Contact Us
Home Resources Architecture Robot Dog

Robot Dog
System Architecture

An integrated reference for multi-sensor perception, AI planning and deterministic quadruped motion control.

Explore the Architecture
Concept illustration of a white and graphite four-legged robot with camera sensors and articulated joints

System Architecture Views

Two views of perception, real-time motion and independent protection across the quadruped platform.

01

System Architecture

Sensor observations, estimated state and bounded motion goals meet at a deterministic controller; protection remains independent.

Scroll horizontally to inspect the full signal map →

Robot dog system architecture Vision and ranging data pass through capture and timestamping to AI compute. A clock-domain correlation maps sensor timestamps to the real-time host clock, so camera observations can be aligned with IMU and odometry state for localization. AI sends bounded motion goals to a real-time controller. Body IMU feeds state estimation in the real-time domain. Joint feedback returns from the leg drives over the cyclic bus to the real-time host, then enters state estimation internally. Real-time status is monitored by an independent safety controller, which supervises hardwired safety I/O to the drives. Power remains separate. CLOCK-DOMAIN CORRELATION Sensor time ↔ RT time SENSING Vision & Ranging Cameras · Depth · LiDAR Environment observations Configuration dependent CAPTURE & TIME Capture & Time Native I/O · acquisition Timestamp / transport Bus selected per build AI COMPUTE AI Compute Perception · Localization Navigation / planning Bounded motion goals INDEPENDENT PROTECTION Safety Controller Hardwired E-stop / limits RT status monitoring Project-defined response PROPRIOCEPTIVE INPUT Body IMU Attitude / angular rate Control-domain acquisition RT-HOST STATE ESTIMATION State Estimation IMU · contact · joint state Pose / velocity estimate MOTION CONTROL RT Motion Control Gait · balance · trajectory Cyclic joint commands HARDWIRED / SAFETY I/O Drive Safety I/O Controlled sit-down / damping Actuator-dependent response POWER DOMAIN Power Distribution Protected rails / monitoring Drive feed illustrated DISTRIBUTED ACTUATION Leg Controllers · Drives · Actuators Front Left Front Right Rear Left Rear Right 2. TALO-B1000 5. AEB-IU300 1. TALO-A1000 4. TALO-M1000
01

Perception path

Observe, timestamp and plan.

Clock correlation aligns sensor timestamps with RT state; AI sends bounded goals to motion control.

02

Control path

Estimate state and execute the joint loop.

Joint feedback returns from the drives over the cyclic bus to the RT host, then enters state estimation.

03

Independent safety path

Separate inputs, decision and drive response.

Select a block to explore its role.

Recommended Albatron platforms

1 Jetson Thor AI Controller TALO-A1000

AI perception and planning.

View product
2 FPGA Sensor Bridge TALO-B1000

Synchronized sensor acquisition.

View product
4 EtherCAT IMU / VRU TALO-M1000

EtherCAT inertial-state feedback.

View product
5 Motion Control Host AEB-IU300 Series

Motion-control host; requires a real-time OS, control stack and cycle validation.

View product

Final selection depends on the robot’s power, thermal and interface requirements. TALO-M1000 belongs to the EtherCAT control path; drive control and independent safety hardware remain project-specific.

Reference design, not a certified safety function or a fixed implementation. Interface protocols, clock correlation, drive commands and the safe-state strategy must be chosen and verified for the robot design.

Build for the physical world

Engineer the control foundation for reliable quadruped autonomy.

Talk to an Expert