Home  |  Notes  |  Builds  |  Photos
Shihao Cao

Shihao Cao

Lodestar: An Electric VTOL Rocket

Back to builds
Lodestar

An homage to SpaceX's Starship, Lodestar is a single point thrust vectoring VTOL drone.

Watch it Fly!

I worked on this project with Govind Chari! Click here to checkout his site!

Introduction

Lodestar is a small scale electric vertical takeoff and landing (VTOL capable) project that I worked on with Govind Chari over the 2020 Summer. I was in Hawthorne, so I designed the craft and wrote the flight software. My partner Govind built and tested the vehicle, and also wrote the GNC code. By the end of the summer, the vehicle could perform a takeoff and land vertically. The objective was to understand the challenges in creating a vehicle that could take off and land vertically through thrust vectoring only similar to the Falcon 9.


Years Before

The first attempts at this project actually started back in 2017 before I had even met Govind. Together with Matthew Cox, and Jude Bedessem, we studied how to build a similar system for recovery of high power rockets. While they worked on the rocketry side of the project, I lead the way on the recovery side with two pathfinder vehicles for an electric motor vertical landing system.


Pathfinder 1 - EDF Thrust Vectoring

This initial pathfinder taught me how to build a thurst vectoring actuators, and I was also able to test using an EDF as a main power system. I realized that the torque delivered from the motor needed significant compensation. Further that an EDF sized system was too large and heavy for safe prototyping. This drove me to target a lighter smaller system on the next iteration.

EDF Vectoring

Pathfinder 2 - Counter Rotating Props and Active Control

On this iteration, I tested using a commercial flight computer the Pixhawk 4, and two counter-rotating propellers. This also used a similar fin-based thrust vectoring system. In the picture below, you'll see it hanging from strings as a partial free-hanging system to test roll control. In this prototype, I found that this split prop design had a much better thrust-to-weight ratio, but that a Pixhawk flight computer did not have sufficient customization for this task. I would repeat the same mechanical and electrical design, but the software and control stack would need upgrading.

EDF Vectoring

Lodestar Design

Using lessons learned from my past designs, the bill of materials and overall system design (mass budget, power budget...) was very easy to get right on the first try. I chose to use two counter-rotating quadcopter motors to cancel torque and ensure centerline thrust. Roll control on the craft is achieved through differential thrust, and pitch & yaw is achieved through thrust vectoring vanes actuated by servos inside the thrust structure. The two motors have separate power systems, and one of them also powers the avionics stack.


Flight Software and Ground Software

Flight software on this vehicle was derived from the flight software on PAN (my satellite team). This gave me the fine-grained control that I did not have from previous commercial flight computers. I wrote all code to parse orientation and position data from the onboard IMU and Barometer. I wrote a mission manager state machine that guided the craft through sensor initialization, flight, and post-landing. To aid with initial development, I also developed a custom NodeJS-based web ground station. I also wrote all the boilerplate code that my partner plugged into for guidance, navigation, and control.

Here's the link to the GitHub repository for Lodestar.

Here's a screenshot of what the groundstation looked like as it relayed live telemetry through the radio system.

Lodestar Groundstation

Iteration

I worked with my partner to develop an iterative testing campaign that would guide us to full mission success. We created a Matlab simulation that verified simple attitude control PID loops. We then built the craft and did thrust testing to verify that it was capable of a greater than one thrust to weight ratio. To aid in the development, and figuring out why flight failures happen, I added SD-card data logging for post-flight analysis.


Roll Control

We verified the full control loop from sensors to actuators by only allowing freedom in the roll axis, and verifying that the craft could maintain a steady roll angle.


Final Flights

After lots of testing and a few total rebuilds of the vehicle, we finally got Lodestar to perform a vertical takeoff and landing after the 23rd test flight.

← Back to home                                                                                                                                                         Source