Inflatable Silicone Robotic Manipulator



I came across research conducted by the Harvard soft robotics team, where they created starfish-shaped silicone robots that move when inflated with air. See this Instructable for videos and info. A lightweight and soft-touch manipulator would be ideal for picking up delicate objects off the ocean floor or collecting sea flora and fauna.

I'd like to adapt this technology for OpenROV: The unique spin on this would be to fill the silicone actuator with surrounding water – instead of air - via a onboard pump which ideally will not dramatically increase the ROV's weight and alter its neutral-buoyancy.

I'm unfamiliar with the engineering concepts surrounding payloads, so I'm not sure how this type of payload will affect buoyancy. I'd love to get feedback from you all as I work through this project!


This is an interesting application for a technology that's near and dear to my heart. Since I first started doing soft robotics experiments people have suggested using soft grippers and tentacles for underwater exploration, especially as they could be hydraulically powered and potentially neutrally buoyant. I've got documentation on the experiments I've done here.

It seems like you've got access to a laser cutter and a shop. You might be able to create a simple mold like Frank Kinnaman does, except constructing it out of laser cut acrylic instead of machining it out of wax.


COOL ¡¡¡¡¡¡¡¡¡¡¡¡¡¡


Very interesting! However, since it is pneumatic, pressure at depth is going to be an issue. Buoyancy will also be affected greatly as air is pushed in and the displaced volume of water increases. I think this is going to be the number one issue, since it will dynamically change the characteristics of the ROV. You will also not be able to direct the gripper unless it was directly attached to the kit and move the ROV to position the manipulator.

1. Dynamic Buoyancy = delta of volume water displaced when gripper actuates.

2. Pressure -- 100m (max depth of ROV) is between 150 and 160psi, I think technically it's about 160 in salt water. Please confirm that number for your calculations. Say the total area of the gripper is 16 in^2 (I'm just guessing for an example and using 150psi) => Your pneumatics would need to supply 2400 lbs of force at max depth...just to equalize. You would need additional for actuation. Not a show stopper by any means, but you would need a small dive bottle or co2 cartridge and then adjust for the change in buoyancy due to the cartridge once it is discharged.

Long story short, you need to address the pressure stresses on the gripper and supply tubing, location/attachment to ROV, and buoyancy affectations on ROV control. I think you'll find, though the idea is in fact cool and under-articulated actuation is the way to go (even our hands are under articulated...I have an issue with anthropomorphic robotics researchers in this area...they spend so much time and money on joint articulation, and they seem to ignore the fact that the human hand is really under articulated given the number of joints. Really it is, can you manipulate every joint in your hand independent of all others? :) ...yet I digress...)

Anyway, now that I've read Gianteye's reply, that's the way to go, oil filled actuation solves these issues, LOL. So, awesome, have fun with this one!


Hi Jim:

Standard sea water density, approved for calculations goes from 1025 Kg/m³ to 1032 Kg/m³ depending on season and salinity.

For ship's building industry, 1025 Kg/m³ is the std one, that is further corrected for other conditions, by following the International Load Lines Convention rules.

For our use, 1025 Kg/m³ is the suitable one.

Hence hydrostatic pressure P at depth D, in sea water, directly comes from:

P=d g D + Patm

P in Pascals (pressure)

d in Kg/m³ (density)

g in m/s² (gravity)

D in meters (depth)

Patm= Atmospheric pressure in Pascals.

If inner pressure of the vessel is constant, and equal to Patm, then, the Pressure working against that vessel results from.

P= d g D

D= 100m

P= 1025 x 9.81 x 100= 1005525 Pa= 1005525 N/m² (Newton/ square meter)= 145.8 psi

I think, the best way for using soft manipulators is by a simple "sorrounding" water pump, able to provide the required pressure increment to the device.

By using external water, buoyancy increment, would be exactly balanced by the weight increment (water buoyancy in water is null; water weight in water is also null), and hence no change in static balance would happen.



Would you have a huge buoyancy change by using compressed gas to displace water in the system, as in having a liquid co2 cartridge push pressurized gas into a chamber filled with water (gas taking the place of the piston in diagram fig. 28 C,here) and using the pressurized water to actuate the soft robot? I haven't considered this much in the context of an underwater robot, but it's how WET and other water feature companies power a lot of their high pressure fountains and it seems like an interesting method to try.


Hi Gianteye:

Well, using liquid CO2 would turn the whole installation into quite a complex thing.

In fact, the liquid phase, into this context, is only useful for storage.

When working pressure is required, liquid is turned into gas by means of the release valve, that makes pressure drop below the vapor tension of the liquid, at the working temperature (See Antoine's equation for aproximate calculation).

As it is gas and not liquid what is working, gas pressure must be equal to:

Working pressure+Ambient pressure.

It means that gas consumption will depend on depth.

A compressor+cooling unit is required for gas phase recovery, in order not to waste gass each time it's used.

Water is the obvious solution.

A centrifugal pump, is a device that is designed for rising pressure, and not for achieving a given one.

I mean that the pump will provide a fixed pressure difference (depending on the pump's curve), that will not depend on the sorrounding `pressure, but only on the pump's RPM.

Hence, it does not require any container vessel, neither any kind of regulating-recovery device.

Water is taken from the environment, pressure is dynamically raised, and once used, released to the environment again. Only an ECS for RPM control(It can be solved with fixed RPM if the pump is correctly choosen) and a release valve for pressure control, are needed.

The valve can be of the type "safety+manual opening".

It will open if the "OPEN" command is received, or if pressure rises over a given maximum "Safety Working Pressure".

Energy consumption would be almost constant.

If a high pressure differential is required (stronger grip capability), while holding the pump size, an alternative pump would make the work. But those devices are slower, energy consumption is higher, and their mechanics a lot more complex.



Let's say the gripping actuator has a volume of 20 cm³.

For such a little device, a single piston (syringe like), directly coupled to the grip, working with water, would make the work.

A motor, driving a screw-nut device, could move the piston inside the syringe.

Regards again.


Have a look at this ¡¡

From: Cornell Creative Machines Lab:


- This gripper works because of a process that is called "jamming." When a granular material such as coffee is compressed, it becomes very rigid. As the pressure increases so does the amount of friction between the individual grains. This effectively locks the grains in place. You may have observed this phenomenon while handling bags of coffee grounds. A vacuum packed bag of coffee grounds is rock hard as long as the seal remains intact. But as soon as the seal is broken, the coffee become soft and pliable and can be poured like a fluid. This process happens with many granular materials such as rice, couscous and even sand. We are utilizing this process to make an amorphous robot gripper. A balloon is filled coffee grounds and attached to an air hose. When balloon is slightly pressurized the grounds are loose and easily rearranged. By pressing the balloon against an object, the grounds will move around it and take its shape. But when the air is sucked out of the balloon, the grounds are compressed and grip the object. The rubber surface of the balloon also helps to keep a hold of the object.-



I am not sure if the Coffee grip is working with water instead of air ...

Dir you already test that the Silicone Manipulator works with liquids instead of air?


How is this coming? Any new developments?