Skip to content

feat: learn every N steps with configurable frequency #9

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
jdbloom merged 1 commit into from
Apr 10, 2026
Merged

feat: learn every N steps with configurable frequency #9

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
26 changes: 12 additions & 14 deletions rl_code/Main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@
from struct import pack, unpack, Struct
import numpy as np
import math
import copy

import zmq
import csv
import os
Expand Down Expand Up @@ -53,6 +53,7 @@
if args.model_path is not None:
model_file_path = os.path.join(containing_folder, args.model_path)
learning_scheme = config['LEARNING_SCHEME']
learn_every = int(config.get('LEARN_EVERY', 1))
port = str(config['PORT'])
test_mode = args.test
train_mode = not test_mode
Expand Down Expand Up @@ -173,7 +174,6 @@
if not exp_done:
time_steps = 0

object_positions = []
agent_prox_flags = []
last_object_heading = None

Expand All @@ -183,7 +183,6 @@

# Receive initial observations from the environment
env_observations, failures, rewards, stats, robot_stats, obj_stats = Utility.parse_msgs(msgs)
object_positions.append([obj_stats[0], obj_stats[1]])
old_cyl_ang = obj_stats[5]

if Utility.params['num_obstacles'] > 0:
Expand Down Expand Up @@ -312,10 +311,6 @@
elif Utility.params['use_gate'] == 1:
gate_stats = Utility.parse_gate_stats(msgs[7])

old_object_positions = copy.deepcopy(object_positions)
object_positions.append([obj_stats[0], obj_stats[1]])


############################## gsp REWARD ##############################################
gsp_reward, label = calculate_gsp_reward(
config['GSP'],
Expand All @@ -331,9 +326,9 @@

old_cyl_ang = obj_stats[5]

old_agent_prox_flags = copy.deepcopy(agent_prox_flags)
neighbors_old_heading_gsp = copy.deepcopy(old_heading_gsp)
old_heading_gsp = copy.deepcopy(next_heading_gsp)
old_agent_prox_flags = list(agent_prox_flags)
neighbors_old_heading_gsp = old_heading_gsp.copy()
old_heading_gsp = next_heading_gsp.copy()

new_agent_states = []
force_mags = []
Expand Down Expand Up @@ -500,11 +495,14 @@
r.append(rewards[i][0])

if train_mode and config['LEARNING_SCHEME'] != 'None':
if args.independent_learning:
for i in range(Utility.params['num_robots']):
loss = models[i].learn()
if time_steps % learn_every == 0:
if args.independent_learning:
for i in range(Utility.params['num_robots']):
loss = models[i].learn()
else:
loss = model.learn()
else:
loss = model.learn()
loss = 0
else:
loss = 0

Expand Down
5 changes: 3 additions & 2 deletions run_baseline_experiments.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,14 +153,15 @@ def make_config(exp_name, gsp, neighbors, num_obstacles, use_gate, gate_curricul
"EPSILON": 1.0,
"EPS_MIN": 0.01,
"EPS_DEC": 0.0001,
"BATCH_SIZE": 64,
"BATCH_SIZE": 256,
"MEM_SIZE": 100000,
"REPLACE_TARGET_COUNTER": 1000,
"NOISE": 0.1,
"UPDATE_ACTOR_ITER": 2,
"WARMUP": 1000,
"GSP_LEARNING_FREQUENCY": 500,
"GSP_BATCH_SIZE": 128,
"LEARN_EVERY": 4,
"GSP_BATCH_SIZE": 256,
}


Expand Down
Loading