robot, it has a movable head, a torso with two arms and legs. The hands are also based on the human model so that the robot can operate the controls in a vehicle made for humans. This also includes two feet, which have a particularly good grip so that the robot does not slip off the pedals of the vehicle.

Videos by heise

Overall, the robot was designed to function similarly to a human driver. To this end, the researchers have given it a kind of artificial skeleton that is controlled by muscle-like actuators. The robot has 74 artificial muscles and 39 joints. It also has actuators and joints in its two hands, with which it can operate the steering wheel of the vehicle as well as the car key for starting, the blinker lever and the handbrake. Unlike a human, however, the robot operates the accelerator and brake with two feet. They are fitted with force sensors that measure in six axes to provide feedback so that the robot can handle the accelerator and brake with sensitivity.

Musashi uses two high-resolution cameras built into its moving eyes for perception. This allows the field of vision to be enlarged to perceive the surroundings and also cover the two rearview mirrors, for example, so that the rear area of the vehicle can also be seen. The robot is not yet able to detect distances. The researchers are currently working on a distance detection system that works exclusively via cameras. Object recognition currently only works in daylight. Musashi cannot yet be used at night.

Empfohlener redaktioneller Inhalt

Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.

Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.

The robot uses a total of four PCs to ensure its functions. Intel NUCs are used for motor control, image and sound processing as well as a Zotac VR Go to implement environment and object recognition. All other processes are processed on an external Intel NUC.

The scientists trained the Musashi robot with the help of videos and real driving situations. The researchers first tested it on the university campus in a small electric car with limited performance. During the test drives at walking pace, the ability to reliably recognize and react to objects was tested in particular. According to the scientists, these tests were "encouraging". However, the results are not yet sufficient to send the robot into road traffic. They now want to further improve Musashi to find out whether a humanoid robot can drive more safely than an average human.

(olb)

Don't miss any news – follow us on Facebook, LinkedIn or Mastodon.

This article was originally published in German. It was translated with technical assistance and editorially reviewed before publication.