Release 2.0

[daphne-cornelisse]

Lots of small edits before release 2.0

Done

• Add roadmap
• Edit release post
• Interactive demo
• Fix release post date
• Make sure default ini settings are preserved
• Switch out the current policy (.bin) for a better one that is tested with resampling in Carla maps
• Update the game files
• Include link to release video (yt)
• Tag the release

[greptile-apps]

Greptile Summary

This PR finalizes the PufferDrive 2.0 release with comprehensive documentation updates, improved simulator code, and new demo materials. The release introduces an interactive browser demo, updated policy weights trained on Carla maps, and enhanced multi-goal scoring logic.

Key changes:

• Added complete 2.0 release announcement with video, roadmap, and historical context
• Restructured documentation (simulator.md, train.md) for better user experience
• Updated policy weights from puffer_drive_csrz3f8j.bin to puffer_drive_weights_carla_town12.bin
• Improved scoring logic in drive.h with adaptive thresholds based on number of goals sampled
• Fixed collision tracking bug where collided_before_goal wasn't reset on respawn
• Added interactive WebAssembly demo with 28MB of game assets
• Cleaned up old README files and moved content to centralized docs
• Updated default configurations for Carla map training

Minor issues found:

• Typo in train.md ("cab" → "can")
• Grammatical error in drive.c comment
• Potentially unintended value in drive.ini controlled experiment config (goal_speed values include 3 which breaks the pattern)

Confidence Score: 4/5

• This PR is safe to merge with only minor issues that should be addressed
• The release contains substantial, well-organized changes with comprehensive documentation. Code changes are focused and improve existing functionality (scoring logic, collision tracking). However, there are three minor issues: a typo in documentation, a grammatical error in a comment, and a potentially incorrect configuration value that could affect controlled experiments. These are non-critical but should be fixed before release.
• pufferlib/config/ocean/drive.ini (check controlled_exp.train.goal_speed values), docs/src/train.md (typo fix needed)

Important Files Changed

Filename	Overview
docs/src/pufferdrive-2.0.md	Release announcement with highlights, interactive demo, history, and roadmap - well structured and complete
pufferlib/config/ocean/drive.ini	Configuration updates: goal_target_distance 25→30, changed controlled_exp parameters for hyperparameter sweeps
pufferlib/ocean/drive/drive.c	Demo settings updated for new policy, removed ini parsing from performance_test, adjusted reward values and episode length
pufferlib/ocean/drive/drive.h	Improved scoring logic for multi-goal scenarios, fixed collision tracking bug, rendering z-coordinate fixes, added tab key to switch agents
docs/src/simulator.md	Complete documentation rewrite - more user-friendly guide format with better structure and examples
pufferlib/resources/drive/puffer_drive_weights_carla_town12.bin	New policy weights trained on Carla maps, replaces old puffer_drive_csrz3f8j.bin

Sequence Diagram

sequenceDiagram
    participant User
    participant Browser
    participant Docs as Documentation
    participant Demo as WebAssembly Demo
    participant Sim as C Simulator
    participant Policy as Neural Network

    User->>Docs: Read release announcement
    Docs->>User: Show PufferDrive 2.0 features
    
    User->>Browser: Load interactive demo
    Browser->>Demo: Initialize game.wasm
    Demo->>Sim: Load map & policy weights
    Sim->>Policy: Initialize puffer_drive_weights_carla_town12.bin
    
    loop Episode (300 steps)
        User->>Demo: Press arrow keys (human control)
        Demo->>Sim: Step environment
        Sim->>Policy: Get actions for AI agents
        Policy->>Sim: Return actions
        Sim->>Sim: Update physics (classic/jerk dynamics)
        Sim->>Sim: Check collisions & goals
        alt Goal reached
            Sim->>Sim: Generate new goal (mode=1)
        end
        Sim->>Demo: Return observations & render
        Demo->>Browser: Display frame
    end
    
    Sim->>Sim: Log metrics (score, collision_rate, etc)
    Sim->>Sim: Reset episode

Loading

[greptile-apps]

_{23 files reviewed, 3 comments}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

style: Grammatical error - extra "that" in the sentence

Suggested change

	// trained with these exact settings, that changing them may lead to
	// trained with these exact settings, changing them may lead to

Prompt To Fix With AI

This is a comment left during a code review.
Path: pufferlib/ocean/drive/drive.c
Line: 37:37

Comment:
**style:** Grammatical error - extra "that" in the sentence

```suggestion
    // trained with these exact settings, changing them may lead to
```

How can I resolve this? If you propose a fix, please make it concise.

[greptile-apps]

logic: The goal_speed values appear to have an inconsistency - the sequence [10, 20, 30, 3] likely should be [10, 20, 30, 40] or [3, 10, 20, 30] based on the pattern

Suggested change

	values = [10, 20, 30, 3]
	values = [10, 20, 30, 40]

Was the value 3 intentional or should this be a more regular sequence?

Prompt To Fix With AI

This is a comment left during a code review.
Path: pufferlib/config/ocean/drive.ini
Line: 161:161

Comment:
**logic:** The `goal_speed` values appear to have an inconsistency - the sequence `[10, 20, 30, 3]` likely should be `[10, 20, 30, 40]` or `[3, 10, 20, 30]` based on the pattern

```suggestion
values = [10, 20, 30, 40]
```

 Was the value 3 intentional or should this be a more regular sequence?

How can I resolve this? If you propose a fix, please make it concise.

[greptile-apps]

syntax: Typo: 'cab' should be 'can'

Suggested change

	This will launch separate training runs for each value in the list, which cab be useful for:
	This will launch separate training runs for each value in the list, which can be useful for:

Prompt To Fix With AI

This is a comment left during a code review.
Path: docs/src/train.md
Line: 63:63

Comment:
**syntax:** Typo: 'cab' should be 'can'

```suggestion
This will launch separate training runs for each value in the list, which can be useful for:
```

How can I resolve this? If you propose a fix, please make it concise.