Robotic platforms capable of self-directed locomotion through environments, navigating without fixed infrastructure guidance by using onboard sensing, mapping, localisation, and path-planning capabilities. Autonomous mobile robots (AMRs) operate in dynamic, human-shared spaces and are distinguished from automated guided vehicles (AGVs) by their ability to adapt routes in real time rather than following pre-defined paths.
Content
- The distinction between AGVs (which follow fixed infrastructure) and AMRs emerged practically in the 2010s, driven by advances in affordable solid-state LiDAR sensors, increased compute density, and the maturation of probabilistic robotics methods. Companies including Kiva Systems (acquired by Amazon in 2012 and rebranded Amazon Robotics) pioneered large-scale warehouse robotics, though the original Kiva platform operated on a constrained grid model rather than free-form navigation. MiR (Mobile Industrial Robots, founded 2013) and Fetch Robotics popularised truly autonomous, infrastructure-free industrial AMRs.
- The technical architecture of an AMR centres on the navigation stack: a sensor perception layer producing local obstacle maps, a global planner computing a path through the pre-built map to the goal, a local planner implementing dynamic obstacle avoidance (DWA, TEB, or similar), and a velocity controller translating planned motion to wheel commands. The Robot Operating System (ROS) and its successor ROS 2 provide the standard middleware framework binding these components. LiDAR-based SLAM (Google Cartographer, SLAM Toolbox, ORB-SLAM3 for vision-based variants) produces 2D occupancy or 3D point-cloud maps. Multi-floor operation and semantic navigation (understanding room labels, human gesture instructions) extend the basic navigation capability.
- AMR deployment spans logistics and fulfilment (Amazon, Ocado, AutoStore), manufacturing intralogistics (delivery of parts to assembly stations), healthcare (medication and supply transport in hospitals), and retail (autonomous shelf-scanning and restocking robots). Fleet management software orchestrates task assignment across tens to hundreds of AMRs in a facility, using queueing theory and multi-agent path finding (MAPF) algorithms to maximise throughput and minimise congestion. Safety standards (ISO 3691-4 for industrial trucks, ISO 13482 for service robots, IEC 61508 for functional safety) govern the design and certification of collision avoidance and emergency-stop systems.
- In 2024–2025, AMR capabilities are extending towards mobile manipulation — combining navigation with robotic arms for pick-and-place tasks — as seen in platforms from 6 River Systems, Locus Robotics, and Boston Dynamics Spot with manipulation attachments. Foundation models are enabling AMRs to interpret natural-language task instructions and reason about novel objects without explicit programming. Human-robot interaction improvements, including social-norm-compliant navigation and verbal status communication, are addressing acceptance barriers in shared workspaces. The global AMR market is forecast to exceed USD 15 billion by 2027, driven by e-commerce growth and labour market pressures in logistics and manufacturing.