Major upgrade to get Q-learning to work successfully - #18

Merged

eisDNV merged 9 commits into

mainfrom

Aug 28, 2026

Commits on Aug 25, 2026

Major upgrade to get Q-learning to work successfully
eisDNV
committed
a028b57

Commits on Aug 27, 2026

Corrected a probem with the r_limit argument of 'analyse_trained' in 'use_q_ide.py' the problem was related to a quick and dirty fix during training.
eisDNV
committed
731ffdc
Corrected a problem in set_state() of controlled_crane_pendulum.py, where the orthogonality check used the wron variable. The related section of the code (w_speed is ndarray) has so fra not been used.
eisDNV
committed
3508d14
Removed trend reporting from dump_results() of q_agent.py. Trend reporting is now done from episode-runs on trained data. The removed code was in any case inaccessible, due to changes in reward_stats.
eisDNV
committed
888f521
fix(q_agent): read reward_stats as dict in analyse_training()
aleksandarbabicdnv
committed
055b417
fix(use_q_ide): pass r_limit as keyword in analyse_all()
aleksandarbabicdnv
committed
800983c
fix(controlled_crane_pendulum): add missing f-prefix to set_state() error
aleksandarbabicdnv
committed
7a68abb
style: apply ruff format (missing spaces around | and ,)
aleksandarbabicdnv
committed
a6ce4a0
fix(ppo_agent): decouple reward plot from env.reward_stats
aleksandarbabicdnv
committed
0d4e18f