Project Leaderboard

Project Leaderboard#

Last updated: 02 February 2026

This leaderboard shows validated student submissions for the final course project (Unit 14: RL-based orbit station-keeping).

Ranking criteria (in order of priority):

  1. Days in orbit (higher = better sustained station-keeping)

  2. Fuel mass remaining (higher = better fuel efficiency)

  3. Score (average episodic reward)

Only submissions manually verified by the instructor are listed.
Students should submit via the designated channel (Google Form / email) including:

  • Link to your notebook/Colab (publicly viewable)

  • Model files (shareable Nexcloud or Drive folder or GitHub release)

  • Self-reported metrics and a short description

Model name (anonymously name to be added to leaderboard): Der Gerät Days in orbit: 180.0 days Fuel mass remaining: 352.15992895985124 kg Score = 253616.6155650075

Rank

Model name

Score

Days in orbit

Fuel mass remaining (kg)

Notes

1.

Der_Gerät

253616.6155650075

180.0 days

352.12 kg

2.

Pointy_end_forward

238863.2635504955

180.0 days

330.54 kg

3.

I_Am_Really_Done_ok

238316.09242013458

180.0 days

329.72 kg

Used PPO with custom reward shaping featuring an altitude deadzone (5km) for fuel efficiency and an alignment penalty for integrated control. The agent acts as a numerical decoupler for the hierarchical control scheme in Sub-Problem 3.

4.

Return_to_Sender_v16

235598.18129603358

180.0 days

325.57 kg

5.

Fawkes V18

233299.97598548885

180.0 days

322.03 kg

6.

Spyder_007

232306.17917220743

180.0 days

320.48 kg

7.

CmπF_The_End_V2

231448.81516948578

180.0 days

319.14 kg

8.

Houston_We_Have_A_Lot_Of_Problems

228407.15166331158

180.0 days

314.34 kg

Minimal change in PPO agent hyperparameters

9.

Phoenix_V2

230374.80509389436

180.0 days

317.45 kg

Reward Function was inspired from the research paper - https://www.sciencedirect.com/science/article/pii/S0273117725004533

10.

Iamconfused_V1

229503.53107895696

180.0 days

316.08 kg

-

11.

Not_Just_PPO_model_v4

229001.36

180.0 days

315.28 kg

-

12.

AstraKeep_SC_1100k

229001.36

180.0 days

315.28 kg

PPO-based high-level station-keeping controller with integrated reward shaping directly inside the environment definition. Instead of using the separate “Reward Shaping” section, I embedded a physics-informed energy-based reward into the PPO training workflow.

13.

Just_PPO_model_v3

228152.73483114876

180.0 days

313.93 kg

-

14.

Bond007

228152.73483114876

180.0 days

313.83 kg

-

15.

Houston_We_Have_Some_Problems

227297.3329480142

180.0 days

312.57 kg

-

16.

CMπF_The_End_V1

226888.9319445231

180.0 days

311.91 kg

-

17.

Houston_We_Have_No_Problem

223453.60

180.0 days

306.35 kg

18.

unknown_spacedude

218444.36739086636

180.0 days

298.07 kg

19.

Humanized

214797.17081679983

180.0 days

291.88 kg

20.

Red_Falcon_V5

213529.6644028123

180.0 days

289.71 kg

21.

Mars??_Not_now

207223.52786880732

180.0 days

278.61 kg

22.

maximum_effort_v3

201071.4436535585

180.0 days

267.34 kg

23.

Space_King

193468.52150566512

180.0 days

252.72 kg

24.

Just_PPO_model_v1

193108.07718246026

180.0 days

252.00 kg

Basic PPO model without any integration of reward functions. The solver from the environment.py is changed to LSODA instead of RK45.

25.

AGENT ZORO

186939.11961551246

180.0 days

239.46 kg

26.

I_swear_I_tried

172391.05231180333

180.0 days

206.86 kg

27.

Miauuu

160291.37954613447

180.0 days

175.19 kg

28.

Schrodingers_cat

150011.77786515633

180.0 days

142.87 kg

29.

minimum_effort_v2

19265.44

180.0 days

0.0 kg

Minimum passing grade as defined by the rules

Legend

  • Score: Average episodic reward over standardised evaluation (higher is better)

  • Days in orbit: Average number of days the spacecraft remains within station-keeping bounds before violation (higher is better)

  • Fuel mass remaining: Remaining propellant mass (kg) at episode end or bound violation (higher is better)