Author Archives: Patrick

High-altitude balloon for the solar eclipse

via Pololu Blog

UNLV mascot Hey Reb! high above Idaho (and a commercial jet) during the partial eclipse.

This summer, Jon and I participated in NASA’s nationwide Eclipse Ballooning Project with the University of Nevada. Specifically, we were members of the University of Nevada, Las Vegas (UNLV) section of Nevada’s team, which also had a section from the University of Nevada, Reno (UNR). Our goal was to make a payload to collect interesting video footage and scientific data, then fly that payload on a high-altitude balloon that would ascend to around 100,000 ft in the totality zone during the 2017 solar eclipse on August 21.

Before the eclipse, the UNLV team did a practice launch near Nipton, California. For this launch the payload included a humidity sensor, an AltIMU-10 v5 Gyro, Accelerometer, Compass, and Altimeter, and an A-Star 32U4 Prime LV with a microSD card for datalogging. All these electronics were contained in a plastic project box with a mobile phone charger as a power source. A heating pad powered through the A-Star Prime’s 5 V regulator kept the electronics in the project box at an operable temperature range, and the project box was insulated in a styrofoam box to prevent heat from escaping during the flight. The payload also included an Automatic Packet Reporting System (APRS), a Spot Tracker, and two GoPro cameras. Telemetry from the APRS allowed us to track the position, altitude, heading, speed, and surrounding temperature of the payload during the flight. Our APRS reported data roughly every three minutes, and we programmed our A-Star to collect sensor data ten times per second.

Electronics from the high-altitude balloon payload.

Project box containing electronics and power supply.

We considered this practice launch a success because our team, which consisted mostly of people who had never launched a high-altitude balloon before, prepared and launched the balloon/payload and recovered it during an unpredicted thunderstorm. Despite the storm, the A-Star and the other electronics stayed warm and collected data through the entire flight without incident. Even though the GoPro cameras ran out of memory during the descent, they obtained some awesome images of desert mountain landscape, Lake Mohave, and the Colorado River from over 90,000 ft.

Lake Mohave and the Colorado River from above during practice launch.

The practice launch also offered an opportunity to highlight any weaknesses or flaws in our design, which we could improve or fix before the actual eclipse mission. For example, we learned that we did not properly take into account how much twisting and rotating the entire payload would undergo. The twisting and rotating was apparently strong enough to completely unscrew our APRS dipole antenna from its holder; it separated during the descent when the balloon was still over 50,000 ft above ground! Thankfully, our Spot Tracker was able to acquire a satellite signal and relay its position so we could recover our payload. For the eclipse launch, we took special care to in securing APRS antenna before liftoff and added another Spot Tracker backup.

Additionally, after reviewing the data collected from the practice launch, we realized that during key events (e.g. liftoff, burst, landing) the payload both accelerated harder and rotated faster than we could measure with the default configuration settings of our accelerometer/gyroscope. Our code uses the Pololu LSM6 library, which by default configures the sensors to use the smallest full scale range (+/-2 g for acceleration and (+/- 240 dps for rate of rotation). We wanted to better characterize the magnitude of the acceleration and rate of rotation during those extremes, so, for the eclipse, we increased the full scale range to the maximum settings ((+/-16 g for acceleration and +/- 2000 dps for rate of rotation).

Exterior sensors.

Another improvement for the eclipse launch was the addition of a second group of sensors to the payload to collect data from outside the insulated styrofoam box. This group of external sensors includes a photocell, a luminosity/lux/light sensor, an infrared/visible light detector, and an ultraviolet light detector, another humidity sensor, and a second AltIMU-10 v5 Gyro, Accelerometer, Compass, and Altimeter. These external sensors also required a heating pad, but adding another pad would push the A-Star Prime‚Äôs 5V regulator to its limit. The phone charger supplies power through a USB-micro connection, so we decided to access the 5V power from the charger’s USB power bus. Thankfully, our mobile phone charger could handle the extra current we wanted to supply. Ideally, we should have used something like Pololu’s USB-micro breakout board. We were short on time though, so we used what we had on hand, a FPF1320 Power Multiplexer Carrier with USB Micro-B Connector (the carrier makes VBUS accessible through 0.1"-spaced headers). In addition to the electronics added to the A-Star system, we also added a microbiology experiment for NASA Ames and a Geiger counter with its own data logging equipment.

Preparing main payload before eclipse launch.

For the eclipse, the UNLV and UNR teams met in Idaho and launched our balloons and payloads north of Hamer, Idaho so that they would be near maximum altitude during the totality. One of the objectives for the UNR section of the team was to live-stream video footage of the balloon flight during the eclipse. The radio equipment they used for this required the receiver to automatically maintain a line-of-sight with the balloon. They accomplished this using a Micro Maestro 6-Channel USB Servo Controller to operate a servo pan-tilt assembly.

Receiver for UNR balloon live-stream.

The eclipse launches were a great success! UNR launched and recovered two balloons and payloads. One balloon landed on a farm, and the second was retrieved from a residential backyard. Although they were not able to successfully live-stream their flight, they were able to collect cool data and GoPro video. The UNLV balloon and payload that Jon and I worked on ascended over 100,000 ft, and it was between 85,000 ft and 90,000 ft during the totality. We recovered the balloon and payload about one mile east of Interstate 15 south of Hamer, Idaho. The next image below shows the flight path of the UNLV balloon during the eclipse. You can more closely examine the path by opening this file (2k kmz) in a desktop Google Earth program.

Payload recovery after the eclipse.

The A-Star successfully logged data from all the sensors during the entire flight. To share and analyze the data, I used a MATLAB script to create and save graphs of the data. I quickly learned that many of the sensor signals also had a lot of interference that distorted the true patterns during the eclipse. In MATLAB I had to create functions that removed this electrical noise from the data sets. The graphs below show the readings throughout the flight from the humidity sensor inside the payload and the humidity sensor exposed to the environment. The graph on the left demonstrates how much interference there was in the raw data, and the graph on the right demonstrates the effect of removing that noise on the clarity of the graph.

Humidity vs time plot without electrical noise removed.

Humidity vs time plot with electrical noise removed.

Below are some more graphs showing some of the interesting data we collected. Keep in mind that liftoff was around 19 minutes after the system was powered on. The totality occurred about 84 minutes after the system was powered on and lasted for about 2 minutes and 20 seconds.

The UNLV balloon had two GoPro cameras during the eclipse flight: one facing down and another facing up and out. You can view the full video from the downward-facing camera here, and the video from the upward-facing camera here. Unfortunately, our cameras stopped filming about 40 minutes into the flight, before the totality. We are unsure why they stopped filming, but some of the team suspects the cameras overheated since at those high altitudes there is less ambient air around to exchange heat with the cameras, which were generating large amounts of heat from recording video at a high resolution. However, we ran the cameras at the same resolution during our practice launch without this problem, so I am not entirely confident that overheating caused the camera shutdown. The camera shutdowns also happen to occur around the same time as the two mysterious radiation spikes measured by the Geiger counter. Perhaps these radiation spikes adversely affected the cameras and caused them to shutdown.

While we didn’t get any pictures of the eclipse during totality from the balloon, we did get some nice pictures from the ground:

Photo of the 2017 solar eclipse taken by Dr. Jeffrey Lacombe (UNR) with a telescope camera.

High-Altitude Balloon Team Nevada.

The Rebel WIP from UNLV

via Pololu Blog

It has been over a year since my last/first blog post about my mini sumo robot. Since then, I have been busy studying to earn a mechanical engineering degree at the University of Nevada, Las Vegas, working at Pololu, and, of course, building cool robots. Recently, I have been working as part of a UNLV team, Rebel Robotics, to build a robot to compete in an intercollegiate competition organized by the American Society of Mechanical Engineers (ASME) called the Student Design Competition. My team’s robot, the Rebel WIP (work in progress), competed against teams from colleges across the western half of the United Sates in the Student Design Competition at an engineering festival hosted by UNLV and ASME called E-Fest West.

The competition

The challenge presented in the Student Design Competition is different every year. This year, the challenge was to build a robot to compete in a series of five events called a Robot Pentathlon. The events included a sprint, a stair climb, a lift, a throw, and a hit. The robots had to start each event in a 50 cm sizing cube and could only use energy from a rechargeable battery. The robots would compete and be placed first-to-last in each event based on their performance. The sum of the places determined the winners of the competition (lowest sum won).

Chassis and drive train for sprint and stair climb

The chassis is built from ServoCity and Tetrix C-channel sections bolted together with corner-brackets. These brackets hold the frame together, but to keep the frame stiff we machined a steel plate that mounts to the underside of the robot and supports other components like the air compressor and battery.

The Rebel WIP’s frame, side view.

For the sprint event, robots had race against time to touch a wall ten meters away and get back to their starting location. In the stair climb, the robots had to ascend and descend a set of three steps as quickly as possible. The steps could each be up to 15 cm high. To make the Rebel WIP competitive in both the sprint and the stair climb, my team designed a differential drive setup to control four tank treads that are each on independently articulated tank arms.

The four tank treads are controlled by two motors at the back of the robot. They are the motors that are in line with the C-channel chassis, not the motors above the chassis. By directing these motors to rotate in the same or opposite direction, the robot will move backwards, forwards, or turn. To make this work, the output shafts of two motors are coupled to 6 mm shafts which are supported by bearings in the C-channels. Each of these shafts at the back of the robot are fixed to a timing belt pulley on the inside of the C-channel and a hub which rotates the rear tank tread. The timing belts on each side run to pulleys at the front of the robot that are also fixed to rotate with 6 mm shafts. These shafts at the front of the robot are fixed to hubs that rotate with the front tank treads.

Drive and tank arm motors located at the back of the Rebel WIP.

The tank arm articulation system is closely linked with the drive system. The tank arms support the tank treads and can rotate about the driving pulleys. They are able to do this independent of the movement of the treads. The aluminum tank arm parts were machined from a larger plate using a CNC (computer numerical control) mill at the UNLV machine shop.

Actuating the tank arms requires four more motors. These four motors sit above the C-channel frame at the front and back of the robot. Each motor drives a small gear that connects to a larger gear fixed to an aluminum arm. The large gear moves together with the aluminum tank arm. A second milled aluminum part sits in a slot in the tank arm. This part supports the idler pulley for the tank tread. Positioning this support in the slot tensions the tank tread. The aluminum tank arms are supported by, but not fixed to, the 6 mm shafts that drive the tank treads. The gears that are fixed to the tank arms and control tank arm rotation are concentric with the 6 mm shafts and drive pulleys. When one of the motors above the frame rotates, the arm it connects to will rotate around the shaft that supports it, but the parts fixed to the shaft (the tank tread and pulleys) do not rotate with the arm.

Forward tank arm in angled position.

During the competition, the articulation of the tank arms allowed the Rebel WIP to approach a step and prop itself up on the raised surface. Rotating the rear treads behind the frame kept the robot supported while it climbed the stairs so that it would not tip over backwards. There were a few flaws in the drive/climbing system that could not be corrected in time for the competition. The motors selected were just barely powerful enough to complete the tasks under ideal circumstances, but were not powerful enough to complete the tasks reliably. We should have either used more motors, or more powerful motors. In fact, the robot had trouble turning because the drive train motors did not have enough power to pivot the robot from a stopped position. Fortunately, we did not need to turn in any of the events. Another problem was that a few components on the robot were too low on the robot and could obstruct the robot from climbing. It also would have been better if we programed the tank arm motors to hold position under load when not being controlled using encoders, but we ran out of time to do this before the competition. The most significant problem with the chassis was that the steel plate caused the frame to bow. The plate had been machined by hand, so some of the designed dimensions were not produced with adequate precision. Despite these shortcomings, the Rebel WIP earned second place in the sprint event, and tied for third in the stair climb event.

The lift

In the lift event, robots had to lift a weight as heavy as possible as high as possible in under one minute. The score used to rank the teams was calculated by multiplying the weight lifted in kilograms and the distance lifted in centimeters. To accomplish this task, the Rebel WIP uses two scissor lift mechanisms to raise a platform that holds a weight. The platform is made of a few sections of C-channel that go around the air cannon, which I will discuss later.

Each scissor lift mechanism is supported at two points at the base of the robot. At one point the connecting scissor lift section can rotate on a bushing but otherwise cannot move. The second point also allows the connecting section to rotate on a bushing, but it can also translate linearly towards or away from the first connection point. This is accomplished using a rack and pinion system connected to a motor geared for high torque. The rack and pinion is setup so that when the motor rotates, a small C-channel section will move linearly on along the gear rack. The motor and the second connection point for the scissor lift are fixed to the moving C-channel section. The same connections are present at the top of the lift, just without the motor.

When the pivot points are moved towards each other, the scissor lift sections, which are connected to each other at their ends and midpoints, are forced to rotate so the lengths of the sections create a greater angle relative to the horizontal. The total length of each scissor lift section is fixed, so as the horizontal component of the space each section covers from end to end shrinks, the vertical space each cross section covers must expand, increasing the total height of the structure.

The two scissor lift mechanisms on the Rebel WIP need to move together to keep the top platform level. We could have done this by using encoders to keep the motors running at the same speed, but it was simpler with our time constraint to make a control option where the two lift motors could be controlled independently. If one motor got ahead of the other, we could manually stop it with the remote control and give the other motor time to catch up.

The Rebel WIP completing the lift at competition.

Building and testing the scissor lift was a significant learning experience for me and my team. Our goal was to lift a relatively light weight and maximize height. This strategy required the robot to have as many scissor lift sections as possible, but each section added to the lift adds weight that the motor has to lift. More importantly, the more scissor lift sections add to the starting inertia that the motor must overcome to get the lift moving, especially in our ideal setup where the sections start nearly parallel to the horizontal. For these reasons we expected motor power to be the greatest limiting factor to improving the lift. Instead, we discovered that the it was our structure around the motor and lift that could not handle the heavy loads caused by the scissor lift actions. Parts of our structure would deform under load and space would form between external gears that connected the motor to the rack and pinion. This would cause the gears to start skipping teeth, which would stop the lift right after it only started to raise the weight. To make the lift work for competition we had to dramatically reduce the number of sections used in the scissor lifts since we did not have time or budget to build a new stronger frame structure.

For the competition, the Rebel WIP lifted a relatively light weight of about 1 kilogram about 1.7 meters. Although our lifting capacity was far less than our goal, we still lifted our weight higher than any of the other robots.

The throw

In the throw event, robots had to throw a preloaded tennis ball as far as possible. The loading and throw had to be completed in one minute. The Rebel WIP completed this event by using compressed air to launch the tennis ball through a PVC tube. An air cannon is a pretty simple concept, but a lot of thought went into optimizing each element of the design to maximize the throw distance. A regulation tennis ball is 2.7 inches in diameter, so we selected 2.5 inch PVC for the cannon to have a tight fit. Tennis balls have a small amount of compression, but our fit was tight enough that a ramming rod had to be used to load the tennis ball into the cannon. Since all the power for the robot had to come from a rechargeable battery our robot needed to use a compressor to build air pressure. In the competition we would launch the tennis ball around 100 psi. This high pressure can be dangerous to work with if you’re not careful, and you definitely do not want to accidentally exceed the safe operating pressure you establish. Rebel WIP has multiple safety components including a pressure gauge, manual pressure release valve, a pressure switch, an air pressure regulator, and a relief valve. Our air reservoir was located right next to the one inch electric valve that would open to launch the tennis ball, and our valve was located just before the cannon. Minimizing the space between the reservoir, valve, and cannon were important design considerations because we wanted the compressed air to be moving with as much force as possible when it contacted the tennis ball.

Air compressor and safety components.

The Rebel WIP’s air cannon.

At the competition, the tennis balls were covered in red chalk so that the ground would be marked where the balls first land. So after the loading and firing the first tennis ball, the PVC white color on the inside of the cannon was replaced by a cool coat of red chalk. The Rebel WIP threw the tennis ball over 145 feet in the competition. We earned second place in this event.

The hit

In the hit event, robots had to hit a golf ball that started on the ground as far as possible in a straight line. The score used to rank the teams was the distance the golf ball traveled before hitting the ground in the specified direction minus the distance the golf ball traveled perpendicular to the specified direction. To complete this event, the Rebel WIP has a powerful motor at the front of the robot without any gearing to reduce the speed of the motor. It has a free run speed of 13,180 rpm. Attached to the motor output shaft was a piece of metal to hit the golf ball. The strategy was to prop up the front of the robot using the front tank arms, turn on the front motor, allow it enough time to reach to full speed, then move the tank arms to come down and hit the golf ball.

The Rebel WIP, front view.

My team and I thought of the hit as the least interesting and most difficult of the challenges. It became the lowest priority challenge for our team and was the last component to be tested. In fact it was not tested until around 12:30 am the day of the competition. Unsurprisingly, the mechanism for completing the hit did not work during the competition and our team tied for last place in that event. Fortunately, our performance in other events compensated for our weak performance in this event.


The Rebel WIP is powered by a 12 V rechargeable battery. We selected the Arduino Mega as the microcontroller for the robot because we could use the familiar Arduino IDE to program the robot, and the Mega has 70 input-output lines available to control the many components on the Rebel WIP. When fully charged, the battery voltage is usually around 14 V, so a Pololu 9V step-up/step-down voltage regulator is used to reduce the voltage supplied to the Arduino (the recommended range for powering Arduino microcontrollers is between 7 V and 12 V). The robot is operated remotely using a PlayStation 2 controller. The controller’s wireless receiver is connected to digital inputs on the Arduino. The receiver is powered through the Arduino’s on-board 5 V regulator. Whenever the Arduino is powered (even if the rest of the robot is not powered), the receiver is also powered, so I can send signals to the robot with the remote control. This was useful for troubleshooting problems when I began programming.

The motors for the drive train, tank arms, and lift are controlled by Pololu second-generation high-power motor drivers. The motors for the drive train and tank arms are each controlled by a 24v13 motor driver. The two lift motors are each controlled by a 18v25 motor driver. The motor used for the hit event is controlled by an old Talon SR motor controller I had previously used for an FRC robot when I was in high school. I used the digital inputs on two high-power MOSFET slide switches to activate the air cannon valve and to turn the air compressor on or off. The Rebel WIP also has a Pololu 5V step-down voltage regulator that could have been used to power many small components like encoders in case the Arduino’s on board regulator could not supply enough power. Unfortunately, I did not have time to add these components before the competition.

The battery is wired in series with a physical switch that powered the rest of the components (regulators, MOSFET switches, and motor drivers) in parallel. This means that when the switch is turned on, all of these components can draw power from the battery without that power having to go through any other circuit elements except for the battery’s built in fuse.

The electronics shelves on the Rebel WIP.

All of the electronics for Rebel WIP are located on two shelves at the back of the robot. When I build robots under a size constraint, I like stacking my electronics like this because it allows me to keep all the electronics consolidated in one area, but it is not as demanding on horizontal space as making one board that carries all the electronics. It also allows me to keep all the jumper wire connections vertical. This way when the robot moves around, gravity is working to help keep things connected instead of actively working to break connections.


The Rebel WIP was a time consuming project, especially as the competition date drew closer and closer. My team’s hard work paid off though. We won third place out of thirteen university teams in the competition!

Any robotics project is a learning opportunity, and I certainly learned a lot from working on the Rebel WIP that will help me improve my other robotics projects. Although my next big task at hand is preparing for finals in May, I look forward to sharing my next cool robot project.

Patrick’s mini sumo robot: Covert Ops

via Pololu Blog

Patrick's mini sumo robot: Covert Ops

Hi, my name is Patrick. I am an engineering intern at Pololu and am studying at the University of Nevada, Las Vegas to earn a mechanical engineering degree. I decided to build a custom robot to compete in the recent LVBots mini-sumo competition here at Pololu. It was my first competition at LVBots. I started out by watching a compilation video of the previous sumo competition at LVBots since I had never competed in a mini-sumo competition before. My goal was to create a robot that could out maneuver other robots and had as few vulnerabilities as possible. To achieve this goal, I decided to build a robot that would be high speed and able to push opponents from both sides of the robot with a lot of force. The result of my efforts is the robot I call Covert Ops.

The frame

Before doing any construction, I created a detailed CAD model of my whole robot with SolidWorks. The inside frame of the robot is composed of multiple 3D-printed parts that snap together like jigsaw puzzle pieces. In addition to being snapped together, each part is bolted to the adjacent frame members in order to make the skeleton of the robot as sturdy as possible. I used a 3D Systems Ekocycle 3D-printer to make the parts. This was the first time I used this printer to create parts that require very precise dimensions. I got pretty excited when I put the frame sections together for the first time and found out that the clearances I set at only a few tenths of a millimeter in some places were perfect for snapping the frame together like I wanted!

Covert Ops has 3D printed blades facing forwards and backwards on the robot so it can push from two directions. I used black material for the blades and side covers since many proximity sensors (such as IR-based sensors like these) struggle to detect darker colors. I was nervous about using plastic blades, but the blades worked out better than I expected and easily held their own against the metal blades of the other robots.

To hide the wiring and electronics and keep my robot looking cool, I 3D-printed a thin black cover that just slips over the structure that supports the electronics. Some images and logos I glued to the cover make Covert Ops look like race car.

Drive train

Covert Ops is driven by four gearmotors. The motors I selected are 50:1 high-power micro-metal gearmotors with carbon brushes. Each motor is directly connected to a green BaneBot wheel. The 50:1 gear ratio makes the robot pretty fast, and since there are four motors, Covert Ops can still push with a lot of force. Each motor is controlled by a DRV8838 Single Brushed DC Motor Driver Carrier. I chose these motor drivers because they can supply adequate power to the micro-metal gearmotors and they take up very little space on the robot.

Sensors and electronics

Covert Ops uses four QTR-1RC reflectance sensors for detecting the border of the sumo ring. There is one reflectance sensor on each corner of the robot. To detect opposing robots there are four Sharp GP2Y0D810Z0F digital distance sensors. One distance sensor faces forward, one faces backwards, and the final two face out from the sides of the robot. I chose the Sharp GP2Y0D810Z0F sensor, which has a range of 10 cm, because I only wanted my robot to charge towards an opponent once they were close. I also have a Pololu MinIMU-9 on my robot. This component has a 3-axis gyro, accelerometer, and compass. Unfortunately though, I ran out of time to create code that would utilize the MinIMU.

Covert Ops is powered by four AA batteries. They are located on the underside of the main body, which keeps the robot’s center of gravity low. The Arduino-compatible A-Star 32U4 Mini LV functions as the controller for my robot. The batteries supply 6V power, so I included a 5V step-up/step-down voltage regulator to power my sensors though I could have used the A-Star’s built-in regulator instead. For power distribution I removed two power distribution boards from a 400 point breadboard and mounted them to my robot. One is used for 6V power distribution and the other for 5V power distribution.


I created the code for my robot using the Arduino programming environment. Covert Ops is programmed so that after the mandatory five second wait, it rushes forward at full speed to the middle of the sumo ring. From there it spins and uses its proximity sensors to scan for the other robot. Since the proximity sensors only detect robots very close to them, I programed Covert Ops so it spins and scans for a short, random amount of time before making a quick move to a new location. This way I hoped to avoid another robot detecting and charging at my robot from far away.

Accomplishments and improvements

Even though Covert Ops did not place, I am still pretty impressed with it. It was one of the quickest moving robots in the competition and never lost a pushing match when it directly engaged another robot face to face. Covert Ops had a unique design and did not suffer from any mechanical failures during the competition.

There is still a lot of room for improvement for my robot. One issue during the competition was that since my proximity sensors had to be really close to detect an opponent, they would often fail to detect the opponent. Another issue was that since my robot used four wheels, it had to have clearance underneath it. This gave two wheel robots with low blades a major advantage over Covert Ops, since they would be able to get underneath my frame. My programming strategy also needs some work. Because my robot moved randomly, it spent a lot of time around the border of the ring, which is a very dangerous place to be in a sumo competition. I also have yet to figure out the programming that would allow me to use the MinIMU’s gyro sensor and accelerometer. I will definitely be making some improvements before the next sumo competition.