Digit 5 is Agility Robotics’ latest humanoid platform, presented as a safety-first machine for high-volume warehouse work. The company says Digit 5 is engineered to lift 23-kilogram bins, reach high shelving, and operate for 90 minutes between charges while recharging in about 9 minutes, with the aim of scaling humanoid labor where jobs are physically demanding or repetitive.
From ATRIAS to Digit
Agility Robotics traces its roots to research at Oregon State University that pursued animal-inspired locomotion through a spring-mass theory of gait. The lab’s early biped, ATRIAS, demonstrated the physics of humanlike walking and running with a minimalist body and compliant legs. That work led to the spinout of Agility Robotics and its first commercial research platform, Cassie.
Cassie introduced dynamic balance to the market, enabling fluid gait through active control. Research teams adopted reinforcement learning to train locomotion in simulation and transfer it to the hardware, helping Cassie recover from stumbles, complete a 5-kilometer campus run, and set a 100-meter dash record for robots.
To extend beyond locomotion, Agility added a torso and arms to Cassie’s legs to create Digit 1, moving toward mobile manipulation.
Why Digit Began With Warehouse Bins
As Agility evaluated first commercial uses, prospective customers pressed for practical value rather than novelty. Bin handling in warehouses and factories emerged as a fit, where two arms, vertical reach, and legged mobility in tight aisles are advantageous.
The task is variable enough to resist single-purpose automation, yet specific enough to be economically viable. Agility began real-world demonstrations in 2019, including a pilot with Ford Motor Company moving packages from an autonomous vehicle to a doorstep.
Subsequent iterations, Digit 2 and 3, supported box handling trials with a global parcel carrier. Amazon evaluated Digit 4 for bin movement and stacking, and Digit became the first humanoid commercially deployed at logistics provider GXO.
In 2024, Digit appeared on the cover of Time magazine.
Early deployments highlighted a gap between research ambitions and customer uptime requirements. Agility dedicated significant resources to keep pilots running, trading margin for learning on issues that extended well beyond core robotics, from service and support to supply chain, compliance, and fleet management.
Turning Digit Into a Product
To commercialize at scale, Agility expanded its leadership team and reworked the platform. What began as a modest update became Digit 4 EVT, a major revision that added larger batteries and broad safety and compliance upgrades on a compressed schedule.
Those additions increased mass and changed the robot’s appearance and performance. While earlier Digits operated without a head, Digit 4 incorporated one to optimize cameras and antennas, a change that also improved human-robot interaction cues.
The redesign process prioritized function over aesthetics, informed by field lessons and a defined roadmap of the first 10 use cases. Those insights set the foundation for Digit 5.
Digit 5 safety architecture and deployment
Digit 5 represents Agility’s transition to a general-purpose system shaped by deployment realities such as CE marking in Europe, UL electrical safety certifications in North America, and FCC emissions limits. Safety emerged as the central barrier to deploying robots near people, since prior iterations often worked behind barriers to satisfy risk and insurance requirements.
Agility designed Digit 5 to operate next to human coworkers under certified safety constraints. The robot uses safety-rated sensing to detect nearby people and follows a conservative response.
As a person approaches, Digit 5 halts its task, sets down any load, lowers itself to the floor, and de-energizes its motors before contact range. The approach is intended to verifiably reduce risk and enable initial side-by-side operations.
The safety architecture influences nearly every subsystem, including protected sensors, isolated safety computers, and body shaping. These design choices add complexity, power demand, mass, and size.
The platform is now over 120 kilograms and targets U.S. National Institute for Occupational Safety and Health lifting guidelines, including repeatedly handling 23-kilogram bins and reaching shelves about 1.75 meters high.
Agility says Digit is built for up to 20 hours of daily work, supported by high-torque, precisely controlled actuators, a large battery, rapid autonomous charging at approximately 240 amperes and 48 volts, and active cooling that occupies most of the torso volume.
Performance trade-offs accompany the gains. Digit 5 does not run or jump like earlier versions, and its larger frame reflects its workload and safety requirements.
Agility also shifted from its earlier “backward” leg configuration to a humanlike leg, addressing squatting limitations and improving heavy lifts from near the floor by coordinating torque contributions from hip and knee drives.
Agility emphasizes that the new form stems from function, not mimicry. According to the company, humanoid proportions align with the demands of repetitive lifting at height, sensing coverage, and verifiable safety in shared spaces.
What Comes After Digit 5
Agility views Digit 5 as its scaling moment, enabling broader deployment while the robot continues with bins and expands to tasks such as item picking, tending machines, preparing parts for assembly and shipping, and eventually assembly itself.
Interchangeable end effectors and modular subsystems are intended to support incremental upgrades, training data collection, and continuous improvement while remaining a sellable product.
Company leaders acknowledge compromises in size, mass, and cost to meet today’s requirements, but contend that these steps are necessary to make humanoids useful at scale.
The long-term vision continues to prioritize agility, dexterity, and natural interaction, with future versions expected to become lighter and more graceful as safety, certification, and workload benchmarks are met.
For more on how AI-focused systems are being developed and standardized for practitioners, see IEEE launches AI Processor Architecture program for engineers.