Tag: robotic

Amazon recently unveiled two new robotic and AI technologies designed for its operations network: the "Blue Jay" multi-arm robotic collaboration system and "Project Eluna," an agent-based AI model. Amazon emphasizes that these innovations aim to reduce repetitive tasks for frontline workers, improve workplace safety, increase productivity, and further accelerate package delivery. Tye Brady, CTO of Amazon Robotics, stated, "Our latest innovations are a prime example of how AI and robotics can create better experiences for our employees and customers." Blue Jay: Multi-arm Collaboration, Process Integration, and Reduced Repetitive Labor. "Blue Jay" is described as an "extra helper" for employees, primarily used to assist with repetitive tasks involving reaching and lifting. It's a next-generation robotic system capable of coordinating multiple robotic arms to perform multiple tasks simultaneously, integrating functions that previously required three separate robotic workstations—"pick," "stow," and "consolidate"—into a single, streamlined workspace. Amazon describes Blue Jay's operation as a "perfectly flawless juggler," handling tens of thousands of fast-moving items while also acting as a conductor to ensure all actions are coordinated. This design not only provides more support for employees but also creates greater efficiency in a smaller physical space. Notably, Blue Jay's development cycle from concept to production took just over a year, far faster than the previous development time of over three years. Currently, Blue Jay is being deployed for production testing at Amazon's South Carolina facility and is reportedly able to handle approximately 75% of different types of inventory. It will become one of the core technologies for Amazon's Same-Day delivery network, potentially shortening customer package delivery times. Project Eluna: Agent-based AI, Reducing Cognitive Load and Predicting Operational Bottlenecks On the other hand, "Project Eluna" is an agent-based AI system aimed at helping operations center managers reduce their cognitive load. Traditionally, managers need to monitor dozens of dashboards simultaneously, handle technical glitches, reallocate resources, and make rapid decisions. Project Eluna is designed with a degree of autonomy, capable of reasoning about complex operational situations and providing action suggestions to operators. It integrates historical and real-time data from across the facility to predict potential bottlenecks and help maintain smooth operations. Eluna will be first tested at an Amazon fulfillment center in Tennessee during this year's holiday shopping season, initially focusing on "sorting process optimization"...

Milan Kovac, former vice president of Tesla who led the development of the humanoid robot "Optimus," has announced his departure, citing "being away from home too long" as the reason. Kovac joined Tesla in 2016 as an engineer on the Autopilot team and began leading the "Optimus" team in 2022. In his announcement, Kovac stated that his role will be taken over by Ashok Elluswamy, the current vice president leading the Autopilot team. Kovac emphasized that his departure was primarily due to being away from home too long, wanting to spend more time with his family, and stressed his unwavering support for Elon Musk and the Tesla team, seemingly avoiding any connection between his departure and the recent falling out between Musk and Trump. Tesla first announced its humanoid robot development in 2021 and officially unveiled "Optimus" in 2022. It is expected to sell for $2 upon mass production, and the robot is projected to generate up to $25 trillion in revenue for Tesla. However, Tesla has not yet confirmed a release date for its humanoid robot "Optimus." Previously, Tesla CEO Elon Musk stated that robot production was affected by China's restrictions on rare earth magnet exports, making it impossible to produce robots smoothly in the short term. Over the past 9+ years, I've had the immense privilege to work with some of the most...

Uber Eats Japan announced that starting June 10th, it will begin using delivery robots for food delivery at four Lawson convenience stores in Osaka. This marks the first time Uber Eats has used robots for convenience store delivery in Japan. Uber Eats has been collaborating with Mitsubishi Electric and Cartken in the Japanese market since March of last year, using Cartken's Model C robots for food delivery. In Osaka, robot delivery service began last November, with the robots also provided by Cartken. The newly announced service will utilize delivery robots at four Lawson convenience stores in Osaka: Fukushima 7-chome Store, Fukushima Seiten-dori Store, Kita-ku Sonezaki Shinchi Store, and JR Shin-Fukushima Station Store. However, the robot delivery service with Lawson convenience stores will only operate from 9:00 AM to 9:00 PM; orders placed outside these hours will not be delivered by robots.

In addition to confirming the Galaxy Unpacked event to be held on January 22nd (US Pacific Time), where the flagship Galaxy S25 series smartphones are expected to be unveiled, Samsung reiterated at its CES 2025 pre-show activities its commitment to bringing artificial intelligence applications to various aspects of daily life. This includes launching the Galaxy Book5 series laptops powered by Intel's latest high-performance laptop processors, and announcing that the previously showcased home robot Ballie will officially enter the market in the first half of this year. The new Galaxy Book5 series laptops feature Intel's "Arrow Lake" Core Ultra 200H series processors, and include search functionality similar to Google's Circle to Search, as well as more artificial intelligence applications, such as intelligently zooming in on photos or taking smart screenshots. The updated Galaxy Book5 series laptops all conform to Intel Evo design, and include the Galaxy Book5 Pro and Galaxy Book5 360. Both feature a 14-inch AMOLED touchscreen with a resolution of 2880 x 1800 and a 120Hz refresh rate. A 15.6-inch version with a resolution of 1920 x 1080 and a 60Hz refresh rate is also available. The new Galaxy Book5 series laptops will be launched in the US, Canada, UK, Germany, and France starting in February this year, with specific pricing yet to be announced. As for Ballie, the home robot showcased at CES 2020 and whose design was "flattened" last year, it is confirmed to enter the market in the first half of this year, but similarly, no specific release date or expected price has been announced. Samsung states that Ballie can serve as a personal home assistant, automatically navigating the home environment and interacting with connected home appliances via wired connections. Through continuous interaction, it learns the user's home habits to provide a better interactive experience. For example, Ballie can assist with pet care at home, allowing users to monitor their pets' condition from afar, and also enables interactive features such as video feeds. Ballie can also serve as a home life helper, for instance, by playing music to promote relaxation or answering phone calls.

After shutting down its AI modeling division for robots in 2021, The Information reports that OpenAI seems poised to re-enter the robotics market. Due to significant advancements in humanoid robot technology in recent years, particularly NVIDIA's Project GR00T this year which accelerated humanoid robot development, companies such as Boston Dynamics, Tesla, 1X, Agility Robotics, Apptronik, and Fourier have already invested in the field, and numerous startups in China are also entering the robotics industry. OpenAI previously proposed Roboschool, a toolset for training robots, in 2017, and demonstrated its ability to train robotic arms to solve Rubik's Cubes in 2019. However, it appears to have disbanded its robotics R&D team in 2021, focusing instead on large-scale natural language models and subsequently launching the ChatGPT service. Now, however, OpenAI seems to be restarting its plans to enter the robotics market. While not officially confirmed, OpenAI's recent recruitment of Caitlin Kalinowski, formerly in charge of AR glasses hardware at Meta, to lead its robotics team, coupled with previous reports of OpenAI considering developing its own processors and its existing AI model capabilities, could all be advantages in its entry into the robotics market. However, with NVIDIA accelerating its development of humanoid robots and competition from companies like Boston Dynamics that have been investing in humanoid robots for years, OpenAI may still face considerable competitive pressure.

As expected, shortly after announcing the retirement of its parkour-capable bipedal robot Atlas, Boston Dynamics released a new video teasing the upcoming debut of the new Atlas. The trailer shows a simpler design for the new Atlas, capable of rising directly from a lying position and walking on two legs with smoother movements. Unlike its predecessor, the new Atlas lacks a bulky body resembling a backpack and has exposed wiring; its overall size is also significantly smaller. It's confirmed to operate entirely on battery power, no longer limited to wired connections. While the specific capabilities of the new Atlas are currently unknown, it should theoretically be able to perform parkour and backflips with ease. Given the significant advancements in generative artificial intelligence technology, Boston Dynamics will likely incorporate more AI applications into the new Atlas. According to Boston Dynamics' official blog, they have spent years researching humanoid robots, emphasizing that designing robots with a human-like appearance makes them easier to integrate into human-dominated work environments and can handle various unstructured, unknown, or harsh terrain navigation and movement needs. While Atlas is designed with a human-like appearance, Boston Dynamics states that it aims to make the robot move in the most efficient way, rather than being limited by its human form. Therefore, the newly teased Atlas adopts a non-anthropomorphic standing posture, further explaining that it will promote robot applications through more complete software and hardware integration, along with services and an ecosystem.

Boston Dynamics has announced the retirement of its Atlas bipedal robot, capable of parkour, backflips, and other maneuvers, which inspired numerous subsequent robot designs. Atlas, named after the Titans of Greek mythology who supported the heavens on their shoulders, was capable of running on two legs, walking on uneven surfaces, and performing parkour, backflips, and even maintaining stability under pressure. However, Boston Dynamics did not specify the reasons for Atlas's retirement, but some market observers believe it may indicate that Boston Dynamics is preparing to launch a follow-up bipedal robot design based on Atlas. Boston Dynamics initially spun out of MIT and was acquired by Google in 2013. In 2017, it was acquired by Softbank from Google's parent company, Alphabet, and subsequently by Hyundai Group of South Korea in 2020, which holds approximately 80% of the shares, with the remaining shares indirectly held by Softbank through its subsidiaries.

Following the recent confirmation of the cancellation of its vehicle product development plans, Apple is reportedly focusing on home robot products, including a home-following robot and a desktop application. If you recall, Apple was previously rumored to be creating a device combining a screen and iPad functionality, which was seen as potentially competing with smart screen products like the Nest Hub or Echo Show. However, according to Bloomberg reporter Mark Gurman, this desktop application will also include features such as mechanical adjustment of the screen angle and even simulated head nodding interactions. The other home-following robot will be able to move around the home and even accompany the user, providing assistance whenever needed. Similar products have been proposed by companies including Samsung and LG, and NVIDIA recently launched Project GR00T at GTC 2024, aiming to help developers and manufacturers create more natural-looking interactive robots. It is currently unclear whether Apple will announce any related information at the upcoming WWDC 2024. Robots may be one of the products that allows Apple to achieve new heights in its business after abandoning its vehicle product plans.

Following its launch of robot-assisted food delivery services in Miami and Fairfax, Virginia, Uber has confirmed a partnership with Mitsubishi Electric and Cartken to begin offering Uber Eats food delivery services in Japan this March, utilizing Cartken's Model C robots. Uber stated that the robot delivery service is not intended to replace human delivery, but rather to complement existing delivery methods using bicycles, motorcycles, light trucks, or on foot. The robots used are the Model C robots, developed in collaboration between Mitsubishi Electric and Cartken. These robots move on six wheels and utilize six cameras and advanced AI models to identify road obstacles, enabling autonomous delivery and even remote operation to resolve movement issues when necessary. The Model C robots have been optimized for Japanese road conditions, with a maximum speed of 5.4 km/h and a carrying capacity of 27 liters and a maximum weight of 20 kg. The robot's dimensions are 71cm (L) × 46cm (W) × 60cm (H), and its insulated design keeps delivered food cold or warm. Prior to partnering with Uber, Mitsubishi Electric had already collaborated with Cartken, and since early 2022 had assisted Starbucks, Rakuten Group, and the supermarket chain SEIYU in providing delivery services in specific regions of Japan. In the United States, it also partnered with Grubhub to provide self-service food delivery on university campuses such as Ohio State University and the University of Arizona.

Following its recent announcement of bringing automatically generative artificial intelligence to the Jetson robotic platform, NVIDIA's research team unveiled a solution called Eureka. This solution uses OpenAI's GPT-4 large-scale natural language model to train a robotic arm, enabling it to twirl a pen between its fingers as quickly as a human. Besides pen-twirling, Eureka allows the robotic arm to perform 30 different finger-based actions, such as opening drawers, using scissors, and catching balls. This training will help the robotic arm perform more delicate movements and complete tasks more efficiently, unlike earlier robotic arms that could only perform approximate actions and often took considerable time to complete successfully. NVIDIA researchers point out that training through AI is about 80% faster than human training through procedural writing, and can improve the robotic arm's efficiency by 50%. In the design of the Eureka solution, the AI ​​training method evaluates the results during the process, allowing the robotic arm to complete training more efficiently and perform more accurate actions. The algorithms and libraries related to the Eureka solution will be made available to NVIDIA Isaac Gym, allowing more application services to train robotic arms and perform the required actions more efficiently.