Articles tagged with "Amazon-Robotics"
Amazon Robotics’ ViTa-Zero solves key robotics challenge
Amazon Robotics has developed ViTa-Zero, a novel zero-shot visuotactile pose estimation framework designed to address the challenge of accurately determining object poses during robotic manipulation tasks. Traditional visuotactile approaches have struggled with generalization due to limited visuotactile data. ViTa-Zero innovates by using a visual model backbone combined with feasibility checks and test-time optimization grounded in physical constraints derived from tactile and proprioceptive sensor data. The system models the interaction between the robot gripper and object as a spring–mass system, where tactile sensors provide attractive forces and proprioception generates repulsive forces, enabling more robust pose estimation. The framework was validated on real-world robotic setups across various manipulation scenarios such as grasping, object picking, and bimanual handover. Experimental results demonstrated significant improvements over purely visual models, including a 55% increase in area under the curve (AUC) for ADD-S, a 60% increase in ADD, and an 80% reduction in
roboticsvisuotactile-sensingpose-estimationmanipulation-tasksAmazon-Roboticszero-shot-learningrobotic-perceptionAmazon launches new AI foundation model, deploys 1 millionth robot - The Robot Report
Amazon has reached a significant milestone by deploying its 1 millionth robot across its global fulfillment network, solidifying its position as the world’s largest operator and manufacturer of industrial mobile robots. This achievement builds on a robotics journey that began with the acquisition of Kiva Systems in 2012 and has since evolved to include advanced autonomous mobile robots (AMRs) like Proteus, Hercules, Pegasus, and Titan, capable of handling various inventory weights and tasks with precision navigation and safety around employees. Alongside this milestone, Amazon introduced DeepFleet, a generative AI foundation model designed to optimize the coordination and movement of its robotic fleet. DeepFleet acts like an intelligent traffic management system, improving robot travel times by 10%, reducing congestion, and enabling faster, more cost-effective package deliveries. This AI leverages Amazon’s extensive inventory data and AWS tools to enhance operational efficiency while supporting the company’s processing of billions of orders annually. Despite the increasing automation, Amazon emphasizes its commitment to workforce development, retraining
robotartificial-intelligenceautonomous-mobile-robotsindustrial-automationAmazon-RoboticsAI-foundation-modelwarehouse-automationSafety and efficiency in robotics design
robotroboticsautomationAIindustrial-robotsforce-sensingAmazon-RoboticsSafety and efficiency in robotics design
robotroboticsautomationAIforce-sensingindustrial-robotsAmazon-RoboticsAmazon’s Vulcan robot uses force sensing to stow items
robotautomationforce-sensingAIAmazon-Roboticsmaterial-handlingoperational-efficiency