Brian Cossette’s Post

Take a look and spend a thought on mass market maturity of humanoid consumer robots by 2030 of US, Chinese, Japanese and Korean vendors for a strong growing market from 2028 to 2050. Than think about an outclassed Germany and Europe in this upcoming consumer mass market with first announcement for 2025: https://lnkd.in/e3t5kT7y What a galloping loss of the future for german and european industries, specially for automotive industries with their OEMs. But Europe is leading in robotic ethics and regulation - what a mess. https://lnkd.in/eCwWhJeJ Humanoid industrial robots, are currently usable and productive and are mainly imported from the USA, China, Japan and Korea at shrinking costs of probably 80.000 € for purchase and 4.000 € yearly for maintenance today too 40.000 € in 2030 realized by series production (Estimation by Horvárth 2024). https://lnkd.in/eXmq95_c

Smart Cleaning Revolution: The AI-Powered #SP50 Vacuum Robot Transforming Senior Living Facilities Discover the future of cleanliness in senior living facilities with our groundbreaking AI-powered vacuum sweeping robot. Equipped with advanced 3D LiDAR technology, this smart robot navigates with unparalleled precision, mapping out its surroundings to ensure no corner is left untouched. Its intelligent AI algorithm devises the most efficient cleaning routes, providing smooth and thorough cleaning without human intervention. The robot autonomously returns to its charging station when needed, ensuring it’s always ready for the next task. With flexible scheduling capabilities, it operates round the clock, adapting to the living facility’s rhythms and needs, promising a consistently pristine environment for our seniors. Embrace the smart cleaning revolution where innovation meets care, ensuring our loved ones live in the cleanest, safest environments possible. #AIInnovation #SmartCleaning #RoboticsInCleaning #EducationalFacilities #FacilityMaintenance #CleanTech #FutureOfCleaning #cenoBots

🌟 ROS-In Series: 2D Mapping with SLAM 🌟 Welcome to the second post in my ROS-In series, where I’ll be sharing my learnings of ROS. Today I've taken the next step into 2D mapping using SLAM (Simultaneous Localization and Mapping). 🚀 What is SLAM? SLAM or Simultaneous Localization and Mapping, is a fundamental technology in robotics that allows a robot to create a map of its environment while simultaneously determining its position within it. By integrating data from sensors like LIDAR, cameras, or depth sensors with motion data (odometry), SLAM algorithms enable robots to explore and navigate unknown spaces autonomously. This process involves identifying landmarks, estimating the robot’s position, and building or updating the map in real time. SLAM is essential for tasks like autonomous navigation, as it eliminates the need for pre-existing maps or external positioning systems, making robots adaptable to dynamic and unstructured environments The result? A detailed 2D map of the environment that Mack navigated. 🗺️ This map will serve as the foundation for autonomous navigation in the next steps of my ROS-In series. It’s fascinating to see how SLAM combines sensor data, odometry, and advanced algorithms to build accurate maps. This milestone feels like a big leap toward making Mack fully autonomous! 🚗💡 Stay tuned for the next post, where I’ll dive into path planning and autonomous navigation. Let’s keep building! 💻 . . . #ROS2 #SLAM #Cartographer #Robotics #AutonomousNavigation #LearningByDoing #Mack #Learning

The field of humanoid robots is exploding. The differentiator is the AI. None of the humanoids are truly intelligent today. Superintelligence Computing Systems is working on sdroud brain with AGI qualities.

𝐇𝐨𝐰 𝐝𝐨𝐞𝐬 𝐚 𝐫𝐨𝐛𝐨𝐭 𝐥𝐢𝐤𝐞 𝐭𝐡𝐞 𝐏𝐔𝐃𝐔 𝐌𝐓1 𝐤𝐧𝐨𝐰 𝐞𝐱𝐚𝐜𝐭𝐥𝐲 𝐰𝐡𝐞𝐫𝐞 𝐢𝐭’𝐬 𝐠𝐨𝐢𝐧𝐠 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐛𝐮𝐦𝐩𝐢𝐧𝐠 𝐢𝐧𝐭𝐨 𝐚𝐧𝐲𝐭𝐡𝐢𝐧𝐠? Introducing the PUDU #MT1 by Pudu Robotics , an AI-powered robotic sweeper that’s not just sweeping up dirt—it’s sweeping away outdated ways of managing large-scale facilities. This machine is designed to take care of huge spaces, offering the kind of efficiency that manual cleaning simply can't match. 𝐓𝐡𝐞 𝐬𝐭𝐚𝐫 𝐨𝐟 𝐭𝐡𝐞 𝐬𝐡𝐨𝐰? #LidarTechnology. Picture Lidar as the robot’s superpower, allowing it to "see" and map its surroundings in real-time with pinpoint accuracy. Whether it's navigating through narrow aisles or dodging obstacles, this isn’t just another cleaning robot—it’s a spatial genius. By combining Lidar with VSLAM (Visual Simultaneous Localization and Mapping), the MT1 builds a detailed map of its environment, constantly adapting its path, almost like a GPS, camera, and AI brain all working together to ensure no spot gets missed. At QWR (Question What's Real), we’re building a compact tracking device that can be mounted onto any robot—or even everyday objects—instantly making them spatially aware. We’re not reinventing the wheel here; this is a tech that robots will increasingly rely on. Felix Zhang | Raymond Pan | Adam Reiver |

A recent paper I co-authored with Paola Iacomussi Brian Deegan Alexander Braun Paul Carlson and Robert Dingess entitled "Is It Time To Consider A Non-Biological Reference Observer?" examines a simple but evidently non-obvious idea - given that all road-marking systems so far are based on human visual perception, what are the new opportunities for computer vision #AI based driving systems in the context of modern image sensing systems and the possibilities afforded by reconsidering the physical properties of road-marking materials? The question led me to naming the corresponding nascent research programme BEtter Live-LAb with ROad MArkings (BELLAROMA), which will inspire many questions and avenues to be substantiated with experiment, and will help to exploit the confluence of road-marking properties and innovations in sensing technologies to make safer transportation networks. Our paper provokes such research questions. The authors constitute an experienced collective of experts who have developed automotive imaging systems on millions of cars, road marking systems on millions of kms of roads as well as the associated standards behind them, and I'm looking forward to developing this provocative question further in the innovative environment of the University of Limerick , as part of our wider and deeper investigations of #ai #imaging and #computervision in the automotive environment. As the abstract puts it - Advanced Driver Assistance Systems (ADAS) play a relevant role in compensating for human physical limitations and increasing road safety. Cameras and sensors (e.g., lidar) are crucial for a vehicle to sense and perceive road surroundings and act to increase driving safety. Thus, driving is no longer a human-only visual task. Camera systems have been developed over many decades with the human visual system as a reference, both as a technical basis (e.g., the choice of colour filter arrays) as well as the final application, providing an image to the driver. Increasingly, this reliance on the human visual system limits the development and performance of ADAS functionality, as the consumption of the images by a computer vision algorithm has distinctly different requirements than the human visual system. In this article, we detail these differing requirements with examples from automotive applications, to support the need for a new non-biological reference observer like the CIE photometric reference observer for colours. Margaret Belska Robin Jenkin Robert Stead Dr. Sven Fleck Trent Victor Alexis Lluis Louis Kerofsky Ciarán Eising Quentin Noir Uwe Artmann Norman Koren Jonathan B. Phillips Bernice Rogowitz Deardriu Lally, PMP, Prince2 Stephen Kinsella Colin Fitzpatrick Patrick Healy Albert Theuwissen Laurent Chanas Darryl Perks Martina Prendergast Padraig Costello BFS APA LIB Kenneth Stanton Senthil Yogamani Pepijn van de Ven Nikola S. Nikolov Lionel Briand Michael Forde Valentina Donzella Dylan Lane Martin Glavin Olivia Wynne Jonathan Horgan

How does a camera perceive the 'yellow' of yellow road markings? Does it even make sense to define road-markings for Computer Vision as colors? What about emerging infrared capabilities? BELLAROMA is here to tackle these questions!

Dr. John Vial's "Become a Roboticist" course - Day 3: Adding sensors The Krytn robot's description (.urdf) file came with a 2D LIDAR sensor joint (rplidar on the left, little black knob on the right) and its scanning capability already in place, both of which can be visualised in Rviz (left side of video) simultaneously with the Krytn robot being teleoperated in the Game City Cafe world simulation in Gazebo (right side). Changing just a couple of little parameters has a noticeable (detrimental) effect on Krytn's ability to gauge its physical surroundings appropriately. Run #1 (default parameters): the robot can effectively detect obstacles or lack thereof at an appropriate height - notice the continuous obstacles to the right, the gaps between obstacles to the left, and the gauging of distance from the pinball machine's feet. Run #2 (changed paramaters - LIDAR sensor raised by 30cm, sampling rate doubled): with the sensor now at a height not appropriate to the rest of the robot's frame, many obstacles are not detected - the seat to the right is invisible as the LIDAR beam hits the wall (infinity symbol under Chun Li) above it, all the seats to the left are invisible as the beam hits the window behind them instead, and the jutting-out front panels of the pinball machine are registered as an obstacle even though realistically the robot can fit underneath them. The increased sampling rate doesn't help at all, especially given the static environment. Imagine having to teleoperate the robot safely through the cafe - or worse, the robot navigating autonomously - with only this LIDAR data to go by, no visuals. I know which version of the parameters I would choose! Yet another example of why it's important to match software to hardware - if you want your simulation to reliably predict the robot's behaviour in the real world, all the details need to match the real robot! Next up: creating a map from sensor data and enabling autonomous navigation! (hmmm.... better bring that LIDAR sensor back down first😅 )

Its exciting to watch Terra AI from Stereolabs take shape on their website! #SpatialAI is finally mature enough for widespread adoption by autonomous robots and they're leading the charge powered by NVIDIA Robotics Jetson boards. As expert roboticists, we recognize the unique challenges inherent with deploying and managing modern AI-based perception systems (doing far more than just 2D detection and segmentation), so having partners like Stereolabs is critical to accelerating your #ROS, #Nav2, or other product to market. I can't wait for a future where Nav2 and autonomy frameworks are managing the standardized outputs of dense and semantic environmental models from #TerraAI and others for our planning, control, behaviors, or even higher level #AI application systems! There will be so much more we can do to make robots safer, more reliable, and more intelligent. Definitely check out the great videos they have there!

Meet the new humanoid robot: the Unitree G1! The Unitree G1 is a next-generation humanoid robot designed to be more accessible, starting at just $16,000. This compact marvel stands at 4 feet 2 inches and weighs 77 pounds, boasting superior flexibility and agility with its 23-43 joints—more than a human. Its advanced hands, powered by AI-based imitation and reinforcement learning, enable the G1 to manipulate objects with precision and continuously improve its skills. Enhanced with a 3D LiDAR and depth camera, the G1 offers improved environmental perception over previous models. Robots are here, and they’re more capable than ever! Follow me MD ROBIUL HOSSEIN for AI and Robotics updates. #Robotics #AI #Innovation #Technology #UnitreeG1