Teaching factory robots to plan their own collision-free paths - in microseconds, not weeks.
On a modern assembly line, the robot arm is the easy purchase. The expensive, slow, and frustrating part is teaching that arm exactly where to move - and re-teaching it every time the product, the fixture, or the neighboring robot changes. Realtime Robotics, a Boston company founded in 2016, built its business on that overlooked bottleneck.
Its software lets industrial and collaborative robots generate their own collision-free motion plans on the fly. Instead of an engineer hand-choreographing every joint movement, the robot computes a safe path in microseconds and adjusts it as the world around it changes. The company's stated mission is plain: to make industrial automation "simpler, faster, and broader," in service of a blunter vision - "to free the world of dull and dangerous tasks."
The idea has academic roots. Co-founder George Konidaris started a real-time motion-planning project at Duke University and, seeing its commercial potential, teamed with fellow professor Daniel Sorin and researchers Sean Murray and William Floyd-Jones to spin it out. The early breakthrough was a specialized motion-planning processor - custom circuitry designed to solve, in hardware, a problem that had bogged down general-purpose computers.
Sources: rtr.ai/about, Business Wire, The Robot Report, MassRobotics.
Traditional industrial robots are precise but rigid. Each move is programmed in advance, usually by hand, and the robot repeats it on a fixed schedule. That works when the environment never changes - and breaks the moment it does. Realtime Robotics' approach flips the model: the robot treats motion planning as a live computation, deciding how to get from A to B without collisions at the instant it needs to move.
The practical payoff is threefold. Deployment gets faster because engineers no longer hand-teach every waypoint. Multi-robot cells get simpler because the software deconflicts several arms sharing one space, a task humans do slowly and error-prone. And the cell gets safer and more flexible because the robot can react to obstacles - including people - that it wasn't explicitly told about.
Its customers are the companies that live and die by cycle time: automotive and industrial manufacturers, the system integrators who build their lines, and the robot makers whose arms sit at the center of it all. The clearest signal of adoption came in 2023, when the BMW Group named Realtime an official motion-control supplier and began using its RapidPlan software on the production line - initially to choreograph robots performing 3D CT scans that check weld quality.
That is the quiet shape of the company's value. You will not spot Realtime Robotics on a factory tour. You will see robots moving a little faster and a little smarter, with fewer engineers hovering over them.
"Fast, collision-free robot path planning that adapts in real time - no manual reprogramming required."
Generates fast, collision-free robot paths and adapts them in real time, removing much of the manual reprogramming needed to deploy and coordinate robots.
Turn-key 3D spatial perception for the Realtime Controller and RapidPlan - workcell monitoring and active obstacle detection so robots work safely in dynamic, unstructured spaces.
Simulates, tests, and optimizes workcell layouts and task allocation before deployment, maximizing throughput and shrinking cycle time.
The controller that executes real-time motion planning, built around the company's specialized motion-planning processor and custom circuitry.
The default way to program a robot is the teach pendant: a human jogs the arm through each pose and records it. Offline tools like Siemens Process Simulate and Visual Components make that easier, but the plan is still authored ahead of time and re-authored when things change. Realtime's difference is that the plan is computed live.
That distinction matters most in two places. First, multi-robot cells, where the number of possible collisions grows fast and manual deconfliction becomes a project of its own. Second, dynamic environments, where a fixed script simply cannot account for a part that shifts or a person who steps in. Rivals and adjacent players range from the robot OEMs' own stacks (FANUC, ABB, KUKA) to software firms like Ready Robotics and Micropsi Industries and open-source frameworks such as MoveIt.
Realtime Robotics sits in the software-and-controls layer of industrial automation - above the robot arm, below the factory's planning systems. Its business is B2B: it sells and licenses RapidPlan, RapidSense, and Resolver to manufacturers, integrators, and robot OEMs, and partners with automation suppliers to embed the technology into workcells.
The company's backers double as a market map. Strategic investors and partners include Mitsubishi Electric, HAHN Automation, Kawasaki Robotics, and Siemens - the kind of names that buy a technology because it lowers a bill they actually pay. When customers become investors, the signal is worth reading.
Series B amount not publicly disclosed. Figures approximate, compiled from public reporting (Business Wire, The Robot Report, Crunchbase).
Spun out of DARPA-funded motion-planning research at Duke University.
Develops its specialized real-time motion-planning controller and processor.
Raises $11.7M to commercialize the real-time motion-planning platform.
Launches its motion-planning and perception products alongside a Series A extension raising $30M+.
Named an official motion-control supplier; RapidPlan goes to the production line.
Mitsubishi Electric leads a strategic Series B and joins the board.
Led four prior companies - two to IPO, one to strategic sale. Holds an MS from MIT.
Started the Duke motion-planning project that became the company's foundation.
Duke professor who co-established the company around the processor breakthrough.
Researchers Sean Murray and William Floyd-Jones, whose work seeded the technology.
It builds software and hardware that let industrial and collaborative robots plan collision-free motion in real time, automating much of the programming needed to deploy and coordinate robots.
RapidPlan (real-time motion planning), RapidSense (3D perception for obstacle avoidance), Resolver (workcell simulation and optimization), and the Realtime Controller.
Automotive and industrial manufacturers, system integrators, and robot OEMs - most notably the BMW Group, which named it an official motion-control supplier.
Founded in 2016 out of Duke University research; co-founders include George Konidaris, Daniel Sorin, Sean Murray, and William Floyd-Jones, with Peter Howard as co-founder and CEO.
Roughly $100M in total, including an $11.7M Series A, a Series A extension of over $30M, and a strategic Series B led by Mitsubishi Electric in 2024.