For only $40,500, a fully-sized humanoid with 450 N·m of joint torque, sufficient for the manipulation of human-sized masses, is now available on the market. EngineAI’s T-800, with its military-grade agility combined with service-oriented flexibility, is one of the most technically challenging humanoid robots available on the market, having been unveiled in Shenzhen with plans to be utilized as an additional member of staff in the retail sector as ‘cyber staff.’
With a height of 173 cm and weighing 75 kg including the battery, it has an aviation-grade aluminum alloy frame that combines strength and lightness. With 29 in the body and a further 7 in each of its highly advanced models’ hands, it is possible to achieve all sorts of movements characteristic of the human figure. The Max series, costing $80,800, is further upgraded in terms of computing capabilities, actuators, and sensor systems.
But one area where KUKA’s new robot shines is in its sensor stack and ability to navigate through spaces. The sensor stack is a function of EngineAI and is made up of a 360° LiDAR sensor, stereo vision capabilities, and omnidirectional radars that assist with simultaneous location mapping and avoidance of obstacles. The EngineAI sensor stack provides millisecond-level environment processing that is crucial for a number of dynamic movements of the robot, including flying kicks and changed directions.
The compute options include Intel compute modules up to the NVIDIA Orin class of AI compute, offering up to 275 TOPS AI compute performance. This allows for real-time sensor fusion, motion planning, and adaptive control with still processing left for secondary development tasks.
The firm touts “secondary development” abilities, although information about SDKs/APIs is not available, which may work as a deterrent for robotics developers needing more integration.
T-800’s actuators are designed for high intensity operations, with active cooling in its leg joints to avoid thermal throttling during long-term operations. Its hot-swappability feature in its solid-state lithium power modules delivers 4-5 hours of operation time, and it can be swappedit without shutting it down, essential for long-term service operations. Though its legs can support heavyweight operations, its three-fingered hands have a 5 kg maximum weight it can lift per hand the lowest in the market at 20-25 kg, surpassing those of Optimus from Tesla or Apollo from Apptronik.
From the engineering perspective, the mobility and longevity of the T800 are related to the actuator technology. Though EngineAI has not yet clarified whether it uses rotary or linear actuators in the bearing joints of the robot, based on industry trends, the use of planetary roller screws, as in the Tesla Optimus Gen-3 robot, might significantly enhance the static efficiency of the robot when standing. For example, rotary actuators are often used in dynamic performance robots. They provide the T800 with the agility required for combat demonstrations.
The perception and navigation skills of the robot are also strengthened by work in heterogeneous place recognition, in which common learning frameworks combine data from radar and LiDAR sensors in a common feature space. This enables reliable localization performance even when GPS signals and visual inputs are weak or unavailable and will enable the T800 to function while navigating a cluttered retail setting and a publicly busy setting. It is essential for humanoids that have to coexist with humans and prevent collusion.
Deployment-wise, EngineAI is marketing the T800 less as a proof of robotics ability and more as a practical retail assistant. Its neck flexibility and ability to bend at the waist facilitate eye contact and manipulation of items at different heights. Its ability to move at a speed of 3 m/s coupled with awareness of the environment makes it able to operate in a crowded space while executing tasks.
Whether in a robotic fight tournament or in a retail store as a customer greeter, the T800 embodies the integration of high-torque actuators, multi-modal sensing, and AI computing normally associated with industrial- or laboratory-scale systems. The appearance of such a system in the marketplace at a mid-five-figure price point indicates the beginning of a new phase of humanoid robotics, one in which laboratory systems are replaced with more general-purpose machines.
