Fluid Dynamics: Maximizing Downforce Between a Robot and the Floor

[sroberti]

Good evening, all!

Some quick background: I am working on a robot for a competition that strictly limits entry weights to 3lb. Robot speed & acceleration are highly desirable in this competition, so I have been focusing on ways to maximize my entry's performance. Drive motors for these weight ranges are extremely power-dense, so my current iteration can apply far more power to its wheels than it can practically use (this is typical for entries in this weight class). The limit to static friction (i.e. the maximum force that can be applied to the ground by the wheels) is determined by the coefficient of friction (which I can improve by using the best-quality wheels available) and the normal force (which, because all entries are limited to 3lb, is hard but not impossible to improve).

I've been investigating the possibility of using a blower fan/pump of some sort to create a downward suction force, increasing the normal force on my robot's wheels and thus increasing the portion of my drive power that can be used. However, the exact characteristics of the system are somewhat beyond me at the moment, and I don't have any formal education in this field to help me sanity check my ideas.
Broadly-speaking, I see two main approaches:

1. Use a high static-pressure fan to reduce the air pressure below the robot's bottom plate by as much as possible.
2. Use a high-volume/velocity fan, coupled with an appropriately-shaped robot bottom surface, to induce a Venturi effect underneath the robot.

The first option would be much simpler, mechanically, but I worry that the air leakage around the edges of the robot will put a severe limit on the pressure difference that can be maintained without spending an insane amount of power on the fan. The second option sounds dramatically more difficult to design, but seems like it would let me treat the "leaky edges" as "input nozzles", and would have a nice side-effect of possibly letting me air-cool some of my hotter components.

Anyway, that's my current mental dilemma. Am I remotely on the right track here? Looking forward to any insights y'all might be willing to share.
Thanks in advance.

 

[jack action]

You might be interested in those race cars:

https://en.wikipedia.org/wiki/Chaparral_2J
https://en.wikipedia.org/wiki/Brabham_BT46

 

[erobz]

1. Use a high static-pressure fan to reduce the air pressure below the robot's bottom plate by as much as possible.
2. Use a high-volume/velocity fan, coupled with an appropriately-shaped robot bottom surface, to induce a Venturi effect underneath the robot.

Jetting air upward through a fan connected to the body, applies a downward force the body of the vehicle and would seem to be more direct than trying to pull vacuum on the underside?

 

[sroberti]

You might be interested in those race cars:

https://en.wikipedia.org/wiki/Chaparral_2J
https://en.wikipedia.org/wiki/Brabham_BT46

Indeed, the Fan Car is a big part of the inspiration here. I would personally be very interested in seeing schematics of the underside of the BT46; specifically, I have no idea if the bulk of the pressure differential is caused by the seal from the skirt, by the change in pressure from the Venturi effect, or just by throwing so much power at the fan that it can do a bit of both (it IS an F1 engine, after all).

Intuitively, I'm guessing that the lower the static air pressure inside the skirt gets, the greater the power requirements to accelerate that air through the fan by some amount (since the ratio of inner pressure to ambient pressure would be lower). Not quite sure if there's a way I should be formalizing this, though.

 

[sroberti]

Jetting air upward through a fan connected to the body, applies a downward force the body of the vehicle and would seem to be more direct than trying to pull vacuum on the underside?

More direct, perhaps, but I imagine it would also be far less power-efficient. I am limited in the mass I can dedicate to batteries and motors, so I'm hoping to make use of ground effect to get the greatest extra downforce per weight investment that I can manage.

 

[erobz]

I feel like the propeller car in the video "Robot Car Climbs Walls" would be fairly light weight. Who knows.

 

[sroberti]

I feel like the propeller car in the video "Robot Car Climbs Walls" would be fairly light weight. Who knows.

It also appears to be dedicating roughly one-third of its mass to its propellers and associated motors, which is far more than I would be able to afford.

Unrelated, but I've also come across an interesting paper on the subject, one which I hadn't found in my previous searches: https://www.mdpi.com/1996-1073/15/15/5549 .

 

[Baluncore]

Robots do not need wheels to have high acceleration, or to be fast.

A robot benefits from low mass, if it hops like a frog.

Consider a grasshopper.
Ground pressure can be very high during the takeoff.
It then flies through the air, and lands, to turn, then jump again.
Foldable wings might give it a glideslope to the finish line.

The sport of parkour employs inertia of the body, on a zigzag path, to increase traction.

 

[sroberti]

Robots do not need wheels to have high acceleration, or to be fast.

A robot benefits from low mass, if it hops like a frog.

Consider a grasshopper.
Ground pressure can be very high during the takeoff.
It then flies through the air, and lands, to turn, then jump again.
Foldable wings might give it a glideslope to the finish line.

The sport of parkour employs inertia of the body, on a zigzag path, to increase traction.

Sadly, this is completely infeasible in this case. For context, the competition in question is NHRL, a combat robotics league; my robot will need to stay grounded to effectively maneuver around opponents. Furthermore, at this scale (remember, this is a 3lb bot), mechanisms to store and release large amounts of energy quickly (as would be required for a "jump") tend to be quite heavy. Instead, it tends to be more efficient to use systems that can use a much smaller amount of power over a longer time. In this case, the continuous acceleration of wheels would allow the associated mechanisms to be far lighter, even for an identical average power usage.
Regardless, I appreciate your insight!

 

[jack action]

I have no idea if the bulk of the pressure differential is caused by the seal from the skirt, by the change in pressure from the Venturi effect,

There is no venturi effect here. The goal is just to remove the air from under the vehicle. Less air, lower pressure.

Intuitively, I'm guessing that the lower the static air pressure inside the skirt gets, the greater the power requirements to accelerate that air through the fan by some amount

The power needed by the fan depends only on the quantity of air leaking in. If one could achieve perfect sealing, one could just pump the air from under the vehicle manually before starting it and no fan would be needed.

Some type of skirt like on hovercrafts seems the ideal method of sealing. Hovercrafts are just the reverse of what you want to achieve: increasing the pressure underneath instead of lowering it.

 

[Baluncore]

The power needed by the fan depends only on the quantity of air leaking in. If one could achieve perfect sealing, one could just pump the air from under the vehicle manually before starting it and no fan would be needed.

A fast spinning aluminium disc, mounted on a vertical shaft below the vehicle, would act like a centrifugal pump and throw air outwards. That will reduce the air pressure between the disc and the floor. The down force on the disc (air pressure above) will pull the motor shaft downwards, pressing the vehicle onto the floor or road surface. No skirt is needed.

Once the low pressure is established, there will be little airflow, as there is no flow inlet to the pump. That suggests a low power requirement, apart from viscous drag of the air that remains in the shear zone, between the disc and the floor. It should be possible to calculate the down force as a function of disc diameter and RPM. We need some idea of the vehicle plan size.

At the periphery of the disc, air will be thrown out from the disc lower surface, then flow back in along the floor. That parasitic-drag flow, will be reduced by having the disc closer to the floor, but then drag in the shear zone will increase. Air thrown out by the top of the disc, can be pulled down through the motor if cooling is needed.

By mounting the motor in soft supports, with some angular compliance, gyroscopic forces will be reduced when crossing a rough surface. It would also reduce the vibration due to any imbalance of the disc.

 

[Tom.G]

Not the question you asked but possibly interesting.

You mentioned that you have much excess torque available for the drive wheels.

I seem to recall that for drag-racing, a 10% overspeed of the tires yields the maximum acceleration.

The technology needed to implement that control may not be practical for your weight limit though. You would need a 5th wheel to sense vehicle speed and a way to sense and control the drive-wheel speed.

I'm sure others here, with much more racing experience, can correct that 10% number if needed.

Cheers,
Tom

p.s. Please keep us updated on the both the project and the competition.

 

[f95toli]

Maybe it would be worth having a look at some micromouse designs?
https://hackaday.com/2023/05/28/the-fascinating-evolution-of-micromouse/

They use fans to increase traction; this is how they can turn so incredibly quickly.

 

[sroberti]

Maybe it would be worth having a look at some micromouse designs?
https://hackaday.com/2023/05/28/the-fascinating-evolution-of-micromouse/

They use fans to increase traction; this is how they can turn so incredibly quickly.

Oh wow, I hadn't seen this implementation before! I wasn't able to find any more detailed articles on their design from my quick search, but on visual inspection, this does look very very close to what I'd like to do.

Not the question you asked but possibly interesting.

You mentioned that you have much excess torque available for the drive wheels.

I seem to recall that for drag-racing, a 10% overspeed of the tires yields the maximum acceleration.

The technology needed to implement that control may not be practical for your weight limit though. You would need a 5th wheel to sense vehicle speed and a way to sense and control the drive-wheel speed.

I'm sure others here, with much more racing experience, can correct that 10% number if needed.

Cheers,
Tom

p.s. Please keep us updated on the both the project and the competition.

I heard something similar for the first time just the other day from other builders, funnily enough. If I were to try and implement throttle-limiting to try and optimize acceleration, I would personally probably try to use hall-effect sensors to track wheel or motor RPM, and maybe estimate a threshold where the throttle/RPM ratio starts to change very suddenly. It's not unheard of to use hall-effect sensors for other reasons in this sport (with an appropriate ESC, they can be used to dramatically improve start-up torque for brushless motors), though it is uncommon for this weight class.

Since there's been a bit of interest, I'll go ahead and share some of the past progress I've made, which is serving as a sort of "starting point" for this stage of the project. This is Morq, my beetleweight combat robot. The version you see here fought twice at NHRL in September, and four times at Makerfaire Orlando in November. In its current form, Morq is a pretty speedy bot; however, now that I have a significant gap before any future competitions, I am spending the time to use my lessons learned to upgrade the robot, and to try my best to add the aero-downforce system that this thread is focused around. Many of the features you see in the images might be likely to change in the redesign, but I'm happy to provide details if anyone feels it might be relevant.

As for the downforce system itself, I'm starting to see a lot of appeal in a "periphery jet" design, based on my readings from the paper I linked above. Essentially, it would eliminate the need for a solid "skirt", which could become snagged on damaged or uneven sections of floor, and would instead push high-pressure air in a thin wall around the entire perimeter of the low-pressure surface. The geometry of the system might be complex, but could be 3D-printed quite easily.

My current dilemma is fan placement. The system described in the paper used a large number of small fans to pull air from the low-pressure volume into the high-pressure chamber; because of the differences in scale, I find it very unlikely that I'll be able to fit more than one or two pressurizing fans, so I'll need to optimize intake placement and other parameters carefully.

This might be a tall order, but does anyone have any suggestions for CFD software that I might be able to use to simulate air pressure, airflow, etc, in a 3D volume? Do any non-commercial options exist for this? I'm fairly tech-savvy, and I'm a decent programmer, so I'm very open to the possibility of using simulation libraries for programming languages.

Oh, and in closing: Thanks a ton to everyone who's participated in this discussion so far!

 

[Gunsuit]

A fast spinning aluminium disc, mounted on a vertical shaft below the vehicle, would act like a centrifugal pump and throw air outwards. That will reduce the air pressure between the disc and the floor. The down force on the disc (air pressure above) will pull the motor shaft downwards, pressing the vehicle onto the floor or road surface. No skirt is needed.

Once the low pressure is established, there will be little airflow, as there is no flow inlet to the pump. That suggests a low power requirement, apart from viscous drag of the air that remains in the shear zone, between the disc and the floor. It should be possible to calculate the down force as a function of disc diameter and RPM. We need some idea of the vehicle plan size.

At the periphery of the disc, air will be thrown out from the disc lower surface, then flow back in along the floor. That parasitic-drag flow, will be reduced by having the disc closer to the floor, but then drag in the shear zone will increase. Air thrown out by the top of the disc, can be pulled down through the motor if cooling is needed.

By mounting the motor in soft supports, with some angular compliance, gyroscopic forces will be reduced when crossing a rough surface. It would also reduce the vibration due to any imbalance of the disc.

This sounds incredibly well suited for the task, and the lack of a skirt makes it an ideal solution. Do you have an example implementation of this, or more info you can point us to?

 

[Baluncore]

Welcome to PF.

This sounds incredibly well suited for the task, and the lack of a skirt makes it an ideal solution. Do you have an example implementation of this, or more info you can point us to?

No example, just an idea of what it might lead to, if a centrifugal air compressor was used as the vacuum pump.

The practical limit will need some computation. When it operates, the density of the air is reduced, which will reduce the drag of the disc's lower surface. Maybe the bottom of the disc will need strakes, or simply a rough lower surface near the perimeter.

I was imagining an octopus sucker, or a limpet stuck to a rock. The force improves more rapidly, as the square of the radius. If the disc is big, only the outer annulus will be operating as a pump. The lower velocity near the centre of the disc makes a poor pump, but provides a useful area of suction.

 

[Gunsuit]

I mocked this up in Lego technic with an impeller fan from a car vac. It doesn't move straight and tends to slide down the wall instead of stick to it in place like the skirted vacuum that I was using before.

Is it better to have straight fins on the disc? I was thinking of 3D printing a disc with strakes like you described. The car totally sucks onto the wall from a couple inches away, but it doesn't quite stay sucked in place unless it's on the ceiling. Should the fins be shorter on the disc than what I have? Should it have any skirting at all, or does that need to ideally be wide open all around?

I found a good example that works very well from the looks of the video, but I don't want to use an expensive EDF if I can get away with something simpler/cheaper.

Suction car robot "Ibex" from Operative RC:​

 

[Baluncore]

A car vac is designed to pull a continuous volume of air. That is why it has the big strakes on the fan. The impeller would need some cowl or skirt to reduce the outer edge leakage. The work done to turn a car vac impeller will turn the vehicle, so we need less total surface drag.

I would experiment with a smooth disc, very close to the surface, not yet mounted on a car. Measure the RPM, diameter, and suction down force. There should be some obvious relationships, that will lead to an optimised design. The motor needs to be high RPM, with a lower torque, so it will be longer and thinner, not short and fat.

I expect it will need to rotate fast and have as big a diameter as possible. That will require balancing, so 3D printing will probably not work. The disc will be subjected to high radial forces, so it will need to be strong to avoid bursting. I would make the disc from aluminium or polycarbonate sheet. I am trying to think of a source for a ready-made disc.

It will need to be light-weight, such as polycarbonate, maybe like a CD, but that will need a one-sided hub. An aluminium disc is also possible, something like the flat lid from a saucepan.

There may be a parallel here, with the flat-disc rotors in a Tesla turbine.
Only this idea sucks, as it is not a driven turbine.
https://en.wikipedia.org/wiki/Tesla_turbine

This time of the year is too busy.
Thinking of a neat and elegant solution will have to wait.

 

[Nik_2213]

IIRC, a recent YouTube video mentioned that XB-70 'Valkyrie' had a small, fifth, 'jockey wheel' on a main, quad-wheel bogies. This provided 'ground truth' of runway speed during hard braking, was more reliable 'in extremis' than era's Doppler sonar or radar. Also pics etc at...
https://aviation.stackexchange.com/...ose-of-the-extra-wheel-on-the-xb-70-main-gear
Conversely, as implied above, farm machinery may have Doppler sensing of true ground speed, as drive-wheels may have ~10% RPM slippage at optimal traction...