Dynastinai

Dynastinai - 3lb Modular Combat Robot

full of potential, but hampered by faulty drive system

  Dynastinai is my first attempt of building a battlebot with modular weapon system. It was greatly inspired by leading 3lb battlebots Silent Spring (link to Jamison's blog) and Margin of Safety (link to Aaron's blog). The name Dynastinai comes from dynastinae, which is a scientific name for rhinoceros beetle. I changed the letter "e" to "i" to respect the naming tradition of Robojackets Battlebot (all battlebot names end with letter "i"). Also, I thank Robojackets and its members for providing me the tools I needed and doing a design review for this project.

  The modular weapon combat robot can be very effective in the competition because the driver can strategically choose which weapon to use against specific type of opponents. Depending of the types, battlebots are usually like rock, paper, scissors; one type of weapon is strong against certain weapon type, but is weak against another weapon type. For example, horizontal weapon is useful against many drum spinners with exposed vertical walls but is quite ineffective against wedges or vertical spinners with wedges at front since those deflect the horizontal hits. However, vertical spinner with two sharp prongs like the one designed above in the picture are effective against wedges or vertical spinners with wedges at front as those prongs put large pressure on to smaller area on the floor and this tend to dig under the wider wedges. This type of vertical spinner is at the same time, weak against horizontal spinners like the one designed above since those prongs are exposed to shear hits.

The goal of this project is to:

• build interchangeable weapon (one horizontal, one vertical)
• all basic structures made out of 3D printed materials
• compete in Motorama 2018

Parts I used for this build:

Aluminum parts on chassis	6061 aluminum
Aluminum parts on weapon modules	7075 aluminum
Wheel guard	HDPE
Chassis & vertical module	carbon fiber + nylon
Horizontal module	Kevlar + nylon
Horizontal weapon	AR500
Vertical weapon	S7 tool steel
Drive Motors	maxon gearbox + 1806 brushless outrunner
Drive Motor Controllers	ZTW Spider Lite 18A opto
Wheels	Fingertech robotics 2.5in*0.75in neoprene foam wheels
Weapon Motor (Horizontal)	Quanum MT 4108 700KV
Weapon Motor (Vertical)	NTM 450 series 1700KV
Weapon Motor Controllers	ZTW Spider 30A opto
Receiver	Frsky V8R4-II
Remote Control	Taranis QX7
Battery	3S 500mah 35C * 2 in series
Connectors	XT30U, MR30
5V BEC	Pololu 5V 600mA step-down regulator
12V BEC	TUZ 5V 12V 3A step-down regulator
Standoffs	M3*35 aluminum column spacer

CAD - horizontal

CAD - vertical

CAD - internal view

  This is the bottom view. The top plate does not have to be removed at all unless it gets damaged. Only one bottom plate can be removed to access batteries and the other bottom plate can be also removed to access drive motors. The dove tail design of the module interface and the aluminum posts makes difficult to rip apart the module from the chassis in all directions.

CAD - 3D printed chassis, electronics removed

  Aluminum standoffs are inserted into the 3D printed pieces. Top and bottom plates are then secured onto those standoffs with screws. This allows the top and bottom plates to clamp on the 3D printed parts and gives them compressive preload and prevents delamination. Screws and nuts hovering in the photo above are used to secure wheel guards on to the 3D printed chassis.

Half section view of horizontal module

  No screws go through the 3D printed horizontal module. The aluminum pieces are clamped from top and bottom to give 3D printed part compressive preload: prevent delamination.

Half section view of vertical module

  There are two bearings on each end of the vertical weapon. The motor is fully enclosed and is well protected inside.

Prototypes first before using expensive carbon fiber & kevlar

3D printing in action (horizontal module)

completed part

that's a lot of support structure to remove

 Vertical module is printed with its one side facing towards the bottom. This way, I could clamp the print layers sideways to prevent delamination.

CNC milling horizontal hub mount

Horizontal hub motor mount straight out of CNC mill

  Sadly, I didn't account for the tool deflection and the inner diameter where the bearing would sit was 0.002 inch smaller. The bearing would have still went in if I made a press fit but I didn't want that for the sake of easy maintenance. I used a dremel to carefully sand off the inner diameter little by little until the bearing could slide in with minimal interference. 

chassis assembly almost complete

On the vertical module, you can notice there are some nylon material extruded out and surrounding the vertical wedges. Without these, vertical wedges would be vulnerable to hits as it is just held by few M3 screws and it doesn't take a lot of force to shear those.

Where else would you find S7 tool steel laying around up for grabs?? :)

 I was fortunate enough to find the S7 tool steel in the scrap pile in Robojackets shop. They weren't gonna be used by anyone and apparently, they have been laying around there for a while. So I cut a small piece off from one of them, made a request to the machining mall to have it lathed, wire EDMed, and then heat treated to make a vertical weapon.

Assembly almost complete. Stator on the vertical module is also machined and epoxy glued.

Fabricated, with horizontal attachment

  Horizontal weapon was made out of 3/16 inch thick AR500 steel - waterjetted.

Bottom side

  Bottom cover is composed of two plates to easily access the internal components that are just needed. Also, too many screws on the horizontal weapon. I think I can reduce number of screws here for the next iteration.

With vertical attachment

Internal view

  Inside is full of spaghetti. I couldn't optimize the wire length because I had to borrow drive motors and drive ESCs from other robojacket team members since I didn't have those.

I can take off this cover and access the drive motors

  The drive motors are composed of gearbox and brushless outrunner. Because of this, rooms for the outrunners to safely spin had to be designed. Here I clamped the gearbox (not spinning part of the drive motor) in an elevated position so that the outrunners have enough vertical space to spin. Also, a thin wall was made to prevent any wires from making contact with the outrunners.

Stator inside the motor mount (top), assembled motor hub (bottom)

  The stator is is machined to fit into the hub motor. Also, it is covered with thermally conductive but electrically insulating epoxy glue to be more resistant to heat and take impacts better when it makes contact with the magnets around it. The aluminum hub is first waterjetted for the holes and general dimension and then tuned on lathe for precise cuts. Magnet ring from the original motor is cut and press fitted into the hub and is covered with same epoxy glue to keep magnets in their place.

horizontal module disassembled

vertical module disassembled

damage taken on the horizontal module during the rumble

  Even though the 3D printed part was clamped, delamination occured on the 3D printed part, and aluminum chunk was taken off from the hub motor mount. I will be strengthening this part in the future iteration.

damage on the right wheel guard

damage on the left wheel guard

 HDPE wheel guards held pretty well. It covered majority portion of the 3D printed parts and the wheels and took almost all the hits. The damages to the wheel guards don't concern me since they are designed to be consumable and are cheap and easy to make (it takes 20 seconds to waterjet one).

Dynastinai testing its horizontal weapon

  I see a lot of potential to this robot. There were many good aspect like the interchangeable weapons, efficient hub motor weapon, 3D printed chassis, 6 cell lipo battery. However, it was ultimately hampered by faulty drive system. Whenever I made sudden left to right movement or vise versa, the ESCs went into a reset mode and the robot was immobile until I put the control joystick at the center and wait for a second. In a fast paced, agility requiring combat arena, Dynastinai was immobile for significant period of time due to this mishap.

  Not having my own drive system was a problem too. I had to borrow other Robojacket member's drive ESCs and motors so I had to spend significant amount of time taking the drive in and out instead of test driving or checking out if there are any other problems.

  Also, I didn't put soft pads around the electronics inside so after hard hit from a vertical spinner, the weapon ESC got damaged and wouldn't spin up the motor anymore.

  Dynastinai never won and its win loss ratio is 0:2 from Motorama 2018. However, now I know what to fix so hopefully I can build a more formidable combat robot (Dynastinai V2) in the future.

For future iteration of this robot, improvements would likely be:

• have my own, reliable drive system (brushless -> brushed motors?)
• less free space for soft mounting drive motor, use double sided foam tape or polyurethane rubber
• more space for the wheels to rotate
• 6 cell -> 4 cell lipo ? (this way I don't have to use separate voltage regulator to power drive system)
• smaller horizontal module's hub motor design (for increasing weapon reach distance)
• horizontal, vertical weapon design change to allow bi-directional rotation
• weapon ESC firmare upgrade to allow bi-directional motor rotation
• aluminum hub or 3D printed hub with high tolerance for vertical module's motor (reduce vibration)
• pad all electronics with foam for protection

Vespa

Vespa - Stretched X racing quadcopter (November 2017)

Trade offs are unavoidable

  Vespa is a second racing quadcopter I built following Wazowski. The name Vespa comes from the scientific name of Hornet, one of the largest type of Wasps. The drone visually is sleek, relatively large for having 5 inch propellers. Hence the name Vespa. Vespa was originally designed to be my primary racing drone, taking Wazowski's place. 

The goal of this project is to build a racing drone focused on:

• aerodynamics (streamlined canopy)
• structural integrity
• high thrust to weight ratio

Parts I used for the build are:

Frame	QAV-XS Stretch-X with PC shell
Motors	Racerstar BR2306 2700KV
Flight Controller/ESC	Racerstar StarF3S 30A
Receiver	Frsky XM+
Beeper	Super Loud 5V active alarm beeper 9*5.5mm
Camera	RunCam Micro Swift 600TVL
Video Transmitter	EWRF-e7082TM
Battery	Infinity 4cell 1500mah 70C
Propeller	DAL T5045 V2
Connector	XT60U
Capacitor	Panasonic 1000uF 25V electrolytic capacitor radial

Mid assembly

  I am using an all-in-one flight controller that includes 4 ESCs, OSD, 5V BEC, etc. It is amazing how drone technology advanced within the last few years. This allows very clean and low profile build and all the components should fit into the canopy's limited space.

Soldering complete

  Vespa is based on stretched-X frame, which is a variation of the traditional X shaped frame. The frame is stretched in length wise so that there is more space between the front propellers and the rear propellers. The idea is that as the quadcopter advances forward, the front propellers create turbulence and the rear propellers end up taking those disturbed air. Giving more distance between the front and back props will reduce this effect thus making cleaner flight characteristics.

Camera mount was designed using solidworks and then 3D printed with PLA plastic

Close view of the internal electronics

   Capacitor is added next to the battery connector. It is common way to reduce electrical noise. More effective way of doing so would be adding capacitors as close as possible to the motor (source of electrical noise) but that would add more weight and since there is limited space inside the canopy, I used only one.

  Also, the video transmitter I am using has smart audio feature. That is, I can adjust the video transmitter's power output and frequency through OSD and radio transmitter without doing so manually. This is a very nice feature to have especially when you are flying with multiple drone racers and have to switch VTX setup frequently.

Build complete, Side view

   Motor wires are covered with electrical tape to protect the wires from external objects such as tree branches.

Front view

Rear view

Georgia Tech Unmanned Flying Club

  Vespa improved on every aspect I found flaws from the previous racing quad I built. The frame is a lot stronger (never broke an arm), canopy is aerodynamic, motor is more powerful. However, those improvements brought problems that weren't present in the previous quad.
  Vespa weighs 305 grams and that is 74 grams heavier than Wazowski. 74 grams may not seem a lot but it affects the flight characteristics significantly. Because of that, Vespa is less agile in flight, which is not a problem when going fast in a straight line, but it is bad when flying in race course that has frequent turns.

  Also, the PC shell canopy wasn't as tough as it should be. The front part of the shell broke off easily from mild crash and long crack lines were starting to form (that's what I get for buying the cheap chinese frame off from ebay for 1/4 of the original price).

Canopy damaged, cracks forming...

  The video transmitter I used for this build had good features (smart audio, light weight), but the video it was transmitting wasn't clean enough for flying comfortablely. Edit: turns out I had the antenna too close to the main power wires. EMI was distorting the video quality and after re-positioning the vtx, the video quality was much cleaner.

  I didn't like the problems above (especially the 74 grams heavier aspect) so Wazowski still remains to be my primary racing drone. Vespa is still a great, reliable quad so I'll be using this primarily for free style flying or as a backup to Wazowski for racing.

  From this build, I learned the importance of minimizing weight and how it can affect the flight characteristics. Every gram counts in the world of quadcopters!

For improvements on future build:

• sub 200 gram racing drone (not including battery)

Radii V2

Radii V2 - Second iteration of Radii, the 3lb drum spinner battlebot (October 2017)

Carbon fiber madness. This placed 7th out of 34 competitors with win loss ratio of 4:2

 Radii V2 follows the design of Radii which competed in Motorama 2017. This is the improved version that takes care of the weakness of the previous Radii. Many thanks to Eunchang for being a good team member for this project, working together from CAD design to fabrication. Also, huge thanks to Robojackets and its members for funding this project and providing design reviews.

Goal of this project was to:

• fix all problems that previous Radii had

• utilize 3D printed armor

• compete in SparkFun AVC 2017 competition

Parts I used for this build are:

All aluminum parts	6061 aluminum
Top & Bottom plates	2mm thick carbon fiber
Drive Motors	9.7:1 Metal Gearmotor 25Dx48L
Drive Motor Controllers	Scorpion Mini
Weapon Motor	Turnigy 2836 Brushless 450-Size Heli Motor 3200kv
Weapon Motor Controllers	Racerstar MS Series 35A opto
Receiver	Hobby King 2.4Ghz Receiver 6Ch V2
Remote Control	Hobby King 2.4Ghz 6Ch Tx and Rx V2
Battery	Turnigy nano-tech 850mAh 3S 45~90C Lipo Pack

CAD overall

Internal CAD view

  All the heavy weighing electronics (weapon motor, battery) are now shifted behind the wheel so that the wheels are closer to center gravity. Mistake from the previous Radii was that the CG was too far away from the wheels and it was extremely difficult to drive. This design partially solved the driving issue. 

Section view of the weapon

  The drum is smaller in length but has slightly thicker wall inside. This is to maximize moment of inertia and minimize the deformation of aluminum that holds the screws upon impact. However, the weapon is lighter in the new version. The weight saved from down sizing the weapon is used to strengthen the side armors and adding the front plate to hold the side plates from shearing.

Playing around with Inventor FEA stress analysis

  Did some FEA stress analysis for reference. The part above was designed to be used as armor (3D printed) and would be one of the primary places to take hits from the opponents. Regular FEA analysis would not be the best method since the 3D printed parts are not isotropic. 3D printed parts are generally strong along the filament (tensile strength) but weak when the force is trying to pull apart the layers (delamination).

Internals

  Since the weapon motor got shifted behind, the belt has to run over one of the drive motors and some wires. This looks sketchy but it works. Due to long belt, I suspected the loss of belt tension from stretch would be a problem. I designed the weapon motor mount so that I can shift the motor more towards the back to adjust belt tension.

Fabricated & comparison with the first Radii

  Side, front and back aluminum plates are cut using waterjet and then pocketed with CNC mill for fitting in motors and extra weight savings. Top and bottom carbon fiber plates are waterjetted too. The aluminum part of the drum and the shafts are machined using lathe and then CNC milled for screw holes. External side plates that covers the wheels and aluminum side plates are 3D printed using invention studio's Markforged 3D printer with nylon and carbon fiber filaments. Infill density was set to 50% and pattern was set to triangular shapes (honeycomb shapes for maximum impact resistance. 6 regular triangles = 1 regular hexagon = honeycomb).

Damage taken from project darkness pic 1

  Most of the damage taken from 2017 Sparkfun AVC was from project darkness, a 3lb vertical spinner with front wedge.

Damage taken from project darkness pic 2

  Steel screws have significant amount of chunks taken out from hits. Too many of this would ultimately cause the weapon to be not balanced and this will increase vibration when spinning.

  Although Radii V2 improved a lot of faults from the old Radii, it wasn't a great success as it brought new faults from changes in design.

  First of all, the drive system wasn't reliable. The wheels were closer to the center of gravity so it drove well. However, when the robot got hit, went airborne, and then landed on the wheels, the shock which the drive motors took from the weight of the robot was greater than the previous Radii and easily broke the gearbox. Hard mounting the drive motors on the rigid aluminum plate with screws and using not-so-soft banebots wheels were also the cause since they would easily transfer load to the gearbox and the shaft whenever those take impacts.

  The weapon was an issue as well. The weapon motor wouldn't spin up to its maximum speed and I think the belt tension was the issue. The belt I received was too small in length so I had to constantly stretch it before the competition. Apparently, that wasn't enough and I believe the belt put too much load on the motor and prevented it from spinning freely.

  Using carbon fiber plate as top and bottom plate was a mistake. Carbon fiber is definitely light weight and rigid enough to hold internal electronics in place but it was too brittle. It shattered immediately after taking hits from vertical spinner. 

  Furthermore, the experience from Motorama 2017 and SparkFun AVC 2017 made me realize the limitation of the drum spinner design. It can take and give out vertical hits well but it is weak against horizontal hits. Unless designed to accommodate different types of opponents (ex: vertical spinner, horizontal spinner, wedges, etc), it would be difficult to reach high placing ranks in the competition. Thus, I end the iteration of Radii here.

Improvements and changes for the future design would be:

• interchangeable weapon system

• hub motor design. no more belts

• vertical weapon made out of steel. no more steel bolts as weapon teeth

• chassis made out of 3D printed materials (carbon fiber/ kevlar/ fiberglass combined with nylon/ onyx)

• 6 cell lipo battery as a main power source

• aluminum or plastic for top and bottom plates

• brushless drive system

• soft foam wheels instead of rigid banebots wheels

• soft mount drive motors instead of hard mounting them with screws

• use of metric screws. no more imperial bs

On a side note...

  I was taking the course at Georgia Tech called ME 3210 Design, Materials, and Manufacture at the same semester and the extra credit was to make a video about manufacturing stuffs. My group decided to just make a video about the manufacturing process of Radii V2 which was already built. Luckily, I had recorded several manufacturing processes (ex: lathe, waterjet, 3D printing, etc) so I sent those to my group members and they helped making an informative video clip about the build summary (Thanks Jeremy, Eddie, Dallas). 

extra credit video for ME 3210