Medical Robotics. Control Modalities

Transcription

1 Università di Roma La Sapienza Medical Robotics Control Modalities Marilena Vendittelli

2 differ for the the way human input is mapped to the motions of surgical instruments strictly related to the degree of autonomy of surgical robots can be combined in mixed control schemes the choice is affected by the function and, hence, by the domain of use domain of use function control modality (strictly related to the degree of autonomy) kinematic architecture sensors and actuators orthopedics machining of rigid surface cooperative or handson MIS constrained manipulation neurosurgery, intervantional radiology, radiotherapy constrained targeting microsurgery micromanipulation shared/cooperative control M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Control Modalities 2

3 semiautonomous motion the motion of the surgical instrument is interactively specified by the surgeon, usually on the basis of medical images the computer fills in the details and motion starts with surgeon concurrence allows the execution of complex paths starting from simple task description examples: selection of tool cutter paths for orthopaedic bone machining and needle target and insertion points for percutaneous therapy teleoperated systems the surgeon manipulates the input device called master and a device slave, on the patient side, follows the input allow filtering and scaling of position and force, remote tool control other human interfaces are also used (e.g., voice control) examples: common telesurgery systems such as the davinci M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Control Modalities 3

4 cooperative or handson systems both the robot and the surgeon interact directly with a surgical instrument a force sensor detects the direction toward which the surgeon wishes to move the tool and the controller makes the robot move to comply surgeons find this form of control to be very convenient and natural for surgical tasks combines the precision, strength, and tremorfree motion of robotic devices with some of the immediacy of freehand surgical manipulation less expensive than telesurgical systems (less hardware) and easier to introduce into existing surgical settings force scaling is allowed but position scaling is inherently not permitted incompatible with any degree of remoteness between the surgeon and the surgical tool it is not practical to provide handson control of instruments with distal dexterity examples: orthopedic surgery, microsurgery M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Control Modalities 4

5 cooperative systems peculiarities typically nonbackdrivable and velocitycontrolled actuators the velocity of such admittancetype systems is typically controlled with a highbandwidth servo controller to be proportional to the operatorapplied force forcetomotion scaling linear admittancetype device control is generally modeled by V (s) = Y (s)f (s) with F (s): measured command force externally applied by the surgeon; Y (s): device admittance; V (s): device desired velocity nonbackdrivability and natural stiffness of the robot are assumed to be sufficient to reject applied external forces from the operator and the environment not used by the controller the selection of Y determines the responsiveness of the system to human inputs, allowing significant scaling between humanapplied force and robot velocity, which improves accuracy smoothness can be enhanced by thresholding and filtering the measured command force and by reducing unintentional motions and tremor the admittance value is limited by stability considerations admittance control can also be modified to provide haptic feedback and virtual fixtures ( ) M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Control Modalities 5

6 The HandsOn Acrobot (Active Constraint Robot) handson robotic system for bone machining in knee replacement surgery preoperative phase CT scan of the knee an interactive software allows elaboration of the CT images to build 3d views of the knee the surgeon decides size and position of the prosthesis on the basis of the 3D models built from CT images once positioned the prosthesis in the model, the software generates the constraints for the motion of the surgical tool M. Vendittelli Medical Robotics (Universita di Roma La Sapienza ) Acrobot 6

7 System description spherical manipulator with three orthogonal axes of motion: yaw, pitch (within [ 30 o, 30 o ]), and extension (within cm) kinematic design: mechanical impedance of the axes is low and similar for all axes allowing the robot to be moved by the surgeon with low force; force generated during hard bone cutting can be detected without the need for force sensors in the robot cutter the surgeon moves the robot by pushing the handle near the tip of the robot the handle incorporates a sixaxes force sensor, which measures the guiding forces and torques a highspeed orthopaedic cutter motor (removable for autoclave sterilization) is mounted at the tip the Acrobot is placed on a sixaxes gross positioning device to increase its small working envelope M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 7

8 Control idea: gradually increase the robot stiffness when approaching a boundary defined in the intervention planning phase the robot is free to move inside the safe region (RI) due to the low mechanical impedance of the robot at the boundary of the safe region, the robot becomes very stiff, thus preventing further motion over the boundary (RIII) the portion of the guiding force normal to the boundary is directly compensated, which substantially reduces the overboundary error (this error is inevitable due to the limited motor power) to avoid instabilities at the boundary, the stiffness increases gradually over a small region (with width D 1 ) close to the boundary (RII) only the stiffness in a direction toward the boundary is adjusted, to allow motion along or away from the boundary with a very low guiding force M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 8

9 Control cont inner pos/vel control loop, 2 ms (550Hz) th X d, e co inn Ẋ d mpe IK J 1 ( d d KP KD gravity + friction J T C robot J T, joi ) F C external forces (organs, patient) law F G, outer control loop implementing the active constraints 1015 ms (100Hz) M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 9

10 Control cont two control loops: an inner joint position/velocity control loop (2 ms) with friction, gravity and surgeon s guiding force : τ = K P (Θ d Θ) + K D ( Θ d Θ) + τ C + f (Θ, Θ) + g (Θ) τ torque output from the controller to the robot motors, K P, K D proportional/derivative gains, Θ d, Θ d (Θ, Θ) desired (measured) joint position and velocity, τ C of the surgeon s guiding force F G, f (Θ, Θ), g (Θ) respectively friction and gravity the outer control loop is slower (1015 ms) and implements the idea of active constraint cartesian control provides the gains of the inner loop K P and K D, the desired position and velocity (in operative space) X d, Ẋ d ( Θ d, Θ d are obtained by kinematic inversion) and the amount of force F C ( τ C = J T F C, quasistatic operational conditions) the controller parameters are adapted on the basis of the endeffector position X and of the force F G determined the point on the boundary X np closest to X, define N np : normal to the boundary at X np d = X X np = (X X np ) N np M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 10

11 Control cont RI (d > D 1 ) the PD gains are small to allow the surgeon to feel the cutting forces X d = X Ẋ d = AF G admittance is the same in all directions F C = 0 the surgeon s guiding force is not compensated in this region th X d, e co inn Ẋ d mpe admittance A IK J 1 ( d d K P K D gravity + friction robot J T external forces (organs, patient), joi ) F G Ẋ d X RI D 1 RII N np : X np RIII F G, M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 11

12 Control cont RII (D 1 d > 0) the control gains increase as the distance to the boundary d decreases if F G N np 0 X d, Ẋ d and F C as in RI otherwise F G is decomposed into a normal and a tangential component F GN = (F G N np )N np Ẋ d = AF GT + A N F GN, A N = d D 1 A F GT = F G F GN F G is compensated in a region with width D 2 < D 1 if d > D 2 F C = 0, otherwise F C = D 2 d D 2 F GN th X d, e co inn Ẋ d mpe admittance A (AN) IK J 1 ( d d K P K D F C JT gravity + friction C robot J T external forces (organs, patient), joi ) RI RII Ẋ d F G N np : X np RIII RI F G Ẋ d RII N np : X np RIII F G, of F GN M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 12

13 Control cont RIII (d 0) the PD gains are high X d = X np if the direction of F G points away from the boundary, Ẋ d and F C are set as in RI otherwise Ẋ d = AF GT, F C = F GN th X d, e co inn Ẋ d mpe admittance A IK J 1 ( d d K P KD F C JT gravity + friction C robot J T external forces (organs, patient), joi ) RI RII N np : X d = X np if the dire RIII RI RII RIII Ẋ d F G N np : X np F G Ẋ d F G, of F GN M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 13

14 the used prosthesis requires a number of flat planes to be cut; a 2.5D constraint is defined: the plane part allows cutting a flat plane into the bone, the 2D outline part provides protection for the surrounding tissue the two parts of the boundary are orthogonal X d, Ẋ d, F c can be computed separately for each part, following the above rules, and then combined before being passed on to the inner loop K P and K D are determined by the nearest point on the whole boundary 1999 the Acrobot Company was set up after 8 years of research at Imperial College London 2010 the Acrobot Company was acquired by the Stanmore Implants Worldwide Ltd (SIW) 2013 SIW sold the Acrobot system to the American company MAKO M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 14

15 Acrobot today Bibliography M. Jakopec, F. Rodriguez y Baena, S. j. Harris, P. Gomes, J. Cobb, B. L. Davies, The handson orthopaedic robot acrobot : Early clinical trials of total knee replacement surgery, IEEE Transactions on Robotics and Automation, Special Issue on Medical Robotics, vol. 19, no. 5, pp , G.D. Hager, A.M. Okamura, P. Kazanzides, L.L. Whitcomb, G. Fichtinger, R. H. Taylor, Surgical and interventional robotics: Part III, IEEE Robotics and Automation Magazine, vol. 15, no. 4, pp , M. Vendittelli Medical Robotics (Università di Roma La Sapienza ) Acrobot 15