How rugvedchauragade/rugved---walk-agent-v1 works
Rugved is a simulation-only quadruped locomotion reinforcement-learning project. The workflow defines a Gymnasium walking environment and reward, trains a PPO policy, evaluates clean and disturbed episodes against random and PID controllers, then exports frame-level schema and telemetry. Phase scripts are the product runtime; plots and build machinery are omitted. Source excerpts show the core environment, reward, training wrapper and downstream evaluation/export behavior. Later-phase links follow explicit README statements and SOURCE INDEX references; unsampled implementations are not independently verified.
Main entry points: walk_agent_phase7.py Unified Scenarios · walk_agent_phase5.py Disturbance Injection · walk_agent_phase2.py Reward Strategies · walk_agent_phase1.py Walk Environment
Components 12
Walking Simulation
- Walk Environment
walk_agent_phase1.py - Reward Strategies
walk_agent_phase2.py - Disturbance Injection
walk_agent_phase5.py
Policy Learning
- PPO Training
walk_agent_phase3.py - Trained PPO Model
walk_agent_model.zip
Evaluation
- Episode Evaluation
walk_agent_phase4.py - RL vs PID
walk_agent_phase6.py - PID Controller
walk_agent_phase6.py
Telemetry Output
- Unified Scenarios
walk_agent_phase7.py - Frame Schema
walk_agent_schema.json - Telemetry Log
walk_agent_telemetry.csv
Other components
- Operator
Connections 23
- Operator → Walk Environment: runs phase
README.md - Operator → Reward Strategies: runs phase
README.md - Operator → PPO Training: runs phase
README.md - Operator → Episode Evaluation: runs phase
README.md - Operator → Disturbance Injection: runs phase
README.md - Operator → RL vs PID: runs phase
README.md - Operator → Unified Scenarios: runs phase
README.md - Reward Strategies → Walk Environment: uses environment
walk_agent_phase2.py - PPO Training → Walk Environment: wraps environment
walk_agent_phase3.py - PPO Training → Reward Strategies: applies reward
walk_agent_phase3.py - PPO Training → Trained PPO Model: saves policy
README.md - Episode Evaluation → Walk Environment: evaluates episodes
walk_agent_phase4.py - Episode Evaluation → Reward Strategies: scores episodes
walk_agent_phase4.py - Disturbance Injection → Walk Environment: extends step
walk_agent_phase5.py - Disturbance Injection → Reward Strategies: scores episodes
walk_agent_phase5.py - RL vs PID → PID Controller: runs controller
walk_agent_phase6.py - RL vs PID → Disturbance Injection: compares disturbed runs
README.md - RL vs PID → Trained PPO Model: loads policy
README.md - Unified Scenarios → Trained PPO Model: loads policy
walk_agent_phase7.py - Unified Scenarios → PID Controller: runs controller
README.md - Unified Scenarios → Disturbance Injection: uses scenarios
README.md - Unified Scenarios → Frame Schema: writes JSON
walk_agent_phase7.py - Unified Scenarios → Telemetry Log: writes CSV
walk_agent_phase7.py