1Force Tracking with Feed-Forward MotionEstimation for Beating Heart SurgeryShelten G. Yuen, Douglas P. Perrin, Nikolay V. Vasilyev,Pedro J. del Nido, and Robert D. Howe, Senior Member, IEEEAbstract—The manipulation of fast moving, delicate tissues inbeating heart procedures presents a considerable challenge to thesurgeon. A robotic force tracking system can assist the surgeonby applying precise contact forces to the beating heart duringsurgical manipulation. Standard force control approaches cannotsafely attain the required bandwidth for this application due tovibratory modes within the robot structure. These vibrations area limitation even for single degree of freedom systems drivinglong surgical instruments. These bandwidth limitations can beovercome by incorporating feed-forward motion terms in thecontrol law. For intracardiac procedures, the required motionestimates can be derived from 3D ultrasound imaging. Dynamicanalysis shows that a force controller with feed-forward motionterms can provide safe and accurate force tracking for contactwith structures within the beating heart. In vivo validationconfirms that this approach confers a 50% reduction in forcefluctuations when compared to a standard force controller anda 75% reduction in fluctuations when compared to manualattempts to maintain the same force.Index Terms—force tracking, beating heart surgery, motioncompensation, 3D ultrasound, medical roboticsI. INTRODUCTIONIn beating heart procedures, the surgeon operates on theheart while it continues to pump. These procedures elim-inate the need for cardiopulmonary bypass and its associ-ated morbidities [1], and allow the surgeon to evaluate theprocedure under physiologic loading conditions. The latter isparticularly useful in the repair of cardiac structures like themitral valve that undergo substantial mechanical loads duringthe heart cycle [2]. However, surgical manipulation of thebeating heart is challenging because heart motion exceedsthe human tracking bandwidth of approximately one Hz [3].The mitral valve annulus, for instance, traverses most of its10–20 mm trajectory and undergoes three direction changesin approximately 100 ms [4]. This makes the application ofprecise forces for surgical tasks like mitral valve annuloplastydifficult. Indeed, recent animal trials indicate that beating heartrepair of the mitral valve cannot be performed reliably due toits fast motion [5].A force controlled robotic surgical system could benefit thesurgeon by applying precise forces to the heart as it moves.Previous work on surgical force control has largely focused onThis work is supported by the US National Institutes of Health undergrant NIH R01 HL073647-06. S. G. Yuen and R. D. Howe are with theHarvard School of Engineering and Applied Sciences, Cambridge, MA 02138USA (e-mail: [email protected]). Howe is also with the Harvard-MITDivision of Health Sciences & Technology, Cambridge, MA 02139 USA.N. V. Vasilyev, D. P. Perrin and P. J. del Nido are with the Department ofCardiovascular Surgery, Children’s Hospital Boston, MA 02115 USA.DelayedTissuePositionMotionCompensationInstrument3D UltrasoundMitral ValveAnnulusPredictive FilterFeed-ForwardForce ControllerTissue Velocity and Acceleration EstimatesReal-TimeTissue TrackerInstrumentPositionForce SensorContactForceDelayedVolumetricImagesUltrasoundProbeFig. 1. The surgical system actuates an instrument to apply precise forcesagainst beating heart structures. The controller uses both force measurementsand feed-forward tissue motion estimates that are derived from a 3D ultra-sound tissue tracker and predictive filter.force feedback for teleoperation of surgical instruments androbots (reviewed in [6]). Force feedback has demonstrated anumber of performance benefits in the execution of remotesurgical tasks [7], [8] and can enhance safety when usedto implement virtual workspace limits [9]. In this setting,the primary role of the force controller is to provide hapticinformation to the user while the user commands the robot tointeract with the surgical target.In contrast, beating heart applications require the robotcontroller to autonomously maintain prescribed forces of theinstrument against the target tissue despite its fast motion. Onemajor concern is safety, given the well-documented occurrenceof instability in force control [10], [11], [12], [13]. A roboticsystem for beating heart surgery must be damped and stableto ensure that it will not overshoot or oscillate in response tochanges in the desired force trajectory or sudden target mo-tions. Furthermore, the system must have sufficient bandwidthto reject the disturbance caused by heart motion. Previousresearch indicates that standard force control strategies canonly achieve stability for low closed-loop bandwidths dueto vibratory modes in the robot structure [11], [12], [13].These findings were obtained in the context of large industrialrobots interacting with stiff targets. To ensure adequate robotperformance and safety, it is essential to determine whetherthe same limitations exist in beating heart surgery where thetarget is soft but rapidly moving.In this work, we study force control in the context of beatingheart surgery and find that the standard force controller doesindeed suffer from bandwidth restrictions due to the vibratorymodes present in long surgical instruments. However, byincorporating feed-forward tissue motion information into the2Linear MotorForceSensorFig. 2. The motion compensation instrument (MCI) is a handheld surgicalanchor deployment device. It is actuated in one degree of freedom to cancel thedominant 1D motion component of the mitral valve annulus. A tip-mountedoptical force sensor [15] measures contact forces against beating heart tissue.Motion compensation instrument(Rigid model)Beating heart targetFig. 3. Rigid body robot model in contact with a moving, compliantenvironment.controller, safe and accurate force tracking can be achievedat low bandwidth. In preliminary work we experimentallydemonstrated the efficacy of the approach [14], and in thefollowing we provide a detailed analysis of the feed-forwardforce controller, as well as in vitro and in vivo validation. In thefirst part of this paper we show that simultaneously achievingan adequately damped system with good disturbance rejectionis challenging because it requires a closed-loop bandwidththat would excite undesired vibratory modes in the robot.Subsequently, we describe a force tracking system