How rupam496/c-v2x-enabled-cooperative-autonomous-driving-using-multi-agent-reinforcement-learning works
The repository describes two product tracks: RAG-assisted automotive test generation and a two-vehicle cooperative-driving learner. The latter is represented in source by a CARLA environment, an idealized V2X channel, and recurrent MAPPO training components. RAG and test-generation stages are documented in the README but have no implementation paths in the supplied tree; their relationships are documented there. V2X wiring into the environment is not shown, so that link is marked unverified.
Main entry points: train.py Training loop · carla_env.py Intersection environment · mappo.py MAPPO optimizer · v2x.py V2X channel
Components 14
Test generation
- Automotive requirement
- Rule retrieval
- Test-case generation
- Test-case simulation
Driving simulation
- Intersection environment
carla_env.py - V2X channel
v2x.py
MAPPO learning
- Training loop
mappo/train.py - Rollout buffer
mappo/buffer.py - MAPPO optimizer
mappo/mappo.py - Saved policy
mappo/mappo_final.pth
Policy networks
- LSTM actor
mappo/actor.py - LSTM critic
mappo/critic.py
Other components
- Requirement author
- CARLA simulator
Connections 13
- Requirement author → Automotive requirement: specifies
README.md - Automotive requirement → Rule retrieval: retrieves rules
README.md - Rule retrieval → Test-case generation: provides context
README.md - Test-case generation → Test-case simulation: simulates cases
README.md - Test-case simulation → CARLA simulator: runs in
README.md - Training loop → Intersection environment: steps
mappo/train.py - Intersection environment → CARLA simulator: connects to
carla_env.py - Training loop → MAPPO optimizer: selects actions and updates
mappo/train.py - Training loop → Rollout buffer: stores rollouts
mappo/train.py - MAPPO optimizer → LSTM actor: calls
mappo/mappo.py - MAPPO optimizer → LSTM critic: calls
mappo/mappo.py - Training loop → Saved policy: saves weights
mappo/train.py - Intersection environment → V2X channel: may exchange state