Add parameterized social dilemma game with N-player support, dynamic …

[sriramsowmithri9807]

Add Parameterized Social Dilemma Game for MARL Research

Summary

This pull request introduces a new OpenSpiel game, python_param_social_dilemma, designed as a flexible benchmark for multi-agent reinforcement learning. The game generalizes classic social dilemma and matrix games by supporting variable numbers of agents, dynamic payoff structures, and optional stochastic rewards.

Resolves: #[issue_number]

---

Key Features

• Supports N-player simultaneous-move games (N ≥ 2)
• Configurable dynamic payoff matrices for non-stationary environments
• Optional stochastic rewards for robustness and exploration studies
• Includes Axelrod-style benchmark bots for baseline evaluation

---

Changes Included

• Core game implementation and registration
• Comprehensive unit tests for the game and bots (21 tests, all passing)
• Example scripts demonstrating usage, scalability, and tournaments
• Documentation covering configuration, usage, and design rationale

---

Motivation

Existing OpenSpiel social dilemma games are primarily limited to 2-player, static, deterministic settings. This addition enables research on:

• Cooperation and competition in larger groups
• Adaptation under changing reward dynamics
• Robust MARL algorithms in stochastic environments
• Standardized benchmarking against classical strategies

---

Technical Overview

• Game type: Simultaneous-move, general-sum
• Information: Perfect information
• Reward model: Deterministic or stochastic
• Dependencies: NumPy only
• Default formulation: Public-goods-style social dilemma

---

Testing and Compliance

• Follows OpenSpiel Python game conventions
• Uses pyspiel.Game and pyspiel.State base classes
• Observer implementation included
• Example scripts verified
• No breaking changes to existing code

---

Future Work

• Additional social dilemma variants
• Support for asymmetric payoff structures
• Visualization tools for dynamics and learning behavior
• Extended benchmarking utilities

---

Acknowledgments

Thanks to @lanctot for implementation guidance and @alexunderch for pointers to relevant OpenSpiel documentation.

[lanctot]

Ok, thanks. A few changes before I trigger the tests.

• Can you remove PR_SUMMARY.md?
• Can you move everything in IMPLEMENTATION_SUMMARY.md to a section at the bottom of param_social_dilamma_README.md, and then also remove that file?
• Can you add the new test in python/CMakeFiles.txt around line 239

[lanctot]

Thanks! Looking good, yes.

I sent a few requests for changes before I trigger the tests.

[sriramsowmithri9807]

@lanctot done sir, checkout the changes!!

[lanctot]

@lanctot done sir, checkout the changes!!

I still see both of the top-level markdown files (see here). Did you remember to push the changes?

[lanctot]

Note: the changes should be committed and pushed in your branch. The last commit above seems to be an empty change, so maybe you forgot to add the files?

[alexunderch]

One more note: to delete already committed files, it's not enough to delete them from your local branch, but to remove its node from the commit tree using, for example, git rm --cached FILENAME.

[sriramsowmithri9807]

@alexunderch done sir, checkout once!!

[lanctot]

@alexunderch done sir, checkout once!!

I am still seeing 10 files: https://github.com/google-deepmind/open_spiel/pull/1457/changes

[lanctot]

You have to commit and then push after the deletions.

[sriramsowmithri9807]

#1458 check out this new PR sir!!

[lanctot]

Replaced by #1458