We spent the offseason making our electrical robust. Here are a few highlights

## Priorities

The electrical system is designed with the following philosophy in mind:

1. Reliability

   95% + drive base uptime

2. Robustness

   Can handle intense competition scenarios

3. Modularity

   Ability to add complex electrical systems, both prototypes and production-grade

4. Speed in assembly

   Faster wiring speeds enable more time for Drive and Programming

5. Performance

   Circuits optimized for efficiency for high current draw applications

## Drivebase

* Machined Aluminium offers ballast and ease of wiring

* Multiple passthrough buses enable compact electrical

* Dual Falcon 500 with CTRE CANcoder driving SDS MK4i modules

* CANivore enables multiple independent CAN busses

## Subsystems

* Neo 550 with Spark MAX controller

* Wago 221 inline connectors for main power splice

  * Enables fast assembly and reliable/optimized connection

* Heavy-duty wrap on major connections to protect wires

  * High tolerance against constant friction

* Inline heat shrink splice connectors for CAN and secondary power/sensors

  * Our testing, enables fast assembly, high tolerance to stress, and reliability

## Prototyping and Modularity

* Wago connectors enable fast prototyping and rapid iteration

* Independent and Integrated drive base electrical system

  * Enables system redundancy through competition conditions

## Pneumatics

As we are running multiple pneumatic systems on this robot, Pneumatics optimization is a priority. We designed a novel system to address this engineering challenge

* Integrated pneumatics assembly

  * Under cone ramp, which is designed to open for access

  * compressor solenoids and control are close to each other enabling reliability

* Backpressure storage

  * Able to store greater quantities of air, and maintain a constant 60 PSI pressure

* Parallelized solenoid systems

  * Optimized for airflow into arm pistons to enable fast deployment.

The Engineering Behind Temaki

If you're reading this, chances are you've experienced the convenience of online shopping. You turn to Amazon for everyday essentials, Nordstrom for your footwear fix, and often find yourself intrigued by various items showcased on Instagram.

In today's world, e-commerce operates within a complex web of algorithms, advertisements, and referral links. Amidst this noise, your pursuit to find products that genuinely meet your needs becomes muddled by an overwhelming barrage of ads. Conversely, businesses struggle to penetrate this clutter and offer truly personalized recommendations.

However, this wasn't always the case.

Rewind to the 19th century, and picture this: seeking a new pair of shoes meant a visit to your local cobbler. Here, a conversation ensued about your specific needs - "What purpose do these shoes serve?" or "What's your budget?" Based on this interaction, the cobbler would suggest shoes that precisely cater to your requirements.

Enter Sears, a game-changer with its mail-order catalog. For the first time, consumers could peruse and order products from afar. While this innovation offered unprecedented convenience, it also stripped away the personal touch from the shopping experience.

As time progressed, catalogs evolved into online stores, mail orders transformed into expedited shipping, and physical currency gave way to digital transactions. Yet, the core principle remained unchanged. To enable customers to discover products, companies lean on dispassionate algorithms which digest myriad data points to produce recommendations, all in the quest to maximize sales.

But what if we could revolutionize this dynamic?

What if every customer has access to their virtual clerk, offering bespoke recommendations tailored to their needs?

This is the vision we're bringing to life at Mobius. We're developing an AI-driven assistant that enables customers to shop interactively. But this is merely the first step. Our vision is to redefine commerce, not as a journey through extensive product libraries and random suggestions, but as a conversational experience, allowing customers to seamlessly discover products that truly resonate with their needs.

**The future of commerce is conversational. And at Mobius, we're at the forefront of defining this new paradigm.**

The New Commerce Paradigm

Vision for the New Non- Profit Economy

Observations about current non-profit economy: A non-profit, from a legal sense, is a tax exemption from IRS. But the social expectation of a non-profit is to serve the community in some way. From that, the first thing that comes to mind is the Bill & Melinda Gates Foundation, the most influential philanthropic organization. However, a singular, highly valuable entity controls worldwide distribution of services and funds. To be clear, the Gates Foundation has done incredible impact around the world across many fields. However, to create true change, a community needs to take on the responsibility of making a positive difference, even if they are to start within their own local spheres.

Recently, there have been a mass influx of non-profit organizations started by high-school students, with the motivation of boosting their college application. These organizations and the talent behind them have incredible potential. But, they are severely limited in their impact due to the fact that they are driven by a timeline and a specific motive – that being college applications. What if these organizations banded together and created a network with the larger goal of enabling true change?

Benefit of creating a network:

All non-profits need money and volunteer support. By banding together, we can create a pool of resources

Since there is an alliance, it is not one massive and bureaucratic organization focusing on serving every need, instead, a network of flexible, agile organizations that can tailor themselves to the niche they service, yet, draw from a common pool of resources.

Such a network would involve a host of teen-run non-profits, each of which offers a different kind of service, that is envisioned by that particular leader. For example, Iridium Tutoring specializes in offering tutoring services. Mehek Box focuses on providing technology for the disabled to learn and create music. ArtForYou brings art and STEM together with themed performances. GRISM is focused on Girls in Math and Science. These are all teen-run non-profits that started in the past few years. If they banded together they would all contribute in their own areas of expertise and yet support each other across several niche spaces.

A vision like this is ambitious, but if implemented, it will be a game changer benefiting the volunteers involved, the founders, and the donors who support these causes, but most importantly the countless number of people who will be better serviced.

The Phoenix Project aims to build toward this vision. With the current peer-to-peer robotics program model, we are laying the groundwork and building the necessary infrastructure. Much like Amazon started off selling just books, and expanded to every possible consumer good, the Phoenix Project establishes its foundations through the peer-to-peer robotics network with the vision to build the new non-profit economy.

A Vision for the New Non-Profit Economy

On Saturday, October 29th at the PNW Block Party Offseason Competition, FRC 7461 experienced a major failure of the Power Distribution Hub. This was the leading cause of FRC 7461’s sub-optimal performance.

This report explains the events that took place, provides tangible action items to avoid such events in the future and concludes the root cause of the issue. 

## **Analysis**

To understand the root cause of PDH failure, it is important to understand the electrical system architecture. Power and Ground wires from the battery route into the PDH. The power wire is spliced by the Circuit Breaker while the ground wire is directly routed into the PDH. During battery replacement, both wires are tensioned. The power wire is mounted on the circuit breaker rendering it immune to said tension. However, the ground wire places direct tension onto the PDH port. Previous generation distribution panels (CTRE PDP) use a nut and washer assembly to mount the ground wire which is far more resilient to external forces. The PDH, in contrast, used WAGO connectors which function based on metallic cage clamps exerting force onto the wires. This system does not have a bolt preventing unsanctioned removal. Over multiple battery replacements, the ground wire can be slipped out of the port.

This was the most likely cause. The lack of thorough electrical inspections did not help in any regard. It is likely the case that the robot entered Block Party with loose wires.

Additionally, after the practice match, or quals 1, the battery replacement may have been aggressive causing the wires to further shift out of the WAGO connectors.

By match 8, the robot was experiencing issues with the breaker tripping. The likely cause was a minor short circuit between the power and ground wires. It was small enough to be undetectable but large enough to cause the breaker to trip.

Breakers are designed so that the threshold for tripping lowers each time it trips. For this reason, breakers tripped in subsequent matches.

The pit crew replaced the breaker before match 19 but did not inspect the PDH directly after. The power wire may have slipped out further in the instance. In match 19, the robot was heavily defended by 4911. During this, 4911 slipped under the robot and caused the ground wire to shift more causing a large short circuit. This melted through other power wires causing a fire.

After this match, Pit Crew replaced the wiring but failed to inspect the PDH due to time constraints and the fact that the PDH was buried underwires. The Pit Crew failed to notice molten residue left inside the port and instead mounted the wires back into the WAGO port. This meant that the ground wire did not have a proper connection to the PDH and was not mounted. over the next two matches, the robot experienced power loss as the wire shifted from contact.

In match 27, the wire likely caused a short circuit that melted through the plastic casing on the WAGO connections.

---

## **Oversights**

## **Design**

1. While some electrical considerations were taken (placement of electrical components, and batteries), the wiring was not thoroughly planned out

2. Components on the electrical board were not placed until the design had been finalized.

3. The electrical subteam was not kept in the loop during the design process (members were absent).

4. Bellypan and electrical board were made of flex poly making damage to the electrical system during heavy defense easy.

5. Bellypan was smoked poly so it had to be removed for every inspection leading to hesitation/time delay for full systems check.

## **Mechanical/Pit Crew**

1. During battery mounting, stress is placed on the connections to the PDH.

2. Electrical fixes were not necessarily prioritized in pits.

## **Electrical**

1. Wiring was not planned well (in conjunction with design).

2. Not all electrical cables were tensioned.

3. Power and Ground wiring were not expected to slip — no effort was made to tension them.

4. Lack of cohesive systems check.

5. Lack of frequent systems checks.

6. PDH breaker and CAN layout not recorded or updated.

7. Wire- protection was not used as extensively as needed.

8. Hot glue for mounting was not used extensively.

## **Organization**

1. Electrical was given minimal time during robot wiring.

2. The electrical subteam has only 3 official members.

## **Drive Team/Programming**

1. No cohesive drive or testing logs led to a lack of contextual data.

2. Battery mounting

## **Business**

1. Electrical was not necessarily prioritized in the budget leading to a lack of spares.

---

## **Recommended Optimizations**

## **Design**

1. Components need to be placed on the CAD.

2. The wiring needs to be taken into consideration.

3. The electrical board needs to be machined

4. The electrical system needs to be better protected (Rigid Poly or integrated electrical)

5. if a belly pan is used, it needs to be transparent.

6. Prebuilt electrical board and drivetrain

## **Electrical**

1. Plan electrical in advance in conjunction with the design

2. Tension all cables

3. Make a cohesive electrical inspection checklist

4. Inspect electrical frequently and between every match

5. Protect all cables

6. Hot glue all connections

7. Record circuit breaker, ports, and CAN layout

## **Pit Crew/Drive Team/Programming**

1. Better battery mounting procedure

2. Prioritize Electrical in pits

---

## **Conclusion**

This report concludes that the direct cause of PDH failure was due to a combination of gradual degradation of ground wire connection integrity, along with heavy shock and stress experienced during [PNW Block Party 2022 Match 19 8](https://www.thebluealliance.com/match/2022wamc_qm19). Further failure was caused by the lack of time, organization, and extensive inspection procedures in the pit leading to greater damage in the PDH.

It also concludes that the indirect cause of PDH failure is poor electrical considerations in design, improper battery mounting procedures, and lack of extensive inspection procedures.

This report details recommended optimizations to multiple processes to prevent both the direct and indirect causes of PDH failure.

Analysis of Power Distribution Hub Failure at PNW Block Party 2022

**ABSTRACT**

236 terawatt hours. That is how much power data centers consumed in 2021. From YouTube to Netflix, Wikipedia to ChatGPT, almost every application we rely on today uses cloud computing as its backbone. The need for cloud data centers is growing exponentially. An average datacenter of 50,000 Sq Ft. consumes 5 Megawatts of power. This is enough to power 5,000 homes. Even with renewable energy options expanding, the demand cannot keep up with the supply. This is not sustainable.

What if cloud resources were placed in outer space? In parallel to the growth in cloud computing usage, space technology has progressed wherein we are able to get ground speed bandwidth through satellite constellations, as SpaceX has proved.

Space has an unlimited supply of solar power and passive cooling solutions. The cost of satellite production, launch, and constellation technology is dropping down to approx. $100/satellite. This opens the possibility for a new solution that I have termed SpaceGrid.

In the SpaceGrid, individual satellites will function as compute racks and a constellation as a datacenter. The vision of SpaceGrid is to prove that compute nodes can be interlinked with laser-driven network backplanes. The one critical element that needs to be addressed to build the SpaceGrid is a low latency, high bandwidth network backplane that interlinks the satellite constellation.

To address this problem, I will develop a network of compute nodes on earth with commercial off-the-shelf (COTS) components, which are connected by laser edges. In this grid, data is transmitted via laser diodes and photovoltaic receivers, using a Fast Fourier Transform-based algorithm. In my model, the transmissions are routed using a Local Topology Sensitive Network. I will develop the algorithm to provide low latency and high bandwidth interconnectivity. This algorithm has the potential to be implemented at an industrial scale. Developing laser edge transmission and networking will enable space-borne grid computing and thereby lower surface energy consumption.

### **MOTIVATION AND APPROACH**

### **Problem**

Cloud services currently consume 1% of the global energy capacity. (Pesce, 2021). In the next few decades, cloud services will only grow exponentially, as resource-intensive applications such as cryptocurrency, artificial intelligence, genomics, and computational sciences continue to grow. For example, Oracle Cloud Infrastructure, the smallest of the four hyper-scale providers expects a 4x growth in the next 5 years. (Oracle, 2022) Since 2012, Moore's law has enabled power-efficient chips to grow at a steady rate. However, this is coming to an end as chip manufacturers are reaching physical limits in transistor density. Chip density cannot scale at the same rate as energy demand. Growth in cloud services far exceeds growth in energy production.    Bloomberg estimates that by 2030, cloud services may take up to 8% of global energy capacity. (Bass et.al 2017). With such a rapid increase in energy demand, it is not feasible to scale the current grid to reach this capacity even if energy production is doubled every year. Continuing to grow cloud data centers on earth-based grids is no longer feasible in the foreseeable future.

## **Current Solutions**

To manage the cloud industry’s power consumption, one current solution is to increase CPU density. CPU density packs more compute capacity into a single silicon, Dynamic Voltage Frequency (DVF) where CPUs that are idle are shut down and power is throttled for CPUs with low utilization. (Mastelic et al., 2015) In addition, Virtualization, where the same CPU cores are shared with multiple virtual servers, ensures improved utilization from <10% to >50%.

Between 2012 and now, while the compute capacity has quadrupled to over a billion active cores over the past decade, the power consumption by data centers has only grown by 20% - 30% (Masanet et al., 2020). This is because of Moore’s Law. However, Moore’s Law will hit its limit over the next decade. Denser Cores and Virtualization will no longer proportionally decrease power consumption.

In parallel, space technology has progressed offering us a viable alternative. For example, SpaceX through their Starlink constellations and Starship launch platform is bringing down the cost of manufacturing and launching satellites to near-earth orbits to about $10 per KG (Are Space Scientists, 2022). MIT Lincoln Laboratory recently demonstrated a sustained throughput of 100 Gbps from space to earth with laser, TBIRD (Communications System, 2022). Other factors like power and cooling have already been solved in space as seen on large GTO satellites and the ISS.

My solution builds on these innovations to enable data centers in space. More specifically, my solution addresses intra-satellite communication with lasers. Satellites can harvest the abundance of solar power to power micro-satellites that essentially act as data center racks. The primary challenge is interconnecting these satellites into a data center grid, similar to how data centers are built on Earth. My project will demonstrate how to connect the grid with a laser in space, with a focus on the key technologies of laser data transmission and high-level network routing.

**Proposal**

I propose to implement a networking layer of protocol and routing optimizations for data transmission within a satellite constellation, using laser technologies.

**Hardware Test Environment**

For testing purposes, I plan to build a network of computing and storage with COTS components to emulate satellites. I plan to test the reliability of transmissions, throughput, and latency. My plan is to use a network of Arduino boards as computing & storage nodes. I will connect these nodes through a laser diode and a photovoltaic cell as a receiver. The system will be powered through the Arduino board. First, I will use a potentiometer to modulate the current source and intensity in the laser diode. The Photoresistor will detect the spike in voltage, when it receives a signal from the laser diode and correlate it to a pulse.

**Transmission**

I plan to use Fast Fourier Transform (FFT) based algorithm for data transformation between nodes.

A laser can provide high throughput over long distances, as proven by MIT’s TBIRD, it is limited to single-bit transmission - “on” represented by a high-intensity pulse and “off” by a low-intensity pulse. This is highly inefficient, especially for large-scale data transmissions, typical in a data center backplane.

I propose modulating the intensity of the laser (regulated by a potentiometer) on the emitter, with which the receiver detects the intensity over a given time interval. Each packet can be encoded using a Frequency Shift Key (FSK) which transmits the digital signal on an analog carrier wave. A composite waveform of each FSK packet can be transmitted to which a FFT can be applied on the receiver to split the waveforms. This enables the transmission of multiple bits in the same packet and by extension multiple channels of communications

One drawback of this method is that the order of bits is lost. I propose to solve this problem by transmitting packet headers with the order of the bits in each transmission.

**Networking**

For networking, I plan to use a Local Topology Sensitive Network (LTSN) routing algorithm.

Datacenters typically need high throughput and low latency over short distances. Internode traffic is usually within a single domain. Traffic usually passes through a handful of nodes. LTSN with a localized A\* search combined with dynamically balanced partitioning will help me achieve this.

For this application, network routing needs to be maximized for efficiency, and speed due to the large data transmission. A localized A\* search within a dynamically balanced partition addresses this issue.

Each node on the network is aware of all edges and nodes n degrees from it. Within each node’s frame of reference, the topology map has a balanced partition that subdivides the nodes into topology-sensitive subsets (Hu, 2021). Each transmission within the network contains a header that defines the packet address and relative direction of transmission. Within the node, parallel A\* searches are conducted to find the optimal path within the frame of reference. Once a solution has been resolved, the packet address is updated with the sub-destination (objective inside the frame of reference) and routed accordingly. The packets are not fully decoded; instead, just the headers are decoded to save time and compute resources.

The key difference between this method vs a homogenous A\* or hub and spoke search is that it does not rely on a single subset of nodes to handle data transmission. Instead, each node can act as its own local map, and therefore find the optimal path within its frame of reference much faster than the absolute optimal path. This will enable me to scale my node constellation dynamically.