Devlog 9: README Finished, Screws Added, and Diagrams Finalized!

This devlog will be short since most of the work involved writing and documentation.

README Finished!

It took me a while to stop procrastinating, sit down, and write the README.md, but it’s finally done!

I gave a basic introduction to each part of ASTRALIS and explained how it works. I also added dedicated sections covering the electronics, software, and bill of materials. I also added some renders which I’ve included on this devlog.

Outside of the README, I added the remaining screws to the CAD and finalized the project diagrams.

Next Steps

• Submit the project
• Revise it based on feedback

Devlog 8: Aerial Manipulator Redone, BOM Revised, and ¼ of Programming Done!

Across the last three days, I’ve had a realization.

My project was a little too expensive for me to want to self-fund, even with the grant. I felt that I was paying too much of a premium for certain items. In addition, after a comment I got on my previous devlog (thanks @Abracadabra3!), I realized the servo motors I was using probably wouldn’t hold enough weight to create good research data.

So, to solve both of these problems (or at least reduce them), I switched from the Dynamixel XL430/XL330 ecosystem to the Waveshare ST3215 serial bus servos.

Reasons to Switch

Advantages

• Much cheaper per servo ($20.63 vs. $27.50)
• Much cheaper power/control distribution board ($4.99 vs. ~$40)
• Documentation is still acceptable, and CAD files are still provided (even if they’re a little janky)
• Significantly higher rated torque (1 N·m vs. 0.28 N·m)
• Standardized servo motors across the entire arm (making assembly much easier)

Accepted Losses

• Slight loss of precision (no adjustable PID/feedforward gains)
  • I deemed this overkill for Astralis.
• Increased weight per servo
  • 12 g heavier than the XL430 and 50 g heavier than the XL330
• Increased current draw (slightly reducing battery life)

Robotic Arm

I adapted the robotic arm to be compatible with the Waveshare ST3215 servos, with an emphasis on reducing the weight of each joint.

The biggest change was redesigning the base joint so that it only encloses the servo instead of mounting it off to the side. I also used more generous material pocketing throughout the design to reduce weight.

I further increased the arm’s maximum reach so that it can reach the landing gear if needed.

I also started doing the math on how much the arm can actually hold (currently using both a whiteboard and Excel), but I haven’t finished it yet. This isn’t a huge priority right now, as I’d like to improve the linkage system and continue working on the code first.

Coding

I started by creating the architecture of my codebase. I also established the basic MAVLink initialization code and built a listener that filters messages such as battery data (voltage and battery remaining) and GPS data.

Files Worked On

• vehicle_states.py
• mavlink_interface.py

Next Steps

• Finalize the tilting pod CAD and select ball end and threaded rod options
• Finish basic drone connectivity
• Write the README and submit for initial review
• Add all fasteners to the CAD model and clean it up for Onshape publication
• Calculate the maximum payload the robotic arm can carry

Devlog 7: Renders, BOMs, and the Start of Software!

After taking a short break from ASTRALIS to avoid burnout (and do college application stuff), I’m back to continue refining the design and begin software development. With the mechanical design mostly complete, I’ve shifted focus toward reducing costs, preparing the project for funding submission, and laying the groundwork for future testing.

General Design Updates

The primary goal during this stage was standardizing hardware and optimizing the Bill of Materials while preparing the project for prototyping.

• Standardized most fasteners around M3×8 self-threading Phillips Plastite screws to simplify assembly and reduce the variety of hardware required (hopefully to use only one small sorting box)
• Updated the Bill of Materials with nearly every required component. The only remaining hardware to source is the tilt pod pushrods and ball ends, which will be selected after prototyping.
• Reused hardware from previous combat robotics projects wherever possible to reduce overall project cost (mostly m3 and m4 fasteners).
• Created renders of ASTRALIS in Fusion 360 for documentation and future presentations (maker portfolios!)
• Finalized the electronics diagram, completing the planned electrical architecture.

Beginning Software Development

With the mechanical design approaching completion, I also began planning the companion software architecture that will run on my Raspberry Pi 4.

The initial implementation will focus on:

• Establishing communication with the Pixhawk flight controller through MAVLink.
• Initializing the onboard camera.
• Developing a preliminary arm-disturbance test routine.
• Recording telemetry and experiment logs for later analysis.

The long-term goal is to use this software as the foundation for testing disturbance rejection using the aerial manipulator.

Future Plans

To prepare for the hardware funding request, I’ll continue improving the design where I can while mainly directing effort towards software development.

• Optimize the remaining Bill of Materials.
• Develop the preliminary companion software.
• Implement a basic arm-disturbance test with automatic log generation.

Devlog 6: Aerial Manipulator Finalized, Tilt Pods Finished, and Frame Finished

Over the past few days, I finalized the overall airframe, completed the aerial manipulator, and finished the design of the thrust-vectoring tilt pods. This was probably one of the toughest and most detailed-focus stages of the project so far (definitely stayed up super late fixing it sometimes!), prompting several redesigns to accommodate the Dynamixel XL330-M288-T servos while also standardizing hardware throughout the drone.

Note: I have another CAD timelapse recorded in Lookout that I’m still trying to transfer into Stardance.

General Design Updates

A major focus during this stage was reducing complexity and improving ease of assembly.

• Standardized as many fasteners as possible across the drone.
• Reused hardware from previous combat robotics and robotics projects whenever practical.
• Organized the CAD assembly by separating self-tapping screws into dedicated PLASTITES folders while leaving machine screws in the default folders to simplify the assembly process.

Landing Gear

The landing gear went through two complete design iterations.

The original concept used a dedicated clamp that attached independently to the carbon fiber arms. After evaluating the design, I integrated the landing gear directly into the arm end cap instead.

This redesign:

• Reduced the total number of printed components.
• Increased the distance between the landing feet, improving stability.
• Maximizes operating space of arm.
• Minimized interference with the thrust-vectoring pod geometry.

Each landing gear assembly consists of three printed components, including TPU feet intended to improve grip and absorb landing impacts.

Tilt Pods

The thrust-vectoring pods are now mechanically complete.

The largest addition during this stage was designing the pushrod linkage connecting each servo horn to its corresponding tilt pod.
At the moment:

• The linkage geometry has been finalized.
• The linkage components will be 3D printed and prototyped.
• Mechanical advantage calculations and linkage calibration still need to be completed before the CAD animation accurately reflects the pod motion.

Aerial Manipulator

The aerial manipulator has now been finalized.

One of the largest changes was redesigning the arm to properly mount the Dynamixel XL330-M288-T servos internally instead of externally.

Benefits of the redesign include:

• Cleaner overall appearance.
• More compact packaging.

The servos will be mounted using the provided M2 self-tapping screws directly into the printed components.

Battery Holder V2

The battery mounting system also received a significant redesign.

Originally, the battery was positioned underneath the frame because I assumed lowering the center of gravity would improve stability. After learning more about quadcopter dynamics, I realized this configuration introduces a pendulum effect that complicates flight control.

The updated design mounts the battery on top of the frame using a PLA battery cradle secured with four M4 machine screws and a Velcro strap.

This redesign also frees the entire underside of the frame for the aerial manipulator, allowing the arm to rotate through a full 360° without the battery obstructing its workspace.

Next Steps

• Assign materials to all CAD components and calculate the drone’s center of gravity.
• Finalize the complete electrical component placement within the body and wiring diagram.
• Characterize the thrust-vectoring linkage and determine its mechanical advantage.
• Begin firmware development and hardware testing routines.
• Write assembly documentation and build guides.
• Begin printing and validating mechanical prototypes.

Devlog 5: Aerial Manipulator Design

Over the past couple of days, I completed the aerial manipulator — the robot arm — for the drone. The current design assumes the use of DYNAMIXEL XL430 and XL330 servos, though the final servo mounting is not fully finalized in CAD yet.

Cool Things

• The arm is animated in CAD
• The full manipulator is now integrated into the drone assembly
• I designed and compared two different gripper concepts

The Arm

The arm uses DYNAMIXEL smart servos because they provide positional feedback and include many features that may be useful later.

The current manipulator is a 4-DOF robotic arm with a gripper:

• Base rotation about the Z-axis
• Shoulder rotation about the X-axis
• Elbow rotation about the X-axis
• Wrist rotation about the X-axis
• Servo-driven gripper as the end effector

The arm is intended to be fully 3D printed, allowing for fast iteration and easier replacement of parts if the design changes.

The Gripper

The bulk of the time was spent on the gripper, partly because I designed two different models:

1. A linkage gripper
2. A standard parallel gripper, inspired by this guide: 3D Printed Parallel Gripper for Robotic Arms

Linkages…

The linkage gripper was designed with compactness in mind. It used a gear-driven linkage system and was intended to use TPU fingers to grip rigid payloads.

However, maintaining printing tolerances and designing around the linkage geometry made the CAD extremely messy. The mechanism had too many small pivot points, tolerance-sensitive parts, and potential failure points. Because of that, I could not justify printing and prototyping this version.

The main focus of the project is not the gripper itself. The focus is on testing how the drone handles disturbances induced by arm movement and held payloads. Because of this, I need full confidence that the gripper will work reliably without spending too much project time reprinting and redesigning it.

Parallel Gripper!

Following the Instructables guide, I designed a simpler parallel gripper using a similar bearing, rod, and rack-and-pinion layout.

This gripper uses:

• A 3D-printed main body
• Rigid 3D-printed fingers instead of TPU fingers
• Two identical gear racks
• A central pinion gear
• Linear rods for jaw guidance

The CAD design was much simpler, and I am more confident that this version will produce reliable performance. The parallel gripper should be easier to print, easier to assemble, and easier to debug than the linkage gripper.

Next Things to Do

• Add push/pull rods in CAD to the tilting pods
• Finalize internal electronics mounting
• Evaluate costs for plates and add weights in CAD
• Bias the center of gravity toward the ideal location
• Research and add drone motors and propellers to the BOM
• Add the DYNAMIXEL XL330 mount in the arm

Devlog 4: CAD Day!

Today, I completed the body frame of the drone with the addition of a Pi Camera Module 2 mount; I also worked on improving the battery holder design, this time ensuring it can be added and removed without extreme disassembly.

Cool Things:

• I tinkered around with the “Convex Polyhedron” OnShape featurescript to make the drone’s body have a low poly aesthetic
• Added Colors!

Main Things:

• I developed the Pi Camera Mount, having the mount protrude outwards from the X-wing configuration and angle thirty degrees downward.
  • The Pi Camera was added because it provides future upgrade possibilities regarding software decisions, as well as gives the opportunity to get some footage of the drone (I also had two Pi Cams in my stash)
  • There was lots of experimentation regarding this part. I tried using lofts to make the camera mount first and then loft outwards, but frustratingly I was not able to get the geometry to work (too tight of geometry to be lofting like that!)
  • I ended up using the Convex Polyhedron Featurescript, which allowed me to select a few points and enclose an area with a low-poly extruded body, which I then hollowed out to create more overflow space for wiring
  • I finally added some material pockets, mainly for aesthetics. I will need to evaluate the printability of this part and the pockets will be the first thing to be edited if it bars printability.
• The battery holder received minor improvements regarding tolerances and the addition of the closer arm, which is mounted by two screws on each side
  • I added more space for the battery
• The spacing between the top and bottom plates received an update, increasing 10mm so that a Dynamixel XL430 can fit within the drone’s frame (for the design of the manipulator)
  • This also resulted in the arm tube clamp being changed

Next Things to Do:

• Start Arm/Manipulator Development
• Finish Tilt Pods (add m2 push rods)
• Update the BOM

Continued work on BOM (found out the project is extremely expensive; will look for ways to cut down cost). Mainly worked on designing the “Servo Pod” part – aimed to be 3D printed – which holds the servo in parallel with the tilting pod. Later, I intend implement a pull rod system to facilitate the tilting motion. The servo pod is inspired by one designs featured in a research paper I read, exhibiting an almost shaft-collar like clamping design, intending to clamp around the CF tubes that are the arms. Tomorrow, I plan to start the manipulator V1 subsystem, while implementing the push rods in CAD.

Added more research papers (will review next time), worked on v1 of electronics diagram, started BOM.

Battery Holder v1 (not final, just something that fits the battery)