Anatomy of a medical robot

In one line

A robot is a machine that senses its world, decides what to do, and acts on it — then does it again, many times a second. Strip away the application and a surgical robot, a rehab exoskeleton and a pharmacy-dispensing arm are the same control loop.

The pieces, assembled

• Sense — sensors measure position (joint encoders), force (so a gripper doesn't crush tissue), vision (cameras/endoscopes), and the patient's own signals. An op-amp front-end conditions the weak ones.
• Think — a controller (an MCU for joints, often a full computer for planning) runs the control loop: compare where the joint is to where it should be, compute the correction. The classic algorithm is PID — proportional-integral-derivative — which drives the error to zero smoothly.
• Act — actuators (servos, steppers) move the joints and end-effector (scalpel, gripper, needle).
• Feedback — the result is measured and fed back, closing the loop. Open-loop guesses; closed-loop corrects. Medicine demands closed-loop.

Patterns in medical robotics

• Teleoperation (e.g. laparoscopic surgical robots): the surgeon moves a console; the robot scales and de-shakes the motion. The human is in the loop.
• Shared / assistive control (rehab, exoskeletons): the robot supplements the patient's own effort, read from EMG or force sensors.
• Autonomous tasks (lab automation, dispensing, disinfection robots): the loop runs without a human in it — so sensing and safety limits carry the whole burden.

Why it's safety-critical

The closer a robot gets to a patient, the more the non-negotiables dominate: redundant sensing, force and travel limits, watchdogs that halt on fault, and emergency stops. The interesting engineering isn't making it move — it's making it fail safely. This is also where hardware meets the governance ideas in the Audit & Compliance lab: bounded autonomy, human-in-the-loop, and an auditable trail of every action.

Key takeaways

• Every robot is one loop: sense → think → act → feedback, many times a second.
• The pieces assemble: sensors + op-amp, controller running PID, actuators, closed by feedback.
• Patterns: teleoperation (human in loop), shared/assistive, autonomous.
• The hard part is failing safely — redundancy, limits, watchdogs, e-stops, audit trail.