RenaultNissan-v0

Category: Validated Case Study Agents: 2 (Nissan, Renault) Difficulty: Advanced Source: coopetition_gym/envs/case_study_envs.py

Overview

RenaultNissan-v0 models the Renault-Nissan Alliance across four distinct phases, as described in TR-2 (arXiv:2510.24909). This environment captures the 25+ year evolution of one of the most complex and enduring automotive alliances.

Unlike SLCD-v0 which models a single joint venture, RenaultNissan-v0 supports multi-phase simulation with different initial conditions for each era of the alliance.

Renault-Nissan Alliance Phases Trust evolution across four alliance phases: Formation (1999-2002), Mature Cooperation (2002-2018), Governance Crisis (2018-2020), and Strained Restructuring (2020-2025). Phase-specific initial conditions capture historical dynamics.

MARL Classification

Property	Value
Game Type	Markov Game (2-player, general-sum) with phase-dependent parameters
Cooperation Structure	Mixed-Motive with evolving power dynamics across phases
Observability	Full
Communication	Implicit (through actions only)
Agent Symmetry	Asymmetric (Renault: rescuer/0.52, Nissan: rescued/0.48)
Reward Structure	Mixed with phase-specific interdependence
Action Space	Continuous: A_Nissan = [0, 90], A_Renault = [0, 100]
State Dynamics	Deterministic with phase-dependent initial conditions
Horizon	Finite, T = 100 per phase
Canonical Comparison	Multi-phase alliance dynamics; cf. Segrestin (2005) “Partnering to Explore”

Validation Status: Parameters derived from TR-2, calibrated to Renault-Nissan Alliance phases (1999-2025).

Formal Specification

This environment is formalized as a 2-player Markov Game with configurable phase-specific initial conditions.

Agents

N = {Nissan, Renault}

Agent	Index	Endowment	Baseline	Bargaining α	Historical Role
Nissan	0	90.0	27.0	0.48	Rescued partner (1999)
Renault	1	100.0	30.0	0.52	Rescuer/controlling partner

Renault’s slight advantage reflects initial controlling stake.

State Space

S ⊆ ℝ¹⁷ (standard dyadic structure)

Action Space

Nissan: A_0 = [0, 90] ⊂ ℝ
Renault: A_1 = [0, 100] ⊂ ℝ

Uniaxial Treatment: This environment uses the single-dimension action space characteristic of Coopetition-Gym v1.x. Renault-Nissan competition in vehicle markets is captured through phase-specific trust dynamics rather than explicit competitive actions.

Phase-Specific Initial Conditions

Phase	Period	τ₀	R₀	Θ₀	Interpretation
`"formation"`	1999-2002	0.45	0.05	0.95	Uncertain but hopeful
`"mature"`	2002-2018	0.70	0.02	0.98	Strong collaboration
`"crisis"`	2018-2020	0.30	0.45	0.55	Post-Ghosn governance crisis
`"strained"`	2020-2025	0.40	0.35	0.65	Restructured, ongoing tension

Validated Trust Parameters (TR-2)

Parameter	Symbol	Value	Description
Trust Building Rate	λ⁺	0.08	Institutional trust (slow)
Trust Erosion Rate	λ⁻	0.25	Moderate erosion
Reputation Damage	$\mu_R$	0.55	Governance-related damage
Reputation Decay	$\delta_R$	0.02	Standard healing
Interdependence Amp.	ξ	0.50	Cross-cultural amplification
Signal Sensitivity	κ	1.0	Standard

Value Function

Parameter	Value	Interpretation
θ	22.0	Higher scale (automotive industry)
γ	0.58	Moderate platform synergies

Episode Structure

Horizon: T = 100 steps per phase
Truncation: t ≥ T
Termination: mean(τ) < 0.05 (alliance dissolution)
Discount: γ = 1.0

Phase Selection

env = coopetition_gym.make("RenaultNissan-v0", phase="crisis")

Phase determines initial τ₀ and R₀, modeling different eras of the alliance.

Historical Background

The Renault-Nissan Alliance (1999-Present)

Partners:

Nissan Motor Corporation: Japanese automaker, near-bankruptcy in 1999
Renault SA: French automaker, seeking global expansion

Key Events:

1999: Renault rescues Nissan with $5.4B investment
2002-2018: Successful collaboration, shared platforms
2018: Carlos Ghosn arrest, governance crisis
2020+: Alliance restructuring, strained relations

Why Multi-Phase?

The alliance demonstrates:

Trust Recovery: Nissan’s trust in Renault rebuilt after rescue
Trust Erosion: Growing tensions over governance
Crisis and Recovery: Post-Ghosn restructuring
Ongoing Dynamics: Current strained but continuing relationship

Environment Specification

Basic Usage

import coopetition_gym
import numpy as np

# Create environment with specific phase
env = coopetition_gym.make("RenaultNissan-v0", phase="formation")

obs, info = env.reset(seed=42)

print(f"Phase: {info['phase']}")
print(f"Period: {info['period']}")
print(f"Initial trust: {info['mean_trust']:.2f}")

Parameters

Parameter	Default	Description
`phase`	“formation”	Alliance phase to simulate
`max_steps`	100	Maximum timesteps per episode
`render_mode`	None	Rendering mode

Alliance Phases

Phase Configuration

Phase	Period	Initial Trust	Initial Rep. Damage	Interpretation
`"formation"`	1999-2002	0.45	0.05	Early uncertain partnership
`"mature"`	2002-2018	0.70	0.02	Successful collaboration
`"crisis"`	2018-2020	0.30	0.45	Post-Ghosn governance crisis
`"strained"`	2020-2025	0.40	0.35	Restructured but tense

Selecting a Phase

# Formation phase (early alliance)
env_formation = coopetition_gym.make("RenaultNissan-v0", phase="formation")

# Mature phase (successful collaboration)
env_mature = coopetition_gym.make("RenaultNissan-v0", phase="mature")

# Crisis phase (governance crisis)
env_crisis = coopetition_gym.make("RenaultNissan-v0", phase="crisis")

# Strained phase (current era)
env_strained = coopetition_gym.make("RenaultNissan-v0", phase="strained")

Agent Configuration

Endowments

Agent	Role	Endowment	Interpretation
Nissan (0)	Japanese Partner	90.0	Slightly smaller (historical)
Renault (1)	French Partner	100.0	Larger (controlling stake)

Bargaining Shares

Agent	Alpha	Interpretation
Nissan	0.48	Slight disadvantage (rescued party)
Renault	0.52	Slight advantage (rescuer)

Baselines

Agent	Baseline	As % of Endowment
Nissan	27.0	30%
Renault	30.0	30%

Interdependence Structure

Phase-Dependent Dependencies

Created using create_renault_nissan_interdependence(phase): The interdependence matrix varies by phase:

Formation: Nissan highly dependent on Renault (rescue)
Mature: More balanced mutual dependency
Crisis: Reduced dependency as alliance frays
Strained: Attempted rebalancing

Trust Dynamics

Parameters (All Phases)

Parameter	Symbol	Value	Source
Trust Building Rate	λ⁺	0.08	TR-2 validated
Trust Erosion Rate	λ⁻	0.25	TR-2 validated
Reputation Damage	$\mu_R$	0.55	TR-2 validated
Reputation Decay	$\delta_R$	0.02	TR-2 validated
Interdependence Amp.	ξ	0.50	TR-2 validated
Signal Sensitivity	κ	1.0	Standard

Phase-Specific Initial Conditions

Formation (1999-2002):

Trust = 0.45: Uncertain but hopeful
Rep. Damage = 0.05: Clean slate

Mature (2002-2018):

Trust = 0.70: Strong collaborative relationship
Rep. Damage = 0.02: Minor accumulated issues

Crisis (2018-2020):

Trust = 0.30: Severely damaged by governance scandal
Rep. Damage = 0.45: Significant institutional damage

Strained (2020-2025):

Trust = 0.40: Partial recovery
Rep. Damage = 0.35: Lingering damage

Value Function

Parameters

Parameter	Value	Description
θ	22.0	Slightly higher scale (larger industry)
γ	0.58	Moderate complementarity

Metrics and Info

The info dictionary includes:

Key	Type	Description
`step`	int	Current timestep
`phase`	str	Current alliance phase
`period`	str	Historical period string
`mean_trust`	float	Average trust level
`nissan_investment`	float	Nissan’s action
`renault_investment`	float	Renault’s action
`trust_matrix`	ndarray	Full trust matrix

Phase-Specific Analysis

Formation Phase Strategy

Historical context: Renault rescues near-bankrupt Nissan. Trust must be built.

env = coopetition_gym.make("RenaultNissan-v0", phase="formation")
obs, info = env.reset()

# Renault should invest heavily (demonstrate commitment)
# Nissan should respond cautiously (rebuilding)
renault_action = 70.0  # High investment
nissan_action = 50.0   # Moderate response

Expected dynamics:

Trust rises from 0.45 toward 0.60-0.70
Nissan gains confidence in Renault’s commitment
Value creation increases over time

Mature Phase Strategy

Historical context: Successful partnership with shared platforms.

env = coopetition_gym.make("RenaultNissan-v0", phase="mature")
obs, info = env.reset()

# Both partners maintain high cooperation
renault_action = 65.0
nissan_action = 60.0

Expected dynamics:

Trust stable around 0.70-0.75
Consistent high value creation
Minor fluctuations from market conditions

Crisis Phase Strategy

Historical context: Post-Ghosn arrest, governance scandal.

env = coopetition_gym.make("RenaultNissan-v0", phase="crisis")
obs, info = env.reset()

# Both partners are cautious
# Must avoid further trust erosion
renault_action = 45.0  # Defensive
nissan_action = 40.0   # Defensive

Expected dynamics:

Trust at risk of further decline
High reputation damage limits recovery
Careful cooperation needed to stabilize

Strained Phase Strategy

Historical context: Restructured alliance, ongoing tensions.

env = coopetition_gym.make("RenaultNissan-v0", phase="strained")
obs, info = env.reset()

# Gradual trust rebuilding
renault_action = 55.0
nissan_action = 50.0

Expected dynamics:

Trust can slowly recover
Reputation damage gradually heals
Long-term view required

Example: Multi-Phase Simulation

import coopetition_gym
import numpy as np

phases = ["formation", "mature", "crisis", "strained"]

for phase in phases: env = coopetition_gym.make("RenaultNissan-v0", phase=phase)
    obs, info = env.reset(seed=42)

    print(f"\n{phase.upper()} Phase ({info['period']})")
    print(f"  Initial trust: {info['mean_trust']:.2f}")

    # Run 50 steps with moderate cooperation
    for step in range(50): actions = np.array([50.0, 55.0])  # Nissan, Renault
        obs, rewards, terminated, truncated, info = env.step(actions)

    print(f"  Final trust: {info['mean_trust']:.2f}")
    print(f"  Trust change: {info['mean_trust'] - 0.5:.+.2f}")

Research Applications

RenaultNissan-v0 is suitable for studying:

Alliance Evolution: How partnerships change over time
Crisis Management: Recovery from governance failures
Cross-Cultural Dynamics: Japanese-French partnership
Long-Term Coopetition: Decades-long relationships
Phase Transitions: How alliances move between states

Comparison with SLCD-v0

Feature	RenaultNissan-v0	SLCD-v0
Phases	4 configurable	1 fixed
Duration	25+ years modeled	8 years modeled
Complexity	Multi-phase dynamics	Single trajectory
Best For	Phase analysis	Benchmark testing

SLCD-v0: Simpler validated case study
RecoveryRace-v0: Crisis recovery focus
PartnerHoldUp-v0: Power asymmetry dynamics

References

Pant, V. & Yu, E. (2025). Computational Foundations for Strategic Coopetition: Formalizing Trust and Reputation Dynamics. arXiv:2510.24909
Segrestin, B. (2005). Partnering to Explore: The Renault–Nissan Alliance. Research Policy.
Freyssenet, M. (2009). The Second Automobile Revolution. Palgrave Macmillan.

RenaultNissan-v0

Overview

MARL Classification

Formal Specification

Agents

State Space

Action Space

Phase-Specific Initial Conditions

Validated Trust Parameters (TR-2)

Value Function

Episode Structure

Phase Selection

Historical Background

The Renault-Nissan Alliance (1999-Present)

Why Multi-Phase?

Environment Specification

Basic Usage

Parameters

Alliance Phases

Phase Configuration

Selecting a Phase

Agent Configuration

Endowments

Bargaining Shares

Baselines

Interdependence Structure

Phase-Dependent Dependencies

Trust Dynamics

Parameters (All Phases)

Phase-Specific Initial Conditions

Value Function

Parameters

Metrics and Info

Phase-Specific Analysis

Formation Phase Strategy

Mature Phase Strategy

Crisis Phase Strategy

Strained Phase Strategy

Example: Multi-Phase Simulation

Research Applications

Comparison with SLCD-v0

Related Environments

References