Force Feedback: Intuitive da Vinci 5 vs Johnson & Johnson Ottava
- Steve Bell
- 17 hours ago
- 18 min read
Brief patent trawl over the last days to try and work out how different systems might be doing their force feedback and how the team at Ottava (if it ever releases force feedback) might be approaching this.
Let me start by saying clearly - I can only derive total speculation from looking at patent filings in the public domain and looking at any data in public press releases. And this is purely my interpretation to try and help explain what might be going on.
The Problem
I’ve intentionally used force feedback here and not the wider term of haptics. I have written on this before and go back through my blog and find the post on this.
What people think surgeons “want” is to regain some form of knowing what forces are being applied to tissue to avoid pulling / pushing / squeezing too much. When we moved from laparoscopic manual instruments to robotic instruments we lost that ability to feel force on the tissues.
(Nearly 20 million procedures so far have been done without it. juts a sayin’) Experienced surgeons adapted using muscle memory and visual clues to “guesstimate” force being applied. This is often called visual haptics. Some are good at it and some are okay , whilst many are bad. Data shows that most surgeons underestimate forces being applied when analysed.
I have written before that the da Vinci is NOT the first system with force feedback. In fact there have been many before. But it is the first to do this at scale. It gives a feedback through the haptic arms (hand controllers) as well as giving “strain gauges” for how much force is being applied at the instruments tip. A number readout.
The theory is that if we give surgeons back the “force feedback” surgeons will apply less force - novices will learn robotics faster and all of this should translate into better clinical outcomes (in theory and yet to be conclusively proven.)
Whatever - the key is that the most modern systems and the systems that are the newest with significant volumes and deployments (DV5) have force feedback - love it or hate it. It’s here.
The problem for everyone else is that they now need force feedback. Asensus was one of the originals, Toumai has it, Edge has it; and several other system are developing systems with it.
Of note: AI needs it to train better for future autonomy. Using visual clues alone is possible - but once you layer in that force feedback - the AI can do much more surgeon like job of controlling instruments..
Look it’s here and it’s coming alright; and if you want to be in tenders in the future… it’s one of the tick boxes that will need to be checked. Leave it off your system at your peril.
How is it measured?
Okay it is super technical - and I want to make this accessible to all, and talk generically and not get into the exact sensing / mathematical side of things. (I’m sure I will.)
But let me explain a few things that need to be overcome; as that will determine how you get force feedback - if it’s any good ,and can you have patents around it to protect it. And like wise - can you apply force feedback without crashing into the vast IP estates that are out there already.
Now how you do force feedback is important because, as has been said to me, “bad force feedback is worse than no force feedback.”
And one of the issues we have in laparoscopic robots is that we have a trocar in the patient that the instrument passes through. And we want to measure forces beyond the trocar with a robot sitting on this side of the trocar. And that trocar - and the patient breathing (their abdominal wall moving) add their own forces that are different to those when you are grabbing tissue. Those “extra” forces create noise in the signal. And if you don’t “deal” with them, then you can end up with giving the user “false force feedback.”Oh just to add to the complexity - all of those forces are dynamic and changing in microseconds.
So the trocar is one issue that must somehow be “removed” or “accounted for”. The trocar adds lots of noise - because in the trocar is often a silicon valve that does all kinds of stuff to the shaft of the instrument as it moves in and out. Vibration - stick slip - dampening - etc etc. It just dirties everything in the force measurements.Plus the abdominal wall moves as the instrument moves - the patient breaths - the gas in the abdomen changes - just a ton of extra noise that gets transmitted through the trocar - adds to the noise - and well - very noisy signal.
All of that “noise” clearly has an impact on what signal is transmitted back to the robot and hence the robot can transmit to the surgeon with fidelity. But that “friction” also has some impact on the forces applied to the tissue by the robot - it’s a feedback loop. The signal beyond the trocar is usually much “cleaner” than the signal on the user side of the trocar.
So that means:
You have two or three choices for how and where you measure the forces on the tissue. Some of that is practicality - but as we’ll see - some of that is IP landscape.
The ideal seems to be to place “sensors” down where the forces are really happening - at the tip of the instrument. So you can measure forces directly as close to the tissue as possible. That would require not only small sensors embedded in the tips of the instruments - but also ways to transmit that data back to the robot. Add to that you may also still need to measure gross forces on the shaft - via the robot arm - and thew closure forces of the jaws of the instrument b y the drive packs. it is complex. And maybe why it has taken Intuitive nearly 20 years to release its first force feedback system.Now, that “direct” measuring is one way to be able to ignore all that noise coming from trocars, valves and abdominal movements.
So come further back and you can say that if the instrument shaft is a snug fit in the trocar - and the trocar is rigid enough - you could theoretically measure forces directly at the trocar. It adds some abdominal noise but is much closer to the tip - has the direct ability to compare and contrast the trocar forces to the instrument forces (noise) and do some mathematical deductions to get an answer.
But you might not have any direct “griping force” sensors - so may also need to do a bit of maths for how much strain is placed on the cables closing the jaws - or how much current gets drawn from the motors to close the jaws - and do some very clever algorithmic maths.
It’s good but is all starts to add more “calculation errors” into the system. Each error adds up - and we are talking about tiny tiny differences in force.
Or you could forget all of that and try and do it all at the instrument interface and the robotic arm outside of the patient. Accept there’s noise from trocars etc and abdomen movement - valves etc. But do say a simple subtraction.
Wave and move the instruments free in the air (free space) and get a baseline of forces needed by the motors of the robot arm. Then place the instrument through a trocar without gripping tissue - get a trocar baseline. And, say, a trocar in an abdomen that moves - and get an abdominal baseline. - And then with all those base numbers you start doing some maths. When you finally get all that baseline data - go grab tissue. Take the new value away from the total baseline - and voila you have a mathematical force feedback.
And if you do some magic by measuring wire forces - motor current draws etc - you could get a decent value. Not the exact measure but a decent value. It requires a lot of fast compute ion the fly - but it is possible. (But is it good enough?)
And for a bit later - not having to drop all those sensors in the instruments means less expensive instruments - longer lives - and every instrument become force feedback day 1 - you don’t need force feedback and non force feedback instruments.
So why am I explaining all this?
I’m getting a LOT of questions about the validity of force feedback: “Do I think it’s hype or real?”
I won’t go into this now as I’ve written extensively that if the clinical data supports it, and you get better outcomes… it will become almost mandatory. And for new start surgeons that don’t have a heap of lap experience to fall back on - “Why not?”
Why not give them as much support as possible to build up their visual haptics. If after that - they don’t want it - they can turn it off.
Although I firmly believe the next generation of surgeons that grow up on force feedback will miss it if its gone. That’s what I predict. The other main thing I get asked is - "Will Ottava have force feedback?" - well it will need it in the future if it wants to compete. So I will speculate just for you!
So based on that I wanted to explain where I think the two companies in terms of the title of this blog are on this. And I’ll throw in a third company, Microport Medbot, into the mix. Now again I can only go from what I can dig up from public documents, and I am sure every company has some secret sauce that I can’t understand. But this is how I see how each company is approaching this from the public domain documents.
INTUITIVE - Da Vinci - How do I think they do it
By studying the IP and digging into some research papers l see what could be the way they do some of this. Not all as I am sure much of it is trade secret.
It seems they have some sensors / strain gauges down at the tip of the instrument. When forces are applied those strain gauges flex (one patent describes it as a small ribbed tube) but hard to know if that is what made it into the device. Anyhow seems in one layer they have a sensor at the distal end of the instrument - right next to the tissue where the force needs to be measured. Data from those sensors runs up the instrument and then that data passes across to the robot. Because it is the robot and the 10,000X compute that needs to interpret those signals and transform them into sensible data to back drive the controllers in the surgeon’s hands.Here’s where I’m speculating. Remember there’s a sterile drape cap between the instrument drive pack and the motor drive pack of the robot. So, somehow, data needs to be wirelessly transmitted across the drape cap (I think). Some kind of electrical or optical contactless signal bridge could be used. I don’t think this is via their normal NFC chip for talking lives and instrument type. I think this is some form of data stream they can pick up from the instrument.
I think that is how they do the lateral forces (X and Y) which causes a slight flex of the instrument tip. My logic then tells me they can then make the Z axis forces through calculated forces by the Z rail with some clever subtractive maths to remove trocar effect. But it seems obvious to read the Z axis force directly by the Z rail. So that then leaves the grip force. Now remember to close a jaw of a DV5 instrument you pull on two sets of cables (and I think they are now non stretch polymeric cables for this reason) - and by measuring the force on the cables (with a lot of fast maths) you can work out the force being transmitted to the tissue. How do they measure the force being pushed to the cables? Well my guess, and from what I can read ,they look at how hard they are turning the rotary interfaces on the instrument - and that could be a direct measure sensor (in the instrument or the actual drive pack of the robot) or it could estimate the force applied by how much power the motor is drawing to get closure.On the surface it all feels a bit messy - but the 10,000X compute would easily be able to handle those models in real time. Put those lateral, Z rail, grip forces (and some turning forces of the instrument) together. Run some algorithms to reduce any noise in the system - do some fancy subtraction and you could get a pretty complete and accurate set of forces being applied to the instrument and hence the tissue.
Look - I’m not an engineer - but logic says they have sensors near the tissue to get lateral forces - plus some arm movement forces and most likely some kind of cable grip assumption = force on tissue. Engineers - anything public - comment on it and correct me if I’m wrong. Or give me your suggestions - but only public info please.
Bottom line they have a lot of IP for sensors beyond the trocar - and as I said earlier that is the most likely way you get to measure forces with less noise. It would make logical sense to me that to try and be as accurate the instruments are different from non force feedback. The addition of sensors at the distal tip - and maybe some up in the housing at the robot connection area would make them different. Would be a reason they had to revalidate lives - and would be a definite reason they are more expensive than regular “non sensor” instruments. And I’ll come on this - as you may get the most high fidelity result but it may be the most complex and costly way to do it.
Medbot - Toumai - what did they do different?
Let’s move up the shaft and hit the trocar. Because me thinks that Intuitive has some solid IP around using distal strain gauges to get to their forces. So when I was over at Medbot - I saw them switch out trocars for a different trocar Model TRF85D that seemed to be a different material than metal. And they explained it was a sensing trocar. I’d understood the way they were measuring forces on the shaft of the instrument was directly through the trocar at the remote centre of motion.
Now from some public documents - Trocar sensing picks up everything passing through that port — shaft friction, body wall reaction forces, and tip forces all mixed together in a noisy signal. It requires significant software to decompose the signal and isolate actual tip interaction force. From a public presentation: “Toumai runs the system and potentially force sensing at 4,000 Hz with a 250 microsecond response time.”
Now it’s very hard for me to pull out the exact details - I need a Dwight Meglan here - because we are talking microseconds not the milliseconds when we talk telesurgery. I’m not sure but whatever they are doing needs to be fast - computer heavy and may rely on specific chip sets not the standard CPU. The trocar is super noisy and so to interpret instrument forces alone it most likely is running heavy compute. And as it is further away from the tissue it may be less accurate?
Why do I say that? Well I’ve sat on a DV5 with force feedback on, and Toumai with force feedback on. I never want to take sides, but for me the DV5 felt more “realistic” - I have to be careful here as I’m not sure from Toumai what is publicly out in the domain and if it is just the force gauge they are still showing or the force feedback to the controllers. Anyhow - having sat on four systems with force feedback, so far the best has been da Vinci.What I’m getting to is that I think it is a little more noticeable when you move from tip sensing to trocar sensing. BUT they do have it and this doesn’t appear to be on Intuitive’s IP
And a question I’ll ask is how good / bad does it have to be to have clinical benefit. Because even DV5 feel “synthetic”. If you take the elastic band (rubber band) and stretch it - I personally feel it is “filtered” as a signal. I know they have force multiplier settings and I know they have done some recent software updates. But even the best of the best DV5 - to me - still doesn’t not match the direct sensation of laparoscopy. And I can understand why some advanced users that do a lot of robotic and lot of lap have confided in me to say “I turned it off.” But I repeat - advanced users with very big experience.
So what about OTTAVA? What might they do?
Look I’m not a total mind reader… but I am a patent reader And as I say, what is written in patents rarely gets exactly reduced to practice. But it gives directional thinking. It also helps me to understand where companies might avoid the IP landscape. How they might work around to get Freedom to operate.
One note - software is much harder to patent. The algorithms if not coupled to specific methodologies are very difficult to secure claims on. So a lot of the IP (intellectual property) for these companies will be in the way they compute this stuff.
I guess that for Ottava to stay out of the bad lands of Intuitive force feedback - they probablywill need to get to force feedback in a slightly different way. I think the safest way is to focus more about what is happening back at the robotic arm - (The base of the instrument) - which keeps them away from the IP of the end mounted force sensors that may be well protected by Intuitive.
Note: Not all sensors are the same. Some are electromechanical and some are optical. I cannot get to the depth of what is and is not protected. But keep in mind there are fibre technologies that can be embedded in the shaft of an instrument that change light signals as those fibres bend. So I cannot be 100% sure that there are zero ways of sensing the forces in the distal part of the instrument that gets around Intuitive IP.
Now what I did dig up: ”Superposition-Based Force Estimation: The system utilizes mathematical superposition models combining information from 6-DOF force/torque sensors at the tool base alongside accelerometer telemetry to precisely isolate and estimate the force exerted at the tip of the instrument versus the friction generated at the entry trocar/access port.”This set of patents talks exactly about avoiding the need for distal sensors - using a highly mathematical model of the forces at the back of the instrument - along side knowing where it is in space; and mathematically deducting the trocar friction forces to get an estimate of forces on the instrument.
If this is how Ottava does do it (my total guess) then a Superposition Model most likely Bypasses Intuitive IP.
Because instead of measuring pure tissue contact and applied forces measured by sensors, JNJs system would measure a "corrupted" total force at the base of the arm. It then breaks down that total force mathematically using two distinct inputs:
The 6-DOF Force/Torque Sensor at the Base: Measures the cumulative force of gravity, arm movement, trocar friction, and tissue impact.
Then the Accelerometer Telemetry: Constantly tracks the exact spatial orientation and acceleration vectors of the arm.
By applying a dynamic superposition algorithm, the system could calculate exactly what the tool should feel like if it were moving through thin air at that specific angle and speed (gravity + inertia + estimated trocar resistance). The system could then subtract that value from the total measured base force, isolating the residual force. That leftover number is assumed to be the force exerted at the tip, which is then fed back via the haptic controllers to the surgeon's console master handles.
And there are some JNJ patents on how they segment out their hap[tic hand controller in a way to stop getting what is a cross contamination of forces of pitch and yaw and roll. Sorry that was a bit complex but they have some IP on allowing those forces to be transmitted back with higher fidelity.
Let me please remind you - this is all hypothetical assumptions by me reading a stack of their patents. Plus imposing my commercial thoughts on how you make sure you stay away from other company’s IP. It doesn’t mean any of this gets made real (reduced to practice) but it looks like Intuitive has tip sensing pretty boxed off - Microport has trocar level sensing deployed… and Ottava could be doing instrument base maths.
Why would this be great for JNJ? If they could do this and get reliable data to feed back through the hand controllers - it would mean that every instrument you click on Ottava is instantly force feedback. And the cost of the instrument doesn’t change. It is the same as a standard instrument. You’re not paying for complex sensory embeds. Plus as those sensors are not being steam sterilised - so the life of those sensors (they don’t exist) is not a worry. In theory if this is all done in sensing back in the robot arm - then going from no force feedback to force feedback is a software drop. Because the sensors and motors to pas that back to the surgeon are already in the robot.
For me it would make sense that Ottava would do a “this side of the trocar” maths analysis to get the forces. Grip force - by measuring the motors in the drive packs. Lateral and pull forces in the arms of the robot and the base of the instrument. That to me would make more sense. Now of course there is no such thing as a free ride - so what is the trade off?Latency: Running real-time superposition math can introduce microscopic delays between the physical tissue impact and the haptic feedback at the console. In surgery, haptic lag can cause over-retraction or tissue tearing. Or (as I have felt on older systems a weirdness to the sensation.)It comes back to “is this method good enough?” Because bad haptics is worse than no haptics.
And I think this is going to be a battle ion “good enough” to have the same clinical impact - not “perfect replication of forces.”Why? Because for near 20 million patients surgeons coped with no force feedback. Surgeons and mere mortals are extremely good at adapting to synthesised force feedback.
Oh yes - do you think your steering wheel in your car is directly coupled to your wheels? No there is power steering in the middle - and it gives you the sensation of the wheels turning - and you get used to different power steerings fast. There are a few auto journalists that will dig deep into the sensation and “feel” of the steering. But for most of us - we are not crashing everyday if it’s “good enough.”
Trocar Variance: A note on this as it is a pain. Remember trocar friction isn't static; it changes if the patient moves, if the incision stretches, or if body fluids lubricate the port during a three-hour case. Any software must continuously update its baseline friction model on the fly without a direct physical reading inside the abdomen. And that is no small challenge. I do wonder if this part of the equation is what will annoy surgeons of not done right as their force feedback will change over time. On the other hand - happens in lap surgery every day.
Feature | Intuitive Surgical IP Profile | JNJ IP Profile |
Sensor Placement | Distal: Directly on the instrument wrist/tip (inside the body). | Proximal: At the tool base/manipulator interface + accelerometer (outside the body). |
Friction Handling | Avoidance: Ignores trocar friction because the sensor is already past the pivot point. | Isolation: Uses a mathematical superposition model to subtract trocar noise from the total load. |
Sensor Type | Micro-strain gauges, in the shaft/tip. | Multi-DOF (6-Axis) force/torque sensors at the base paired with accelerometers. |
Economic Impact | Higher per-procedure disposable cost; complex instrument manufacturing. | Lower disposable tool cost; higher computational and calibration overhead on the robot arm. |
So what does all this mean?
If you’ve continued reading this far - well done ! But you could be asking “so what?”
For me it is a big “so what.” As if the trend continues - then there will be those that have force feedback and those that don’t. And those that don’t may be out of much of the contracting game (especially for higher end robotic installs.)
I think force feedback becomes the “standard” and I can even force a day where guidelines will start to say..
“The evidence is clear - especially in the novice subgroups - that the use of force feedback reduces over application of forces and tissue trauma that results in poorer patient outcomes and higher risks. Our guidelines say that for the first 300 cases force feedback should be used.”IF and it’s a big if - that every starts to come out. Like ICG in GYN in some countries is a strong guideline to use it. Then you will commercially not be viable as not that many institutions want to go against guidelines.
It will for sure be another tender tick box. And it will (oddly) start to feel obligatory for telesurgery. As a sort of belt and braces. Now I won’t go into why that is technically nonsense as 2 feet or 200 miles should not make a difference. But as psychologically driven animals - force feedback might just make the surgeon feel “more connected” and give more illusion they are right there.
Small realities can have big impacts in performance events like surgery.
Also we are barreling to wards automated steps - and the robot completing automations - and then chaining those automations into automated surgery steps - and chaining to supervised autonomy in surgery. I know many think it’s decades away…. Hmmmmm.
Grooth-H from NVIDIA may change that faster than we think. I won’t go into it but it’s a Vision language Action. Model (VLA) that will help robots to use visual data to perform real world actions.
I am a firm believer that augmenting that visual dat with physical sensor data will give way better results. It is no coincidence Intuitive is giving numeric forces back to the user and to the system.
Bottom line - get Force Feedback or be a dinosaur.
Haptics - the next frontier
I will not go into this now - but see force feedback as the early dawn of full haptics. As technology gets better - sensors get smaller / cheaper and we learn how to represent that data to surgeons in a high fidelity way. Palpation, roughness, wetness, temperature, stickiness etc etc will all find ways of being sense by surgeons at a remote console.
How much real world value that will have will be along debate - but if you can - why not if cost doesn’t go crazy.
The robot - will become more standardised. Just look at the 16 Xi clones. The standard instruments will become equal. The staplers will soon become equal(ish) and vision will become a war of ever increasing vision modes.Data - autonomy - and instrument features such as force feedback and haptics will (in my mind) be the subtle differences that mean I choose system A vs system B
I do believe it all brings marginal clinical gains - but it will bring significant commercial feature differentiation.
Just may feeling (get it - you see what I did there…)
Note - these are assumptions and opinions of the author and are meant for educational purposes only. The assumptions in thje post are based on public documents like Patents, websites and public company statements. Don't take them as fact.
