File size: 3,457 Bytes
604e535
 
8e384df
604e535
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8e384df
604e535
 
 
 
cc396fd
 
 
604e535
 
 
 
 
 
 
 
 
8e384df
604e535
 
 
 
 
 
 
 
 
 
 
db21b01
 
604e535
 
 
 
 
 
 
 
 
 
 
 
 
 
ccf9f1b
 
604e535
 
 
 
 
 
 
 
 
ccf9f1b
604e535
 
 
 
 
 
 
ccf9f1b
604e535
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
# Experiments

This directory contains the paper-facing experiment code, checkpoints, results, GIFs, tables, and reports.

This directory contains two formal experiment categories:

- **A. Learned world models**: trainable image-input WMs evaluated on rollout prediction and WM-based planning.
- **B. Traditional non-WM controllers**: hand-designed control baselines evaluated on the same downstream tasks.

Main method:

- **FlowMo**: Flow-Momentum World Model, the proposed drift-aware world model for surface vehicles.

Category A learned WM comparisons:

- **LeWorldModel**: JEPA-style latent predictor under the shared clean-image protocol.
- **PlaNet RSSM**: recurrent state-space world-model baseline under the shared clean-image protocol.
- **TD-MPC2 Dynamics**: task-oriented latent dynamics baseline under the shared clean-image protocol.

Purpose of Category A: compare world-model architectures under identical image data, optimizer budget, rollout target, and evaluation protocol.

Category B traditional controllers:

- **PID/LOS controller**
- **No-Flow LOS Controller**
- **Current-Estimator LOS Controller**
- **Oracle-Flow LOS Controller**

Purpose of Category B: compare downstream task behavior against non-neural controllers that do not train a world model.

Baseline details are documented in `BASELINES.md`; the full experiment protocol is documented in `docs/EXPERIMENT_PROTOCOL.md`.

Design principles:

- Shared simulator, datasets, planning utilities, metrics, and visualization live in `shared/`.
- Each method has its own directory with `src/`, `checkpoint/`, and `result/`.
- Paper artifacts are collected under `reports/`.
- Method names should be explicit and readable. Avoid cryptic suffixes in paper-facing file names.

Standard method interface:

```text
src/model.py      # build_model(), load_model()
src/train.py      # train(config)
src/predict.py    # rollout(model, batch)
src/config.py     # default_config()
```

Closed-loop planning for learned world models is implemented once in `evaluate_image_planning.py` so every learned method is evaluated through the same CEM interface.

Traditional controllers use:

```text
src/controller.py or src/mpc.py
src/evaluate.py
src/config.py
```

Formal clean-image configuration:

```text
image_size=160
visual_scale=2.5
train=data/paper/train.npz
test=data/paper/test.npz
flow_families=noflow, uniform, vortex_center, double_gyre, source_sink, source_sink_pair, gradient, shear, turbulent_patch, random_fourier
```

Full paper-facing image pipeline:

```bash
python -m experiments.run_paper_image_pipeline
```

The default command runs the paper configuration end to end: train all learned world models, evaluate long rollout prediction, run FlowMo latent probes, evaluate closed-loop planning against traditional controllers, generate GIFs, and write the final report. Images are rendered online from simulator states, so no separate image-cache preparation step is required.
All flow fields are static. Localized flow structures are sampled near task routes so that boat trajectories encounter non-uniform current in the shared train/test/final protocol.

Manual image training:

```bash
python -m experiments.train_image_world_models
python -m experiments.evaluate_image_world_models
python -m experiments.evaluate_flowmo_latent_probes
python -m experiments.evaluate_image_planning --task reach_target --boat twin
python -m experiments.summarize_paper_image_results
```