Compressed Air Movement System

I wanted to explore compressed air as a means of "unconventional locomotion" to gain a 50% weight bonus in beetle-weight combat robotics. Some existing examples of "unconventional locomotion" are gyro walkers, mechanical walkers, bristle drives, flyers, and shufflers. I noticed a lack of air based drive systems and decided it would be worthwhile to explore that gap. In this project, I worked with SJSU ASME to design and construct a prototype of Project Downforce (a misnomer): a Compressed Air Movement System.

In this project, I used Autodesk Fusion, 3D printing, and robotic components to design and build a partially functioning prototype which is capable of generating thrust using compressed air from an impeller and control movement in two degrees of freedom: forward/backward translation and left/right yaw rotation, like an air-powered tank drive. Unfortunately, the locomotion system ultimately fell short of the goal of independent movement due to power constraints and inefficiencies of the system.

In this Instructable, I will give high level instructions on how I designed and built Project Downforce and it's components, including how I overcame several hurdles in the process.

Key Components: Impeller Compressor Assembly, Thrust Vectoring Nozzle

pics: title, full project, full project CAD, impeller CAD, compressor assembly CAD, Thrust Nozzle CAD, Chassis CAD,

1. CAD software (Autodesk Fusion)
2. 3D Printer (Bambu Labs A1, P1S)
3. I used PLA for most of the bodies, and PETG/PLA for the impellers
4. Robotic Components
5. Battery/Power Supply (1050mAh, 6S, 135C)
6. 2X Motors (A3110 900KV Brushless FPV motors)
7. 2X BL ESCs (Pariah 70A AM32 ESC (preferred) / Aria 70A BLHeli Brushless ESC(used ::sadface::))
8. 2X Servos (MG996R)
9. Transceiver (Flysky FS-i6 6CH 2.4GHz Controller)
10. Power distribution board (JustCuz Robotics MOBO)
11. Receiver (FlySky FS2A 4CH AFHDS-2A Mini Receiver)
12. Other
13. Hardware (M2, M3 machine screws/nuts)
14. Electronics (wire, XT30 & XT60 connectors, Dupont connectors)
15. Tools
16. Soldering iron workstation (iron, solder, brass +sponge tip cleaner, helping hands, fume extractor)
17. 3D printer

TLDR: I talk a lot about design and Downforce lore.

The goal of Project Downforce was to build an air based movement system capable of moving a 4.5 Lb (~2040g) combat robot around an enclosed cage. This project was optimized for completion (as most individual project are), so most design decisions were based on what I could most feasibly accomplish, rather than the best engineering practice or most efficiency. I decided to try compressed air using impellers because I wanted a low center of thrust and traditional propellers would add a lot of height and result in unwieldy center of thrust far above the center of mass. Impellers are also slightly more durable than propellers, and are easier to protect. Unfortunately, I lack the aerodynamics skills to determine the effective power of impellers relative to propellers, and essentially hoped for the best (fluid mechanics is magic until I take ME111).

I knew I would need around 20N(2000g) of thrust to effectively move the robot around, but decided it would ultimately be easier to test by proof and build the entire robot rather than using testing rigs. I chose motors based on that value, hoping that the effective thrust would be in the ballpark of the stated manufacturer values (unfortunately 6000g of potential thrust was not enough). The first motor was a much smaller 2207.5 with ~1000g of potential thrust, which fried itself after running for about a minute during testing and got me in trouble for stinking up the club room with electronic smoke.

The name Downforce originated from a concept for a new combat robot: to use air-driven downforce to increase wheel traction and improve control (like the Mythen student car). However that idea quickly fell victim to scope creep as I wanted to use air for the entire locomotive system. The name was kept because it sounded cool (Win for aesthetics based design) (Loss for naming it CRAMS: CompRessed Air Movement System). In initial drawings of the project, there was a single impeller attached directly to the weapon, but that would prove hard to accomplish, and the thrust vectoring would be more difficult to design from one asymmetrical source of thrust. Instead I switched to two independently driven impellers, which I could more simply mirror and use like a tank drive.

The thrust vectoring would be difficult to accomplish, since most thrust vectoring designs are in a different application, so I designed my own system using a single servo to vector thrust along a 180 degree arc in a single plane of rotation. I accepted that I couldn't vector thrust without major inefficiencies (air velocity lost to changing directions).

In interest of future designs utilizing some aspects of Project Downforce (and my feeling bad about it not working), here's a non-exhaustive list of reasons/inefficiencies for the project failing to meet the goal: Lost thrust in thrust vectoring, lack of fully airtight enclosure (compressed air leaks out of the cracks between different parts), oscillating motor output at high throttle (might be due to ESC issues or motor issues, unresolved), non-optimal motors for application (FPV motors are meant for propellers, not impellers, and I don't understand motor choice well enough to choose better motors), design optimization for completion (I would prefer to optimize for performance, but I do not have unlimited time, energy, or motivation for personal projects), aerodynamic optimization of compressor assembly (I didn't have resources to test multiple volute sizes or impeller geometries).

I used Autodesk Fusion CAD software to model the impellers and other components.

High level step by step CAD guide to model the impellers:

1. Sketch (YZ plane) cross section of impeller and rotate top (aka shroud) and bottom (aka base) as surfaces, and duplicate on a rotational offset plane.
2. Sketch (XY plane) cross section of impeller blades and project curve onto top and bottom surfaces.
3. Loft a surface guided by the projected curves and thicken to create blades, cut top with rotational extrude and surface to surface fillet to smooth edge, then use circular pattern to duplicate.
4. Rotation extrude impeller base and shroud to join blades.

Note: thickening the blade surfaces sometimes fails due to excessive curvature, usually around the tip. Guide left abstract due to complex geometries required with different dimensions, blade geometries, aerodynamic optimization etc.

To fit the motor, an outrunner canister motor:

1. Rotation extrude cut to fit motor cylinder and bolt shaft.
2. Create square helix, duplicate in cylindrical pattern, and subtract to create cooling fan to the motor.

To model the volute (compressor assembly body):

1. Sketch circular base and holes for motor mounting and wiring
2. Create helix and project onto sketch for volute spiral. Offset spiral curve for wall.
3. Extrude base, walls (join to base), and lid (new body).

To attach the lid to the base of volute:

1. Offset spiral curve again for wall extension and extrude to slide lid on
2. Create sketch of cross section with holes for hardware, then extrude as new body, copy by circular pattern, and join with base walls by subtraction.

I modeled some simple 3D printed clips as hardware to attach lid to base, and tested to ensure they fit.

I also modeled and tested a mount for the motors.

Assembling the compressor was straightforward, but I'll later list some design considerations I made to ensure everything fit together.

The impellers were printed in PETG with PLA supports. Note that due to the complex shape of the impeller and unsupported shroud during printing, supports in between the blades are necessary, and must be removed in post-processing (ram them out with a pencil and hope for the best).

1. Bolt impeller onto motor (Tightly! otherwise impeller can slip against motor)
2. Affix motor mounts to volute base with screws
3. Attach lid to compressor body (I used clips)

Due to tolerances and plane tilt offsets in construction, I had to experiment with wall height of the compressor volute to make sure that the top of the impeller shroud made minimal friction contact with the compressor body. Essentially, make sure there are plane offsets between the bottom of the impeller and floor of the compressor body, as well as the top of the impeller shroud and ceiling of the compressor body lid. Also consider that at high RPM, the impeller may warp due to creep strain.

The thrust vectoring nozzle was the most complex component, but is simple in concept. It uses a servo to rotate the nozzle 180 degrees using two gears. It has many parts due to the complexity of 3D printing and connecting the bodies, and also that I needed to be able to iterate designs for tolerance and be able to swap in new components. I used primarily nuts and bolts to fix bodies to each other.

I'll list each body and a short high level how-to on how I modeled it. I use face offsets when possible to allow for tolerancing for 3D printed parts.

Stationary nozzle mount base (interfaces with compressor assembly output): Sketch (XY Plane) Base, extrude base, extrude join walls, extrude join lid with hole for nozzle

Rotating nozzle clip base(allows rotation but not sliding on mount base): (from same sketch base) Extrude base, extrude core, sketch holes for clips on construction plane 45 degree offset around Z-axis and extrude cut into core.

Rotating nozzle (directs air): on 45 degree offset plane Sketch walls, rotate extrude 180 degrees, and extrude join walls out. From Sketch base extrude join bottom circle walls to interface with clip base and extrude cut clip holes. Sketch (top surface) interface mounts to gear clip and extrude join.

Nozzle clips (to connect clip base and nozzle): Sketch (XY Plane) (complex geometry, requires testing), extrude.

Nozzle gear (to drive rotation of nozzle): Utility program Create Spur Gear. Move for face connect and axis align to nozzle top. Extrude cut interface mounts, and extrude join circle to interface with gear brace.

Stationary servo mount (mounts servo, connects to stationary nozzle mount base, gear brace, and chassis): Sketch (XY plane) base, with mount rectangle walls for servo (test), servo hardware mount holes (test), and gear brace mount tower. Extrude floor, extrude join walls, extrude join gear brace mount tower, extrude join hardware mount holes. Sketch (YZ plane) hardware mount holes to chassis and nozzle mount base, and extrude cut into walls. Extrude cut gear brace mount holes.

Gear brace (interfaces with nozzle gear and connects to servo mount tower): Sketch (from servo mount base) gear brace base and extrude.

Servo gear (attaches to servo arm): Sketch (XY Plane) servo arm (test for fit) and clip mount holes. Create with utility program Spur Gear and move to interface with servo arm. Extrude cut servo arm partially through gear, and extrude cut clip mount holes.

Servo gear clips (holds servo gear on servo arm): Sketch (YZ Plane) (Complex geometry, requires testing), extrude.

Assembly was again straightforward from the design. I'll again list design considerations for assembly:

1. Slide rotating nozzle clip base inside stationary nozzle mount base and affix to rotating nozzle on top with nozzle clips (requires some dexterity due to size).
2. Attach rotating nozzle gear to top of rotating nozzle using interface mounts.
3. Bolt stationary nozzle mount base to stationary servo mount.
4. Bolt gear brace to gear brace mount tower, ensuring it interfaces properly with nozzle gear.
5. Clip servo gear to servo arm, and slide servo into mount, then bolt to servo mount holes (not too tight, otherwise can deflect and gears wont interface).
6. Bolt to chassis with chassis mount holes.

Allow for tolerances in mounting hardware by testing fit first (ex I use 2.0 mm holes to directly interface and hold M2 screws, but 3.2 mm holes to loosely slide and hold M3 screws). Allow for tolerancing gear interface, I was able to just move the servo into place and bolt down, but everything can deflect just a little bit, allowing for the two gears to lose interface (which is useful for adjusting the rotation phase offset). There probably will be a gap between the nozzle base and the compressor assembly. Make sure that this assembly has full range of motion and doesn't conflict with the compressor body.

I designed the chassis using an assembly of the compressors and thrust vectoring nozzles as reference, then designed mounts for both. I modeled the chassis in one sketch layer (XY plane) and extruded the base and walls, with holes extruded as needed. I also created the support arms that mount the compressor bodies, which include a mounting hole to reduce deflection due to the distance between the mount to the chassis and the motor. The arms also have a channel for the wires to sit inside. I created a surface sketch to replicate the mounts for the thrust vectoring nozzle.

The assembly is completed by bolting everything together at the mounts.

Setting up the robotics was similar to other beetle-weight robots, with the robotic components being controlled by a radio transmitter and electronics running off of a market motherboard (power distributor and receiver combo).

Controller robotics:

1. Transmitter/controller sends PWM signals to the receiver's 4 channels
2. Receiver is soldered onto MOBO, which has Male Dupont connectors to CH1-4
3. CH 3&4 control servo rotation (~90 degrees rotation, 1.0ms-2.0ms PWM @50Hz. Note: I didn't use a travel tuner but one could get 180 degrees rotation with a wider pulse range e.g. 0.5ms-2.5ms PWM).
4. Ch 2 controls the two ESCs which control the two impeller motors, which are signal connected with DIY Dupont connectors soldered together to duplicate the signal to the two ESCs' signal wires.
5. Controller setup: I swapped the left and right sticks input in the back so I could use right stick for CH 3&4 for the servo movement. I mixed CH 3&4 for tank drive. Sticks mode 2. Bind transceiver to receiver. Reverse output for CH 3/4 in settings or physically swap CH3/4 to ensure servo output is as desired. Set transceiver endpoints for CH 3/4 to ensure that the servos don't try to ram the nozzle into anything.

Power/electronics:

1. Soldered XT30 connectors onto MOBO (F for battery, M for power output)
2. Soldered converter from XT60F-XT30M for battery (MOBO has soldered XT30F power input)
3. Soldered XT30F connectors onto ESCs
4. Soldered Banana connectors onto motor wires and ESC output
5. Made sure to heat wrap as much as possible, could also hot glue over the MOBO for protection

With set up complete, plug everything in where it needs to go, just make sure all connectors are in the right orientation and that connections are solid. The electronics all fit into the chassis as shown, but it is a bit tight and I didn't bother to use a lid for the prototype.

Note: The motherboard is not necessary but it simplifies electronics setup, takes up less volume, and allows for 7.4V servo power from a separate BEC.

To operate:

Use left stick up/down to throttle the impeller motors. With an air flow, use right stick as a normal tank drive for forward/backward and yaw movement, with up/down controlling forward/backward movement, and left/right controlling CCW/CW movement.

Video Demonstration: a short demo of Project Downforce turning (very slowly) at ~60% throttle (beyond which motor control is lost).

Autodesk Fusion Files: CAD models of Impeller, compressor housing, and thrust nozzles for reference.

Finishing Notes:

Hopefully you've found this Instructable useful or interesting. As my first solo engineering project, I certainly learned a lot, from robotics controls like PWM to design and modeling skills and even some personal insights on working on personal projects, such as scope creep and compromising for completion. If you have any questions or comments, feel free to leave them below.