Compact, power-efficient compute and vision stacks for AMRs, cobots and service robots.
On a mobile robot every watt is battery runtime and every cubic centimetre is payload. ARM platforms win here not because they are cheaper, but because 12 W of fanless compute buys hours that a 60 W GPU box takes away.
| Power and thermal budget | Compute competes directly with runtime, and there is no room for a fan intake. Fanless RK3588 platforms delivering 6 TOPS at around 12 W, with conduction cooling into the robot chassis. |
|---|---|
| Sensor fusion timing | Cameras, IMU, LiDAR and wheel odometry must share a coherent timebase or SLAM drifts. Hardware trigger distribution, PTP/IEEE 1588 Ethernet timestamping and a PREEMPT_RT kernel option. |
| Vibration and shock | Continuous vibration destroys connectors and unsupported components. Locking connectors, board-level staking, no socketed memory, and EN 60068-2 shock and vibration qualification. |
| Software stack integration | ROS 2, DDS middleware and NPU inference need to coexist without fighting for cores. Pre-built ROS 2 Humble/Jazzy images with CPU isolation, core pinning and NPU inference exposed as a ROS node. |
8-core flagship with a 6 TOPS NPU, four display outputs and 8K decode — a fanless replacement for entry GPU boxes.
RK3588 compute, synchronised MIPI cameras, tuned ISP and a running inference pipeline — evaluate in days, not months.
Quad Cortex-A55 industrial board with dual Gigabit Ethernet, dual CAN-FD and a 10-year supply commitment.
Customer names withheld under NDA; technical details shared with permission.
A single RK3588 board replaced a GPU module plus a separate motion controller. Removing the fan eliminated the dust-ingress failures the fleet had been seeing, and runtime per charge improved by roughly 18%.
A four-camera kit provides bin-picking guidance with hardware-synchronised capture; pose estimation runs on the NPU while the A76 cores handle ROS 2 and trajectory planning.
Send us the panel size, the interfaces you need and the environment it lives in. You will get a concrete hardware recommendation, indicative pricing and lead time within one business day.
Describe the equipment, the environment and the certification path you need to follow. We will respond with a platform recommendation and the engineering effort required to get there.