Put The engineers have created a lightning ping-pong robot which does not only return shots with a human speed and precision, but also imitates rotation and aiming strategies.
Built from components of a humanoid robot and powered by advanced prediction algorithms, it has a success rate of 88% in tests. Now, researchers are working to make it more mobile, with future applications in everything, from sports training to research and rescue robotics.
MIT engineers built a powerful and light ping-pong robot that makes waves in the world of intelligent robotics. This high -speed machine is not only fast – it is impressive.
At the heart of the system is a robotic arm with several articulations mounted at one end of a standard ping-pong table. Holding a regular paddle, the arm uses a network of high -speed cameras and an advanced predictive control system to follow the incoming balls. He then selects one of the many swing styles – such as toppin curls, straight readers or delicate backscin chops – to send the ball exactly where it is targeted, with just the right spin.
During real world tests, engineers launched 150 robot balls in rapid succession. The bot has returned nearly 88% of them in all types of swing. Its striking speed is even competed on high -level human players and surpasses previous robotic table tennis systems.
The MIT team is now striving to extend the robot game range, which allows it to return more types of fire from a larger area. They believe that the system could become a powerful tool for intelligent robotic training and simulation.
Looking beyond table tennis, the same technology could help stimulate the agility and responsiveness of humanoid robots. Applications may include research and rescue missions or in any environment where rapid real -time reactions are essential.
“The problems we solve, specifically linked to the interception of objects very quickly and precisely, could potentially be useful in the scenarios where a robot must perform dynamic maneuvers and plan where his final effector will meet an object, in real time,” explains David Nguyen, a student graduated from the MIT.
Nguyen is a co-author of the new study, as well as the student graduated from the MIT Kendrick CANCIO and Sangbae Kim, Associate Professor of Mechanical Engineering and Head of the MIT BIOMETICS ROBOTICS LAB. Researchers will present the results of this experiences in an article in the International Conference of the IEEE on robotics and automation (ICRA) this month.
Building robots to play ping-pong is a challenge that researchers have met since the 1980s. The problem requires a unique combination of technologies, including high-speed industrial vision, fast and agile engines and actuators, precise control of manipulators and a precise prediction in real time, as well as planning of the game strategy.
“If you think of the spectrum of control problems in robotics, we have the manipulation at one end, which is generally slow and very precise, as to pick up an object and ensure that you enter it. At the other end, you have locomotion, which consists in being dynamic and adapting the disturbances in your system,” explains Nguyen. “Ping pong is between them. You always do a manipulation, in that you need to be precise to hit the ball, but you have to hit it in the 300 milliseconds. Therefore, it balances similar problems of dynamic locomotion and precise manipulation. “
Ping-pong robots have gone a long way since the 1980s, more recently with conceptions of Omron and Google Deepmind that use artificial intelligence Techniques to “learn” previous ping-pong data, in order to improve the performance of a robot against an increasing variety of strokes and plans. These conceptions have proven to be faster and precise to rally with intermediate human players.
“These are truly specialized robots designed to play ping-pong,” explains CANCIO. “With our robot, we explore how the techniques used to play ping-pong could result in a more generalized system, such as a humanoid or anthropomorphic robot which can do many different and useful things.”
For their new design, the researchers have changed a light and high power robotic arm that Kim’s Lab developed as part of MIT Humanoid – a two -arms robot that is the size of a small child. The group uses the robot to test various dynamic maneuvers, including unequal and variable field navigation as well as to jump, execute and do backflips, in order to deploy these robots for research and resistance operations.
Each of the humanoid arms has four joints, or degrees of freedom, which are each controlled by an electric motor. CANCIO, NGUYEN and KIM built a similar robotic arm, which they adapted to the ping-pong by adding an additional degree of freedom to the wrist to allow the control of a paddle.
The team fixed the robotic arm at a table at one end of a standard ping-pong table and installed high-speed motion capture cameras around the table to follow the balls that are plugged in the robot. They have also developed optimal control algorithms which predict, depending on the principles of mathematics and physics, what speed orientation the arm must execute to strike a incoming ball with a particular type of swing: loop (or topspin), train (right) or hop (backspin).
They implemented the algorithms using three computers which simultaneously treated the images of the camera, estimated the state in real time of a bullet and translates these estimates in command so that the robot engines react quickly and take a swing.
After having bounced consecutively 150 balls in the arm, they found the Hit Robot rate, or precision Returning from the ball, was almost the same for the three types of swings: 88.4% for loop strikes, 89.2% for chops and 87.5% for discs. They have since set the robot’s reaction time and found that the arm was hitting the balls faster than existing systems, at speeds of 20 meters per second.
In their article, the team reports that the robot strike speed, or the speed at which the paddle reaches the ball, is on average 11 meters per second. Advanced human players are known to return balls at speeds between 21 and 25 meters in second. Since writing the results of their initial experiences, researchers have further changed the system and have recorded striking speeds up to 19 meters per second (around 42 miles per hour).
“Part of the objective of this project is to say that we can reach the same level of athletics as people,” says Nguyen. “And in terms of typing speed, we really become very close.”
Their follow -up work also allowed the robot to aim. The team has incorporated control algorithms into the system that predict not only how to hit an incoming ball. With its latest iteration, researchers can define a target location on the table and the robot will hit a bullet in the same place.
Because it is fixed at the table, the robot has limited mobility and scope, and can above all return bullets that arrive in a crescent -shaped area around the median line of the table. In the future, engineers plan to fuck the bot on a portico or wheeled platform, allowing it to cover more from the table and return a greater variety of shots.
“A great thing about table tennis is to predict the spin and the trajectory of the ball, given the way your opponent has struck him, which is information that an automatic ball launcher will not give you,” said Canco. “A robot like this could imitate the maneuvers that an opponent would make in a game environment, in a way that helps humans to play and improve.”
Reference: “High -speed robotic table tennis swing using light equipment with predictive control of the model” by David Nguyen, Kendrick D. Cancoo and Sangbae Kim, May 2, 2025, International IEEE conference on robotics and automation (ICRA).
Arxiv: 2505,01617
This research is supported, in part, by robotics and the AI Institute.
Never miss a breakthrough: Join the Scitechdaily newsletter.
Nicole Scherzinger won the Tony price For the best leading actress in a musical Sunday,…
Three months after the developer of Chinese artificial intelligence Deepseek overturned The world of technology…
When the thermometer starts to climb this summer, your electricity bill will probably climb with…
"Maybe a happy ending," A moving Broadway musical about two thrown robots that make a…
The decline of global investors against long -term public debt transforms what would normally be…
The former Indycar and Nascar pilot, Danica Patrick, proudly admitted to having taken the "red…