Connect with us

Robot

Robots Can Now Sprint & Will Soon Compete With Professional Athletes!

Published

on

Cassie the Robot Blue

Cassie, a robot designed at Oregon State University, is officially the fastest bipedal robot on the planet after running the 100m sprint in 24.73 seconds. The robot started in a standing position and naturally sprinted along OSU’s Whyte Track and Field Center. Cassie stayed in its lane and didn’t even stumble.

This OSU robot isn’t just a sprinter. Cassie also successfully completed a 5k run in 53 minutes as it explored OSU’s campus by itself during a single battery charge. It’s the first bipedal robot that successfully used machine learning to run outdoors.

Cassie Cost $1 Million to Build

Cassie is the brainchild of Oregon State robotics professor Jonathan Hurst. The professor received a $1 million grant from the Defense Advanced Research Projects Agency (DARPA). Then, just over a year later, this impressive bipedal robot was born.

Cassie is now being engineered by Agility Robotics, which is associated with OSU. The current human record for the 100m event is 9.58, so the OSU has a long way to go until it can beat people. Cassie is also considerably slower than the average person who can sprint 100 meters in 20 seconds or less.

However, robots are slowly getting faster. Within five years, we should see a bipedal robot that can break Usain Bolt’s world record.

OSU Takes a New Approach With Reinforcement Learning

Boston Dynamics’ Atlas robot may be capable of amazing agility and physical stunts. However, the robot is coded by hand and isn’t able to spontaneously perform complicated physical tasks in random environments with unseen challenges. Cassie, on the other hand, can! 

The roboticists were able to simulate trial and error through reinforcement learning. Thanks to this cutting-edge AI-powered technique, Cassie can run and jump without being coded to perform the movements. 

The OSU showed off its athleticism by having it perform broad jumps, high jumps, and even sprint 400 meters in under three minutes. Reinforcement learning is basically trial and error and mimics the way humans learn skills. Cassie is rewarded if it performs the movement correctly and punished if it doesn’t. Thanks to this innovative approach, the robot can adapt to new scenarios and come up with the right movement based on the situation.

The goal of this latest research is to get Cassie moving dynamically, making the robot move like a human. Instead of waiting years to teach the robot how to jump and sprint, the OSU team could compress the training time down to just a few weeks through the use of simulation. In Cassie’s mind, the simulation training is just as effective as running on a track.

Initially, they uploaded data of humans and animations walking and jumping up and down and instructed Cassie to copy these movements. It didn’t take long for the neural network which controls the OSU robot to master the basic movements.

Training Cassie for Dynamic Movement

Now that the base model was complete, the researchers started adding more advanced commands. Again, Cassie was rewarded when it performed the new skills correctly. Once Cassie was performing well during the simulation, it was time to add an element of randomness to the training.

Through a training method known as task randomization, Cassie had to perform tasks while dealing with unexpected challenges. The researchers would push and pull the robot at random intervals, and Cassie was expected to maintain its balance and continue walking. It didn’t take long before the OSU robot was leaping 1.4m in a long jump.

The OSU robot doesn’t have any sensors and cameras, and according to researchers, this actually makes it easier for the robot. The next challenge will be equipping Cassie with cameras and using the same reinforcement learning method, which hasn’t been done before.

Ultimately, the team wants to create robots that have all of the capabilities of humans, meaning they can perform complex tasks, move around with ease, plan and think, and interact with the entire world. When that finally happens, we’ll could entire companies and workplaces staffed by robots.

That reality is a long way to go, but expect robots to become increasingly mobile over the next few years. A humanoid robot Olympics is definitely on the horizon within the next 10 years, especially for simple events like the 100m sprint. It’s inevitable that a bipedal robot like Cassie will eventually break Usain Bolt’s record and further push the horizon of Machina Sports!

Sport Enthusiast, Builder of brands, and proud founder of Machina Sports, dedicated to pioneering the fusion of human athleticism with cutting-edge technology. Committed to creating a global platform and brand that celebrates the excitement and innovation inherent in Machina Sports while engaging a diverse community of enthusiasts and athletes worldwide.

Robot

Matches in the National Havoc Robot League Have Reached a New Level

Published

on

NHRL Banner

The 2024 National Havoc Robot League Season (NHRL) is powering along as 3lb, 12lb, and 30lb combat robots from across the globe battle to become NHRL’s 2024 World Champion. 

Robot combat is the most popular Machina Sport in the world and is also the most accessible. The sport is simple: two robots enter the cage, and only one leaves. The goal is to destroy the opposing robot and render it unable to continue to battle.

Robot combat has been around since the 90s with popular leagues like Battlebots and Robot Wars. The NHRL is continuing this tradition with seven yearly events held in the House of Havoc, located in Norwalk, CT. The NHRL is now the biggest robot combat league in the world, with shows selling out and hundreds of thousands of fans around the world tuning in to broadcasts.

Builders use cutting-edge technology and ground-breaking ideas to create new and unique robots that are capable of destroying an opposing robot within seconds. The NHRL offers fantastic cash prizes, and the world champions get to lift the coveted Golden Dumpster trophy.

So far, two NHRL events have taken place, and there are still five more to go. Let’s look back at what has transpired this season and see which robots are likely to become world champions!

The First NHRL Event of the Year

The 2024 NHRL season started with a bang on January 20, 2024, with a special event reserved solely for robots who had never competed in the NHRL before. 3lb, 12lb, and 30lb bots took part in the event, and the top four from each category received an invitation to the 2024 World Championships, which will feature a $50,000 prize pool.

There were some incredibly designed robots at the event, from flamethrowers to saws to classic flipping bots and even one very nasty robot armed with a nail gun. 

As there were many new robots competing, there were a healthy number of malfunctions. Some robots just stopped working, while others exploded. Bots were able to win matches by simply surviving.

The 12lb competition was by far the most competitive. Eight fiercely designed bots entered the cage and did their absolute best to destroy their opponents, but only one emerged victorious: Questionable Choices. 

This compact robot is built low to the ground and is armed with a buzz saw in the middle. Then, just below the buzz saw are two sharp spikes that Questionable Choices is fond of ramming into its opponents. The lighting-quick robot also can get underneath opposing bots and flip them high into the air.

In the final, Questionable Choices ended up destroying the drive of Blue Marlin, a similarly designed robot, and exerting superior control.

In the 3lb bracket, Scurryfest was by far the best robot and cruised to the finals, where it defeated Repeater. Then, in the 30lb bracket, Moccasin crushed Colossal Avian.

NHRL Brings the Heat With Its Second Event of the Season

The second NHRL event of the year took place in March, and again, 3lb, 12lb, and 30lb battle bots were competing for a place in the World Championships. The 3lb final between Red Panda and Eruption was a particularly fiery affair, with sparks literally flying.

Eruption came into the fight with a 51-26 record and didn’t disappoint. Channeling Mike Tyson, Eruption launched itself at Red Panda and, within the opening seconds of the fight, had Red Panda upside down and helpless. After continuing to violently ram Red Panda over and over again, Eruption secured the victory.

It only took a little over one minute for Pramheda to KO Black Jack and win the 12lb title. In the 30lb final, defending world champion Emulsifier managed to flip Vorion over. Then, while Vorion was in a vulnerable position, Emulsifier crept and drove its buzz raw right into the stranded robot. Smoke flew out of Vorion, and that was the end of the fight.

The next NHRL is on April 20 and there will be events running monthly up until October. This will be followed by the NHRL World Championship held in November, which is a can’t-miss event as the best battle bots throw down to see who is the best on the planet!

Continue Reading

Robot

Robots on the Court? AI Ball Machine Heats Up Tennis Training

Published

on

Tennis Bll on Blue Hard Court

Ball machines are nothing new in tennis and have been around since the 1970s. However, machines that can mimic players and are powered by AI are. Volley, an innovative sports tech company, has developed a cutting-edge training device that can be used for virtually all racquet sports, including tennis, padel, and pickleball.

Instead of just lobbing balls to you at different velocities, this new-age ball machine actively analyzes your gameplay and then unleashes balls. The goal of Volley’s training device is to recreate gameplay and give players the exact type of feeding they require to improve. 

Volley’s Ball Machine Is Tech Heavy

This AI-powered ball machine comes loaded with three cameras. One camera closely tracks the ball, while another camera tracks you, recording your every stroke. The goal is to get you hitting forehands like prime Roger Federer. The third camera is located within the ball machine and gives customer support a view of exactly what’s going on in case of issues.

Volley’s creation also comes with an LED screen where you can program different workouts. The ball machine is completely adjustable, so it’s capable of hitting booming first serves as well as gentle drop volleys, and everything in between. The training device stands at 87 inches and can tilt, twist, and rotate, ensuring the ball can be launched all over the court.

You can also download the Volley mobile app, where you’ll find workout plans and different statistics and even watch yourself hitting. Another cool feature is the remote control element. Directly from the app, you can instruct the ball machine exactly where you want to bounce, allowing you to work on specific elements of your game.

Volley hopes to ditch the remote control element and allow players to instruct the ball trainer via hand gestures. The team is constantly looking for new features they can add to the ball machine.

AI Ball Machine Sells Out in Less Than Four Months

It didn’t take long for Volley’s ball machine to grow an avid following within the tennis community. It was released in September 2023 and, within a couple of months, sold out. You can find Volley’s trainer at 45 different tennis clubs dotted around the US, including in NJ, FL, MA, OH, and PA. Volley costs a pretty penny and is currently being leased by tennis clubs for up to $3,000 per month.

Racquet sports technology companies are booming as they scramble to meet the needs of players who are no longer satisfied with a static ball machine. Slinger is another smart ball machine that can track your shots and offer stroke advice, while Proton comes loaded with sensors, resulting in laser-accurate ball delivery to all parts of the court.

Volley believes it can serve the needs of pros and hobbyists who are looking to recreate gameplay. With the help of AI, the company hopes to offer a next-generation solution similar to golf simulators.

Volley’s trainer is particularly effective for paddle, which traditional ball trainers have ignored. Volley’s ball machine can create common shots in paddles, which standard trainers can’t. Expect to see this new style of ball machine popping up at tennis clubs all over the world very soon.

Volley’s Ball Machine Can’t Hit the Ball Back

Despite being powered by AI and having the skills to recreate all of your favorite shots, Volley’s trainer can’t engage in rallies. There are currently no commercially available tennis-playing robots as the technology just isn’t there yet. However, a team at Georgia Tech has built a robot that can sustain rallies.

ESTHER features two wheels and an arm that holds a tennis racquet. The wheelchair-designed robot plays tennis at a very low level and is a fan of hitting high-looping shots. However, in a feat of engineering ingenuity, ESTHER can use its cameras to identify where the ball is going to land, move into position, and then strike it. More often than not, the ball lands in the court.

Tennis robots are advancing quickly. It’s reasonable to think that within 30 years, there’ll be a robot that can defeat the best pro tennis players. Also, local tennis coaches may be a thing of the past, with players opting to be trained by robots.

Continue Reading

Robot

H1 Robot Is Now Faster Than Boston Dynamics’ Atlas, Reaching 7.38 Mph

Published

on

Humanoid Robot Sprinter in a track

Unitree Robotics, based in Hangzhou, China, has smashed the robot speed record. Their humanoid robot, dubbed H1, is two miles per hour faster than Boston Dynamics’ Atlas robot and can reach impressive speeds of up to 3.3 meters per second. 

H1 won’t win any points for his running style, but the speed shuffle is effective, and he’s currently the fastest humanoid robot on the planet. The robot clocked in at just under 7.4 mph, which was achieved on a concrete pavement. The previous record was set by Atlas at 5.59 mph. H1 even broke the record while wearing pants! 

Meet China’s Answer to Atlas

Build-wise, H1 is comparable to a very slender human, standing at 71 inches tall and weighing just 100 pounds. The robot is loaded with 3D sensors and a depth camera, which gives him superhuman vision.

This Chinese robot has a completely hollow torso where you’ll find all of the electrical wiring, which acts as H1’s veins. While H1 does have arms, the humanoid robot is currently running around without hands. Unitree Robotics is trying to build hands that can rival Atlas’. The robot won’t be very effective if it can’t lift and put things down.

While Unitree’s H1 robot isn’t nearly as advanced as Boston Dynamics’ Atlas robot, the Chinese company is trying to compete on price. You can purchase H1 for as little as $90,000, while Atlast starts at $150,000 for the very basic version.

There’s a huge demand for humanoid robots as companies all over the world dream of employing them to take over every role imaginable, from mechanic to nurse to builder. However, we’re a long way away from automated workplaces staffed by H1s and Atlases.

What Can H1 Do?

H1 shows incredible balance. In a recent video, it was shown being kicked repeatedly from different angles, but the robot was consistently able to adjust its footing and never came close to falling over. The robot also could sense the tester near him and briefly paused before continuing to walk.

This Chinese robot has an odd gait with overly bent knees, but it’s proven effective. H1 can navigate stairs like a pro and even walk down backwards and while spinning. There are numerous people who would struggle with that feat!

After the engineers attached hand molds, the humanoid robot can now pick up certain objects and hold them securely. It’ll be interesting to see what the robot can do when it has fully functional hands.

H1 can jump as high as the average person and was even out jumping the tester. The Unitree Robotics creation even displayed impressive all-body coordination by dancing. Both arms and legs were moving at the same time and keeping up the beat. While the little jig won’t win any dancing competitions, it shows the robot is developing quickly.

In an even more incredible feat, the Unitree robot performed a standing backflip and stuck the landing. It didn’t leap very high in the air and stumbled a bit, but still, it remained standing.

Not Quite on Atlas’ Level Yet

Boston Dynamics’ Atlas is the gold standard when it comes to humanoid robots. Atlas can complete an obstacle course faster than most people and looks natural doing so. This American robot has a far more natural gait than H1 and a higher level of coordination.

Atlas can do backflips with a tuck, performing the challenging movement multiple times in a row and landing perfectly. Recently, the Boston Dynamics’ robot displayed its skills on a mock building site. Atlas was able to pick up a heavy tool bag and then clamber up multi-story scaffolding, and safely deliver the bag to the builder at the top. Then, in style did a flip off a box to celebrate his performance.

H1 may be the fastest humanoid robot on the planet, but it still can’t perform many basic tasks, especially without functional hands. The Atlas robot is far more nimble, as evidenced by its parkour tricks. It also demonstrated an ability to pick up and move heavy objects, tasks that could transform the manufacturing and building industry. Currently, neither of these humanoid robots can sprint, but it’s only a matter of time before they break the 100-meter record, too!

Read More: Cassie – The Fastest Bipedal Robot

Continue Reading

Trending