# **meeting 12/19** **Advisor: Prof. Chih-Yu Wang \ Presenter: Shao-Heng Chen \ Date: Dec 19, 2023** <!-- Chih-Yu Wang --> <!-- Wei-Ho Chung --> ## **Titles** DRL-Based Passive Beamforming in Downlink RIS-assisted mmWave MU-MISO Systems Under Hardware Impairments and Imperfect CSI Worst-case MSE minimization for RIS-assisted mmWave MU-MISO Systems with Hardware Impairments and Imperfect CSI: A Deep Reinforcement Learning Approach DRL-Based RIS Configuration in RIS-assisted mmWave MU-MISO Systems for Min-Max MSE Optimization with Hardware Impairments and Imperfect CSI ## **Experiment results** ![image](https://hackmd.io/_uploads/SJo_Fd0Ia.png) #### ```PPO-2-16-4``` ```shell -------------------------------- random action: mean: -20.310898220866918 std: 15.086675874090526 max: -0.5428093075752258 min: -94.74134063720703 shape: (220,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-4-mse: mean: -1.9433095455169678 std: 1.6839855909347534 max: -0.15906357765197754 min: -15.97573184967041 shape: (1, 220) -------------------------------- ``` <img src='https://hackmd.io/_uploads/HJS9Wg3I6.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/ryt5be28T.png' width=50% weight=50%> #### ```PPO-2-16-9``` ```shell -------------------------------- random action: mean: -19.026134564965965 std: 14.486501828473566 max: -0.9003727436065674 min: -113.47931671142578 shape: (415,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-9-mse-4: mean: -1.313175082206726 std: 0.4976070821285248 max: -0.49376580119132996 min: -4.867458343505859 shape: (1, 415) -------------------------------- ``` <img src='https://hackmd.io/_uploads/SkvYAJh8a.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/Hy2tAJ3Ua.png' width=50% weight=50%> #### ```PPO-2-16-16``` ```shell -------------------------------- random action: mean: -22.75751398229599 std: 17.16149108280305 max: -0.7223409414291382 min: -120.68282318115234 shape: (688,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-16-mse-4: mean: -1.0519797801971436 std: 0.4774532914161682 max: -0.3476378917694092 min: -4.11301326751709 shape: (1, 688) -------------------------------- ``` <img src='https://hackmd.io/_uploads/Hye4vJ286.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/H1m4wk286.png' width=50% weight=50%> #### ```PPO-2-16-25``` ```shell -------------------------------- random action: mean: -22.255270360410215 std: 17.87927955660399 max: -0.5812353491783142 min: -173.7725830078125 shape: (1039,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-25-mse: mean: -0.6489496827125549 std: 0.26560965180397034 max: -0.10836577415466309 min: -2.0491623878479004 shape: (1, 1039) -------------------------------- ``` <img src='https://hackmd.io/_uploads/rkKlqFnUT.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/SkjeqKhLa.png' width=50% weight=50%> #### ```PPO-2-16-36``` ```shell -------------------------------- random action: mean: -21.94508735051751 std: 14.160902118570835 max: -0.6340930461883545 min: -86.26004028320312 shape: (1468,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-36-mse-4: mean: -0.2223290354013443 std: 0.11829397082328796 max: -0.06411361694335938 min: -1.2988332509994507 shape: (1, 1468) -------------------------------- ``` <img src='https://hackmd.io/_uploads/By2IIIaUa.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/ByRLU8aIT.png' width=50% weight=50%> #### ```PPO-2-16-64``` ```shell -------------------------------- random action: mean: -22.27344212000072 std: 16.785474114019266 max: -0.4368678033351898 min: -128.384033203125 shape: (2560,) -------------------------------- -------------------------------- model inference of PPO-2023-11-14-2-16-64-mse: mean: -0.1619022786617279 std: 0.07155207544565201 max: -0.042396366596221924 min: -0.6167527437210083 shape: (1, 2560) -------------------------------- ``` <img src='https://hackmd.io/_uploads/rylxvq0LT.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/rJSlPcRL6.png' width=50% weight=50%> ### **Comparison of different settings** <img src='https://hackmd.io/_uploads/rJQmQuC86.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/S1PXX_0UT.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/r1WJmOCUp.png' width=50% weight=50%> <img src='https://hackmd.io/_uploads/r131muC86.png' width=50% weight=50%>