The Tesla AI Day 2022 was worth the wait. From a person in a robot suit to unveiling a working prototype of the Optimus robot, Tesla has come a long way. At the event, the company not only explained the hardware and software side of the robot, but also revealed some interesting new updates and breakthroughs in its full self-driving (FSD) beta version, DOJO, and more.  

Tesla’s chief Elon Musk believes these advancements could one day make a meaningful contribution to AGI. 

The team said that Optimus is powered by the same self-driving computer that runs the Tesla cars. “This is literally the first time the robot has operated without a tether on stage tonight,” said Lizzie Miskovetz, the mechanical design engineer.

“You might have seen a very impressive humanoid robot demonstration. That is great. But what are they missing? They are missing a brain that isn’t having the intelligence to navigate the world by themselves,” said Musk. He said once the robot is fully ready and functional, it would cost much less than a car. “I would say probably less than $20,000,” he added. 

“Our goal is to make a useful humanoid robot as quickly as possible,” he added, saying that the company is looking to make the robot in high volume at low cost with high reliability. 

Musk did acknowledge that the robot is still in its early stages of development. “There’s still a lot of work to be done to refine Optimus and improve it, obviously, because it’s just Optimus version one,” Musk added, “That’s really why we’re holding this event which is to convince some of the most talented people in the world like you guys to join Tesla and help make it a reality and bring it to fruition at scale, such that it can help millions of people.”

Check out the Tesla AI Day 2022 below: 

Interestingly, the teams said that the first version of Optimus was built within six months, and they have been working on software integration and hardware upgrades over the month or so, alongside designing the next generation robot with a clear focus on the true human form. 

Comparing the robot with humans, the Tesla Optimus team said that they have stripped down the fundamental degree of freedom of the robot to 28 structural actuators and 11 hand motions. “Humans are pretty efficient at some things and not so efficient at other things,” said Miskovetz, citing how humans are inefficient even when they are resting and are burning energy. The same is not the case with robots, she added, showcasing the Tesla bot, which has many actuators, electrical systems, etc. 

“We are using that vehicle design foundation. So, we are taking it from concept through design and analysis, and then build and validate,” she added, saying that they are going to optimise cost and efficiency as it is one of the critical metrics to take this humanoid to scale, where they look to reduce the part count and power consumption of every element possible. “We are going to reduce the sensing and the wiring at our extremities,” she added. 

Further, she said that the latest generation of Optimus comes with a 2.5 kWh battery pack, placed at the centre of the torso, which is ideal for a full day’s worth of work. Also, what is interesting about this battery pack is that it has all of the battery electronics integrated into a single PCB within the pack. In other words, from sensing to fusing, charge management, and power distribution are all in one place. “We are also leveraging both our vehicle and energy products to roll all of those key features into the battery,” said Miskovetz. 

The team leverages the autopilot hardware and software for building the humanoid platform (Bot Brain). However, the requirements are quite different, noted the team. Also, it supports communication as it is equipped with wireless connectivity and audio support. “It also has hardware level security gestures which are important to protect both the robot and the people around it,” he added. 

Besides these, the team spoke in length about both hardware and software capabilities, including integrated structural control design, human-inspired design of knee joint, and optimised joint for efficient actuators, alongside the difference of powertrain between car and robot, actuator evaluation for a single joint (focussed on system cost for a cloud actuator) and commonality study of reducing the unique actuator design. 

Also, the team touched upon various actuators, their configurations, and how the hand is inspired by the biological hand, alongside talking about its specification and performance. The mechanical hands include components like clutching finger drives and adaptive fingers.  

On the software side of things, Optimus uses computer vision with the help of multiple cameras, including left pillar, fisheye, and right pillar cameras, where it understands the world based on 3D occupancy, panoptic segmentation, and YUV, alongside leveraging methods like synthetic view rendering, volumetric depth rendering, visual navigation techniques and others. 

Touching upon the challenges of balancing in the real world, the team said that it ran multiple tests in a simulation environment, including locomotion planning. The team handled destabilisation issues in the real world using motion control, alongside relying on manipulation techniques. This includes natural motion reference, online motion adaptation, and robot controllers. 

“We started doing something usable, but it is far from being useful. There is still a long and exciting road ahead of us,” said Milan Kovac, director of engineering at Tesla. He said that in the next few weeks, the team would get Optimus at least at par with ‘Bumblecee’ (the first iteration of the Tesla bot that walked onto the stage at the opening of Tesla AI Day 2022).  

Further, he said that they would start focusing on the real use cases at one of their factories, try to nail this down, and iron out all the elements needed to deploy this product in the real world. 

Conference, in-person (Bangalore) Machine Learning Developers Summit (MLDS) 2023 19-20th Jan, 2023

Conference, in-person (Bangalore) Rising 2023 | Women in Tech Conference 16-17th Mar, 2023

Conference, in-person (Bangalore) Data Engineering Summit (DES) 2023 27-28th Apr, 2023

Conference, in-person (Bangalore) MachineCon 2023 23rd Jun, 2023

Stay Connected with a larger ecosystem of data science and ML Professionals

Discover special offers, top stories, upcoming events, and more.

Stay up to date with our latest news, receive exclusive deals, and more.

© Analytics India Magazine Pvt Ltd 2022