Algorithm Introduction
You can choose language by
This project provides 3 types of environment implementations, and 3 baseline algorithms for users to design algorithms and compare their performance. For detailed documentation on environment classes, see: Environment Wrapping; for detailed documentation on algorithm classes, see: Model Construction
Quick Start (Running baseline algorithms)
Baseline1: BenchEnv_Multi environment + A3CLSTM-E2E algorithm
Environment code is located at
Alg_Base/DAT_Benchmark/envs/environment.py
Algorithm code is located at Alg_Base/DAT_Benchmark/models/A3CLSTM_E2E/
You can run the following code to quickly start the algorithm:
cd Alg_Base/DAT_Benchmark/
# Test mode
# Test using Cumulative Reward (CR)
python ./models/A3CLSTM_E2E/main.py --Mode 0 --Scene "citystreet" --Weather "day" --delay 20 --Test_Param "CityStreet-d" --Test_Mode AR
# Test using Tracking Success Rate (TSR)
python ./models/A3CLSTM_E2E/main.py --Mode 0 --Scene "citystreet" --Weather "day" --delay 20 --Test_Param "CityStreet-d" --Test_Mode TSR
# New training mode
python ./models/A3CLSTM_E2E/main.py --Mode 1 --workers 35 --Scene "citystreet" --Weather "day" --delay 20 --Freq 125 --New_Train
# Resuming training mode
python ./models/A3CLSTM_E2E/main.py --Mode 1 --workers 35 --Scene "citystreet" --Weather "day" --delay 20 --Freq 125
Baseline2: UAV_VAT_Gymnasium environment + D-VAT algorithm
Environment code can be found at
Alg_Base/DAT_Benchmark/models/D_VAT/DVAT_envs.py
Algorithm code can be found at Alg_Base/DAT_Benchmark/models/D_VAT/
You can run the following code to quickly start the algorithm:
cd Alg_Base/DAT_Benchmark/
# Test mode
# Test using Cumulative Reward (CR)
python ./models/D_VAT/DVAT_main.py -w 1 -m citystreet-day.wbt --train_mode 0 --Test_Mode CR
# Test using Tracking Success Rate (TSR)
python ./models/D_VAT/DVAT_main.py -w 1 -m citystreet-day.wbt --train_mode 0 --Test_Mode TSR
# New training mode
python ./models/D_VAT/DVAT_main.py -w 35 -m citystreet-day.wbt --train_mode 1 --New_Train
# Resume training mode
python ./models/D_VAT/DVAT_main.py -w 35 -m citystreet-day.wbt --train_mode 1
Baseline3(Ours): Envs Environment + R-VAT Algorithm
Environment code can be found at
Alg_Base/DAT_Benchmark/envs/envs_parallel.py
Algorithm code can be found at Alg_Base/DAT_Benchmark/models/R_VAT/
You can run the following code to quickly start the algorithm:
cd Alg_Base/DAT_Benchmark/
# Test mode
# Test using Cumulative Reward (CR)
python ./models/R_VAT/RVAT.py -w 1 -m citystreet-day.wbt --train_mode 0 --Test_Mode AR
# Test using Tracking Success Rate (TSR)
python ./models/R_VAT/RVAT.py -w 1 -m citystreet-day.wbt --train_mode 0 --Test_Mode TSR
# New training mode
python ./models/R_VAT/RVAT.py -w 35 -m citystreet-day.wbt --train_mode 1 --New_Train
# Resume training mode
python ./models/R_VAT/RVAT.py -w 35 -m citystreet-day.wbt --train_mode 1