Simulated Roadster with cold-gas thrusters nails loops and flips

Nov 07, 2026 Last reply: 21 hours ago 4 Replies
Jump to replies (4)

When everything is running, the white sports car hammers up a large vertical loop, holds still upside down at the apex, drops back to the ground near a red semi-truck, and pulls off vertical jumps with exhaust plumes plus an aerial maneuver labeled KICKFLIP.



The project is a reverse digital twin of the Roadster fitted with fictional cold-gas thrusters. The vehicle is modeled as a four-wheel skid-steer, with no steering joints anywhere. On the ground it turns by running the left and right wheels at different speeds; in the air, the thrusters manage attitude. It has six free joints, three of them linear and three rotational, and the flight computer sums all 14 thrusters into a single net force and moment. Total mass comes to roughly 1,800 kg with a 1,700 kg chassis, and suction fans are on board to keep the car pressed to the track.



What stands out to me is the control loop. The control system is self-developed and runs in real time, taking the car's state and tuning the thrust levels moment to moment. Rough calculations of the vehicle's weight and the thrust it would need fed into the setup as well. The stack runs on Sodero Cloud with R2kernel as the software runtime, and rendering comes from NVIDIA Omniverse Kit in RTX realtime.



One gap remains open. The maker notes the propulsion is not realistic, particularly on peak thrust and gas budget, and the thrusters' gas flow is left out of the simulation. A single run would need about 700 kg of nitrogen, and in my view that figure shows how far the propellant side sits from reality.


How does the car steer while it is on the ground? I keep wondering whether there is a steering mechanism built into the model or whether the turning comes from something else entirely.

Modeling the thrusters as simple units with just a position, a direction and a thrust level strikes me as a clever call. It keeps the control problem lean, though the trade-off is that the gas side stays abstract, which matches the flow not being simulated.

I would suggest simulating the gas flow of the thrusters. The maker has said the gas budget is not realistic and a single run needs about 700 kg of nitrogen, so modeling the flow would show where the propellant actually goes during a jump.

The maker has covered that one. The vehicle is a four-wheel skid-steer with no steering joints, and it turns by spinning the left and right wheels at different speeds. The thrusters only take over attitude once the car leaves the ground.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required