Views
No views yet
| Parameter | Value |
|---|---|
| algorithm | TD3+BC |
| alpha | 2.5 |
| lr | 0.0003 |
| batch_size | 256 |
| discount | 0.99 |
| tau | 0.005 |
| policy_delay | 2 |
| num_iterations | 100000 |
| hidden_dims | [256, 256] |
| state_normalization | zero_mean_unit_var |
| action_normalization | [-1, 1] via joint limits |
| Metric | Value |
|---|---|
| action_mse | 0.374415 |
| gazebo_success_rate | 0/5 (0%) |
| gazebo_avg_lift | 0.0002 |
checkpoint.pt - Model weights (PyTorch)training_code.py - Training implementationtraining_log.csv - Training metrics over timeeval_gazebo.csv - Gazebo evaluation resultsdataset_stats.json - Dataset normalization statisticsconfig.json - Model configuration1import torch
2import numpy as np
3
4# Load checkpoint
5ckpt = torch.load("checkpoint.pt", map_location="cpu", weights_only=True)
6
7# Load dataset stats for normalization
8import json
9with open("dataset_stats.json") as f:
10 stats = json.load(f)
11state_mean = torch.tensor(stats["state_mean"])
12state_std = torch.tensor(stats["state_std"])1JOINTS = [
2 'shoulder_pan_joint', # idx 0
3 'shoulder_lift_joint', # idx 1
4 'upperarm_roll_joint', # idx 2
5 'elbow_flex_joint', # idx 3
6 'forearm_roll_joint', # idx 4
7 'wrist_flex_joint', # idx 5
8 'wrist_roll_joint', # idx 6
9 'l_gripper_finger_joint',# idx 7
10 'r_gripper_finger_joint',# idx 8
11]1@misc{fetch_offline_rl_pilot,
2 title={Offline RL Pilot Study for Fetch Robot Pick-and-Place},
3 year={2026},
4}