-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathver_hm3d_robot_nav.yaml
134 lines (123 loc) · 4.1 KB
/
ver_hm3d_robot_nav.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
# @package _global_
defaults:
- /habitat: habitat_config_base
- /habitat/task: pointnav
- /habitat/dataset/pointnav: hm3d
- /habitat_baselines: habitat_baselines_rl_config_base
- /habitat/simulator/sim_sensors@habitat.simulator.agents.main_agent.sim_sensors.spot_left_depth_sensor: spot_left_depth_sensor
- /habitat/simulator/sim_sensors@habitat.simulator.agents.main_agent.sim_sensors.spot_right_depth_sensor: spot_right_depth_sensor
- /habitat_baselines/rl/policy/obs_transforms:
- stitch_spot_front_vision
- override /habitat/task/actions:
- kinematic_velocity_control
- override /habitat/task/measurements:
- distance_to_goal
- collision_count
- backwards_motion_count
- auto_stop_success
- spl
- distance_to_goal_reward
- kinematic_nav_reward
- _self_
habitat:
task:
actions:
kinematic_velocity_control:
lin_vel_range: [ -0.5, 0.5 ]
ang_vel_range: [ -30.0, 30.0 ]
min_abs_lin_speed: 0.05 # meters/sec
min_abs_ang_speed: 3.0 # deg/sec
time_step: 0.5 # seconds
robot_urdf: "data/robots/hab_spot_arm/urdf/hab_spot_arm.urdf"
nominal_joints: [ 0, -180, 0, 135, 90, 0, -90, 0, 0, 60, -120, 0, 60, -120, 0, 60, -120, 0, 60, -120 ]
nominal_position: [ 0.0, 0.5, 0.0 ] # horizontal, vertical, and forward axes
nominal_rotation: [ 0.0, 0.0, 180.0 ] # roll, pitch, yaw in degrees
reward_measure: "kinematic_nav_reward"
measurements:
auto_stop_success:
success_distance: 0.425
simulator:
habitat_sim_v0:
# Physics must be enabled for EmbodimentCollisionChecker to work (sim.contact_test)
enable_physics: True
agents:
main_agent:
sim_sensors:
spot_left_depth_sensor:
height: 212
width: 120
hfov: 58.286
min_depth: 0.0
max_depth: 3.5
position: [ -0.03740343144695029, 0.5, -0.4164822634134684 ]
orientation: [ -0.4415683, -0.57648225, 0.270526 ]
spot_right_depth_sensor:
height: 212
width: 120
hfov: 58.286
min_depth: 0.0
max_depth: 3.5
position: [ 0.03614789234067159, 0.5, -0.4164822634134684 ]
orientation: [ -0.4415683, 0.57648225, -0.270526 ]
habitat_baselines:
verbose: False
trainer_name: "ver"
torch_gpu_id: 0
tensorboard_dir: "tb"
video_dir: "video_dir"
test_episode_count: -1
eval_ckpt_path_dir: "data/new_checkpoints"
num_environments: 64
checkpoint_folder: "data/new_checkpoints"
num_updates: -1
total_num_steps: 2.5e9
log_interval: 10
num_checkpoints: 100
# Force PyTorch to be single threaded as
# this improves performance considerably
force_torch_single_threaded: True
rl:
policy:
action_distribution_type: "gaussian"
name: "PointNavResNetPolicy"
ppo:
# ppo params
clip_param: 0.2
ppo_epoch: 2
num_mini_batch: 2
value_loss_coef: 0.5
entropy_coef: 0.01
lr: 2.5e-4
eps: 1e-5
max_grad_norm: 0.2
num_steps: 128
use_gae: True
gamma: 0.99
tau: 0.95
use_linear_clip_decay: False
use_linear_lr_decay: False
reward_window_size: 50
use_normalized_advantage: False
hidden_size: 512
# Use double buffered sampling, typically helps
# when environment time is similar or larger than
# policy inference time during rollout generation
use_double_buffered_sampler: False
ddppo:
sync_frac: 0.6
# The PyTorch distributed backend to use
distrib_backend: NCCL
# Visual encoder backbone
pretrained_weights: data/ddppo-models/gibson-2plus-resnet50.pth
# Initialize with pretrained weights
pretrained: False
# Initialize just the visual encoder backbone with pretrained weights
pretrained_encoder: False
# Whether the visual encoder backbone will be trained.
train_encoder: True
# Whether to reset the critic linear layer
reset_critic: True
# Model parameters
backbone: resnet18
rnn_type: LSTM
num_recurrent_layers: 2