Embedded Files
Background
Surgical instrument handover between a human surgeon and a robotic assistant is a safety-critical, bidirectional interaction: the robot must both receive instruments from the surgeon and hand them back, adapting its physical compliance to the surgeon's intent rather than following a fixed trajectory. My research addresses this by designing a robotic arm and simulating an adaptive admittance controller that infers surgeon intent (idle, hold, pull, push) from wrist-force and velocity signals and adjusts its virtual mass and damping in real time, so the arm feels compliant when actively exchanging an instrument and stiff/stable when idle.
Methodology
The manipulator was modeled as a 6-DOF robotic arm with a two-finger gripper, using SolidWorks, then imported into Simulink as a rigid multibody chain (world frame → seven revolute joints → two gripper-finger joints) via Simscape Multibody. The controller was built as an "Adaptive_Admittance" subsystem taking wrist force, joint position/velocity, and equilibrium pose as inputs, and routing them through an Intent Estimator.
Results
Over a simulated 63 s sequence of repeated handovers, the intent state cleanly cycled through idle, hold, pull, and push. Virtual damping stepped between discrete levels (roughly 50, 25, and 10 N·s/m) depending on intent, while virtual mass spiked briefly at transitions. Wrist-force magnitude ramped smoothly during each interaction and reset at cycle end, growing to about 6 N by the most vigorous handover. A finer-grained 19 s trace resolved two force peaks (~4.4 N and ~6 N) corresponding to grasp and release, with the eight joint trajectories staying comfortably within their limits and the gripper fingers opening and closing twice, matching the two-way exchange. The Lyapunov function rose during each engagement and decayed back toward zero afterward, staying bounded across all cycles.
Analysis
The adaptive law behaved as intended: stiffer and high-damping when idle or holding (resisting unwanted drift) and compliant with low-damping when actively pushed or pulled (enabling low-effort handover). The bounded, decaying Lyapunov trace across cycles of increasing force amplitude is evidence the controller remains stable even as its own parameters adapt which is an important safety property for physical human-robot interaction. Joint limits being respected throughout, together with correctly timed gripper actuation, confirms the whole-arm motion, not just end-effector compliance, stayed within the manipulator's physical constraints.
Skills developed
Robotic Kinematic and Dynamical Modelling
CAD Design of a 6-DOF Robotic Arm
PID Controller Design
Adaptive Admittance Controller Design
Software used
MATLAB/Simulink
SolidWorks
Python IDE
Page updated
Google Sites
Report abuse