DANIEL CAMPOS

B.S. Mechanical Engineering & CS Minor (SDSU)

Recent ME grad looking to work on robotics and electromechanical hardware. This site is a collection of my design reports, CAD models, competition builds, and work from my internships.

Click the resume to download a PDF copy.

SDSU LUNABOTICS COMPETITION TEAM

ATHENA is a competitive lunar rover built for the NASA Lunabotics program and ran in the highly abrasive regolith (lunar soil) at the Exolith Lab and Kennedy Space Center Swamp Works lab. Despite being a completely new, first-year team with zero returning members or prior competition experience to lean on, we placed 5th out of 73 teams nationwide.

As the Lead Engineer for the excavation system, my job was to design, manufacture, and integrate the physical hardware that digs the regolith. I ran simulations in SolidWorks and Onshape to ensure our load-bearing linkages could withstand extreme torque spikes, specifically basing our factors of safety around a worst-case belt jam scenario where the motors would attempt to over-tension and collapse the sidearm. I manufactured the components in-house at the SDSU machine shops using waterjets, manual mills, and lathes. On the electromechanical side, I routed the wiring for the actuators and control electronics and spent our physical testing sessions tracking down system brownouts and fixing hardware failures when the rover was pushed to its limits.

KUKA iiwa 7-DOF Manipulator

Pick-and-Place Kinematics Modeling

KUKA LBR iiwa 7 R800 (Image via Robots.com)

Modeled KUKA Robot Arm (Orange) movement through the multi-layer environment with obstacles.

To execute a real-world pick-and-place operation, a robotic arm must navigate complex workspaces without colliding with the environment or itself. Using MATLAB's Robotics System Toolbox, I stripped the standard URDF collision meshes from a 7-axis KUKA iiwa manipulator, rebuilt a simplified custom collision architecture, and generated collision-free inverse kinematics paths across a multi-tiered obstacle course.

This project models the collision-free trajectory of a 7-degrees of freedom KUKA iiwa robotic manipulator in MATLAB. The objective was to compute an inverse-kinematics path that moves target objects between platforms inside a dense obstacle environment without the arm colliding with itself or the surroundings. Built using the MATLAB Robotics System Toolbox, the motion planning stack combines a numerical Inverse Kinematics solver (inverseKinematics) with an RRT obstacle planner (manipulatorRRT), passing the resulting waypoints through a pchip polynomial algorithm to give smooth and continuous joint movement.

Environment & Kinematic Setup

Instead of building a rigid-body tree link by link, the pre-structured URDF model was loaded directly into the workspace. To optimize the computational speed of the collision-checking algorithm, the imported visual meshes were stripped and replaced with simplified custom primitives (collisionBox, collisionCylinder, and collisionSphere).

The simulated workspace contained three separate landing platforms, two vertical obstacle pillars, and two spherical target objects:

Denavit-Hartenberg (DH) Parameters — KUKA iiwa 7-R

Serial manipulator joint frame assignments and Jacobian derivatives (Fu & Spyrakos-Papastavridis, 2020).

Joint (i) Type Twist (α) Length (a) Offset (d) Angle (θ) Role
1 Revolute 0 mm 360 mm θ₁ Base Rotation
2 Revolute -90° 0 mm 0 mm θ₂ Shoulder Pitch
3 Revolute +90° 0 mm 420 mm θ₃ Shoulder Roll
4 Revolute +90° 0 mm 0 mm θ₄ Elbow Pitch
5 Revolute -90° 0 mm 400 mm θ₅ Elbow Roll
6 Revolute -90° 0 mm 0 mm θ₆ Wrist Pitch
7 Revolute +90° 0 mm 126 mm θ₇ End-Effector Roll

Algorithmic Motion Planning

The simulated cycle was broken into three sequential routines: home-to-pick, object grasp, and pick-to-place.

  • A numerical solver was weighted heavily toward Cartesian positional accuracy over joint orientation to successfully converge on reachable grasp coordinates above the spheres.

  • Waypoints were mapped through a Rapidly-exploring Random Tree planner (manipulatorRRT) with self-collision checking enabled to navigate around the obstacles in the environment, such as the vertical pillars or platforms.

  • Since the raw RRT output creates jagged, instantaneous directional shifts, the waypoints were passed through a pchip interpolation algorithm. This generated 200 continuous frames of motion, preventing infinite jerk on the simulated actuators.

Technical Bottlenecks & Debugging

Developing the solver required resolving several common manipulator edge-cases:

  • Early iterations caused the manipulator to pass directly through the square obstacle pillars. This was resolved by re-centering the primitive bounding boxes relative to the link coordinate frames.

  • When target coordinates sat near the outer boundary of the workspace, the solver would terminate early or generate erratic "hunting" motions before settling on a trajectory. Tuning the RRT step-size thresholds (MaxConnectionDistance = 0.2, ValidationDistance = 0.05) stabilized path convergence.

SELF-HOSTED CLOUD INFRASTRUCTURE

Private Linux File Server & Tailscale Tunneling

This project covers the deployment and ongoing administration of a private, self-hosted Nextcloud file server built to replace commercial cloud storage subscriptions. The objective was to make a secure, 4-Terabyte remote storage vault running on repurposed desktop hardware in El Paso, Texas, accessible transparently across Windows, Android, and mobile devices. Built on a headless Ubuntu Server command-line interface, the system bypasses traditional home router port-forwarding by routing all client synchronization, Windows WebDAV mounting, and remote SSH terminal management exclusively through an encrypted Tailscale WireGuard mesh tunnel.

Class files being accessed from an Android Phone

During my time at myCREcloud, I deployed enterprise QNAP file servers, configured Proxmox hypervisors for remote virtual machines, and managed user permissions across Windows Domain Controllers. This exposure to IT infrastructure highlighted the potential of self-hosted networking.

Concurrently hitting the storage ceiling on my commercial cloud accounts, I repurposed an unused desktop tower pc sitting 700 miles away in my hometown of El Paso, Texas. The personal challenge was to step away from graphical interfaces and build a headless Linux environment managed exclusively via command line.

System Architecture & Security

The server is built on an Ubuntu CLI environment. Navigating conflicting apt and snap package configurations required several full system tear-downs before getting a stable Nextcloud deployment.

Rather than relying on traditional port-forwarding, which exposes home router IP addresses to automated botnet scanners, remote access is routed exclusively through Tailscale. This creates an encrypted peer-to-peer mesh network, allowing me to SSH into the Host machine in El Paso or access files from anywhere over standard Wi-Fi or cellular data through the Nextcloud app.

Class Files being access from a windows computer via the file explore

Configured Windows WebDAV integration to mount the remote El Paso server natively inside Windows File Explorer as a local drive, while utilizing Nextcloud mobile clients to automatically ingest and consolidate over a decade of archived phone photography into a single unified database.

Technical Bottlenecks & Roadmap

  • Binary Dependency Mapping: Server-side automated facial and object recognition is currently offline. The Nextcloud instance fails to recognize the active system Node.js binary path despite the installed package matching the version requirement. Troubleshooting is ongoing to resolve the environment variable mapping.

  • Future Hardware Acceleration: Once the Node.js pipeline is stabilized, the objective is to offload the image processing models from the host CPU to the dedicated system GPU to reduce background indexing overhead.

  • Lack of Data Redundancy: The 4TB storage array currently operates as a single point of failure (JBOD) with zero RAID parity mirroring. Mitigating this risk by implementing an automated off-site backup cron job or converting the storage pool into a ZFS mirrored array is the highest priority for the next hardware iteration.

EXPERIENCE

Full Torsion Bar Assembly Summary

HIGH-LOAD TORSION BAR SUSPENSION

Suspension Design for a 5,750 lb SUV

Designed to handle rough terrain on a heavy passenger SUV, this 4130 steel torsion bar acts as a compact, high-torque spring. Sweeping across a 4-degree lever arm range, the shaft absorbs an average 1,700 lbf dynamic wheel load while protecting its root splines from shear failure.

Specifications

  • Target Vehicle is a ~5,750 lb SUV (e.g., Ford Bronco) supporting 1,500 lbf static weight per wheel

  • The Shaft is a 145-inch hollow bar with a 2.15-inch inner diameter and 2.751-inch outer diameter

  • Factor of safety of 10.41 in static torsion against a 1.15 baseline requirement

  • Spline Interface has a ANSI B92.1 20-tooth involute profile with a 30° pressure angle

Range of Motion of the Lever Arm

Because the system relies on a fixed steel shear modulus, the bar's inner and outer diameters strictly dictate its travel suspension. Tuning those dimensions allowed the bar to absorb 889.6 lbf at its 47° minimum operational angle and 2,511.6 lbf at its 51° maximum bottom-out angle.

Torsion Bar Validation - Goodman Fatigue & Joint Stiffness

Dynamic Fatigue Failsafes

S-N Fatigue Graph for 4140 Q&T 205°C 400°F Torsion bar.

To protect against severe highway washboarding, the shaft was modeled against a fully reversed modified Goodman fatigue criterion rather than standard unidirectional cyclic loading.

  • Surface Finish Salvage: Specified a precision ground finish (a = 1.34, b = -0.085) to preserve the bending endurance limit.

  • Fatigue Margin: Cleared the critical fatigue threshold with a 2.71 factor of safety at 99.9% reliability.

Spline & Fitting Optimization

Torsion Bar Detail

Fixed Coupling Detail

  • Spline Tooth Shear: Sized the outer shaft diameter down to 2.673 inches to minimize polar moment of inertia, clearing tooth shear at a lean 1.058 safety factor.

  • Bearing Support Lug will be machined from a single 0.5-inch thick block of 3003-H16 aluminum to resist offset loading, yielding a 4.16 safety factor.

  • Joint Separation: Sized SAE Grade 2 mounting bolts to hold a 2,862 lbf preload, maintaining a joint separation margin of 4.46.

Bearing Support Detail Dimensions

Distributed Load, Shear Force, and Moment Diagrams of Axle

Distributed Load Diagram of Axle

Shear Force Diagram and Bending Moment Diagram for Axle

AIRCRAFT ACCESSORY GEARBOX

Packaging 4 Systems Under a 260 lbf Limit

Modern commercial aircraft engines pack massive power into tight, curved spaces. Integrated directly beneath the compressor casing, this accessory drive gearbox takes 750.3 in-lbf of input torque at 8400 RPM from the turbine transfer shaft and splits it to drive four vital flight systems simultaneously.

Specs:

  • Input Power of 750.3 ± 0.5 in-lbf delivered at 8400 RPM

  • Total Weight of 248.72 lbf achieved against a 260 lbf maximum ceiling

  • 15 meshing spur gears and 1 input bevel gear forged from Nitralloy 135 M

  • Shaft made of AISI 1050 steel Q&T sized for infinite fatigue life

Packaging

Accessory Component Locations

Fitting a 4.75-inch wide gear train inside the curved profile of an engine compressor casing required compounding three sets of gear shafts. The system was tuned to hit four distinct component target speeds:

  • Oil Pump: 259.5 in-lbf delivered at 8500 RPM

  • Fuel Pump: 252.1 in-lbf delivered at 7504 RPM

  • Electrical Generator: 193.9 in-lbf delivered at 6503 RPM

  • Hydraulic Pump: 270.1 in-lbf delivered at 3500 RPM

Gearbox Validation - AGMA Stress & Neuber Shaft Analysis

AGMA Gear Stress Validation

Shear And Bending Moment Diagrams For Drive Shaft (Portion Inside Gear Box, Gears 2 and 9)

All 16 gears were designed to AGMA No. 10 quality standards under moderate shock loading. Nitralloy 135 M was selected across Grades 1 through 3 to optimize contact and bending strength post-nitration. Gear 9 exhibited the lowest bending safety factor at 1.36 against a 1.30 project requirement. Gear 9 also represented the primary pitting risk, clearing the allowable contact threshold with a 3.07 safety factor.

Shaft Fatigue & Stress Concentrations

S-N Curve of 1050 Steel Q&T 400F at the Most Loaded Location of the Drive Shaft

The assembly utilizes twelve uniform 1-inch outer diameter shafts to standardize bearing depths. The main input drive shaft absorbed the most severe compound loads:

  • Retaining Ring Grooves: Achieved a 1.639 safety factor calculated via Neuber shear and bending constants.

  • Keyway Joints: Cleared dynamic loading with a 1.987 safety factor.

  • S-N Life Assessment: Plotted onto the 1050 Steel Q&T 400°F curve, proving the shaft operates strictly in its infinite life zone.

Shear and Bending Moment Diagrams

Shear And Bending Moment Diagrams For Drive Shaft (Portion Inside Gear Box, Gears 2 and 9)

Shear And Bending Moment Diagrams For Drive Shaft (Portion Outside Gear Box, Bevel Gear)

The last day of my Internship at ASML with the Configuration Management Team

Behringer, my son.

Accepting the Outstanding Council Award on behalf of our board during my 2025–2026 Presidency.

Cave Exploring in Julian CA