S
svish robotics
Back to all posts
Mechatronics & Kinematics • • 3 min read

Why Stiff Position Control Fails on the Assembly Floor

Why compliant joint torque control, PREEMPT_RT deterministic kernels, and real-time digital twins are essential for industrial human-robot collaboration.

#Robotics #Impedance Control #ROS2 #PREEMPT_RT #Digital Twins #Industrial Automation

For forty years, industrial automation treated the factory floor as a geometric grid of high-stiffness position control. Massive cast-iron robotic arms tracked pre-programmed spline trajectories with immense motor torque and zero compliance.

If an operator walked into the path, or if a fixture was misaligned by $0.5$ millimeters, the machine didn't pause: it crushed the workpiece, sheared its gearbox, or caused a severe human casualty. This is why conventional robots must be quarantined behind safety fencing.

At Svish Robotics, we reject the notion that industrial automation requires isolating human workers behind interlocked steel cages.


1. Compliance Through Joint Torque Sensing

The Svish Apex 7 arm integrates custom strain-gauge torque sensors directly at the output of every harmonic drive on all 7 joints.

By measuring the real output torque rather than estimating it from motor stator currents (which are corrupted by harmonic gear friction and temperature hysteresis), our controller achieves:


2. Deterministic Real-Time Architecture with ROS2 & Zenoh

Modern industrial robotics cannot depend on unpredictable user-space operating systems. Svish robots run on a hardened Linux kernel with the PREEMPT_RT patch, maintaining deterministic cycle jitter under 15 microseconds.

Communication between the low-level motor controllers, the on-edge GPU running perception, and our central fleet console uses Zenoh and ROS2 DDS. This yields:


3. The Digital Twin as a Live Operational Mirror

Every Svish robot continuously streams joint telemetry, motor thermal states, and Cartesian tool-center-point poses to app.svish.dev.

Rather than treating the digital twin as a passive 3D CAD visualization, our digital twin actively monitors:

  1. Torque Signatures: Detecting bearing wear or harmonic gear tooth degradation 100 operating hours before physical failure.
  2. Thermal Load Balancing: Dynamically adjusting joint trajectory acceleration profiles during hot shift cycles to prevent thermal shutdown.
  3. Remote Fleet Teleoperation: Enabling supervisors to take control of mobile manipulators over low-latency WebRTC streams when unexpected floor obstructions occur.

Engineering for Human Safety

Physical robotics is fundamentally an engineering discipline of empathy. By replacing rigid, unforgiving machinery with compliant, intelligent, human-collaborative systems, we can eliminate the dangerous conditions that have historically injured factory workers—creating a workplace built around human longevity.