100,000 dominoes made into Super Mario, this robot completed in 1 day

Of course, the Dominator robot is not the result of Mark Rober himself. He said that it is the culmination of years of work by their four-person team, which has experienced many failures before the current Dominator.

This is not the first work by Mark Rober, who has created a dart board that can automatically track darts, and a basketball stand that can automatically adjust the angle of the backboard according to the flight path of the basketball, allowing you to throw darts or shoot. When you can hit a hundred hits. These works made him popular on YouTube.

The Dominator story begins in 2019.

At the time Mark Rober was giving a talk in the San Francisco Bay Area. In his speech, he talked about how he had always wanted to build a robot that could build tons of dominoes and fill an entire stadium overnight. But he made some attempts without success.

This idea attracted the interest of Alex Baucom, who has experience developing Robots and usually builds robots as his biggest hobby, so he immediately got in touch with Mark Rober.

In addition, the presentation also attracted Stanford mechanical engineering students John Luke and Josh. The four hit it off and a Domino robot team was born.

The creation of Dominator wasn’t all smooth sailing, and the team found three major challenges in making it a reality when they first started work.

The first challenge is that the scale of the initiative is so large that it exceeds what was originally envisioned.

Mark Robert wants the robot to be able to place at least 100,000 dominoes, and initially they envisioned the robot carrying all the dominoes and placing them one by one. But the team did calculations and found that 100,000 dominoes weighed 850 kilograms and would take at least three and a half days to place even at a speed of 3 seconds per block. So they had to change their initial thinking and use a tray to hold some of the dominoes, and then put the dominoes in the correct position.

The second challenge is to be extremely precise in positioning the robot.

The robot needs to be able to place the dominoes close enough for the dominoes to fall and then knock down other dominoes. But don’t put the dominoes too close, otherwise, the dominoes may be knocked down during the placement process and the previous efforts will be lost. At the same time, the robot itself must have accurate positioning, otherwise it will knock down the already placed dominoes.

The team initially tried to use ultrasonic sensors to determine the relative position of the dominoes to the robot, but that proved unsuccessful. The team later tried the Marvelind indoor “GPS” sensor and found that it didn’t work as well.

The third challenge is to keep the robot on an accurate path of movement.

Even if the problem of robot positioning can be solved, if it cannot control its own movement, it still cannot place the dominoes accurately enough, and it is easier to knock the dominoes down. The team decided to simplify the problem with an omnidirectional drive system that allows the robot to perform small corrective movements to align with the dominoes.

They bought a Mecanum wheel kit to test, built their first small Robotic prototype, and found that the idea worked. But the team quickly found that the first prototype they built was too small to be of any real help, so they built a larger robot and replaced the McCanham rotation with a real one. Omni wheel.

This new prototype helped them develop the first motion system and control loop, and did many of the groundwork.

When the team decided to transition from a prototype to a real robot, it found that several difficulties still awaited them.

First they simply don’t know how to choose the right motor for the robot. The team initially tried stepper motors, which turned out to be too small to carry the required loads and allow the robot to move at high speeds.

They then tried DC motors and found that such motors could not provide enough torque to control the robot, and sometimes the wheels of the robot would slip, causing the robot to spin like crazy.

Mark Rober had to contact other people for help, and finally, under the guidance of professionals, they chose ClearPath servo motors (servos motors) to drive the robot. Using this motor can precisely provide a large amount of torque to better control the movement of the robot and ensure that the wheels do not slip.

While choosing the motors, they also decided to switch the robot’s main computer from an Arduino to a Raspberry Pi. Using a Raspberry Pi makes the sensor latency much lower, but also complicates the code.

The team encountered some challenges in porting the software needed for the robot, such as the overall system becoming more complex and the need to relearn how to build the system in C++. But eventually they completed the main computer replacement.

After the above problems are solved, the robot can finally achieve stable and controllable driving driven by the new software system. Next, the team needs to add a tray to the front of the machine for placing dominoes.

They wrote a simple program to control the robot’s actions to place and return the dominoes, and the device started to work very smoothly. Alex Baucom called it one of the few parts of the entire project that didn’t have major problems.

Based on the test results of an early prototype, the team plans to use the Marvelmind sensor to position the robot because it already has sub-centimeter accuracy. If that’s not enough, they also plan to add ultrasonic sensors to help detect dominoes.

But things didn’t go so well here, and they found that both plans still didn’t work, and the dominoes were so small that the sensors sometimes couldn’t detect them effectively.

After several days of trying, they decided to ditch the ultrasonic sensors and tweak the location of the Marvelmind sensors, before adding ground-facing cameras to the robot to detect markers on the ground. After testing and debugging, it is proved that this scheme is effective.

Finally, all the parts of the robot are ready, and the dominoes can be placed.

From the outside, the main body of this robot consists of a frame with a steel base plate, and the drive system is mounted on the base plate. In front of the robot is a large tray capable of placing more than 300 dominoes at a time. On top of the tray is a 3D-printed funnel to make it easier for the dominoes to fall into their corresponding grids. A large counterweight has been added to the back of the robot to ensure enough friction between the rear wheels and the ground to avoid slipping.

Before the domino placement work began, the team also challenged domino artist and YouTuber Lily Hevesh. In the end, the robot did all the work in just 24 hours, putting the 100,000 dominoes in their correct positions. Alex Baucom called it a new world record.

With a crazy dream of filling a stadium with dominoes, Mark Rober and three other partners worked hard for two years to create a robot like the Dominator.

While the job doesn’t seem complicated, just having the robot carry the dominoes and put them in the correct position, it’s been incredibly difficult to accomplish. Achieving this work requires not only the robot to be able to accurately locate, but also to maintain a precise walking route, and the placement process must remain absolutely stable, otherwise any slight shaking will “annihilate” the already placed dominoes.

In the past two years, Mark Rober’s team encountered difficulties one after another, and solved them one by one. In the process, they gained not only the satisfaction of the realization of their dreams, but also valuable practical experience. These practical experiences can also provide reference value for handling robots that have been used in factories, warehousing and logistics and other fields.