running nao test simulations

nao-guy

Run the following in different tabs in terminal

MDP Constant
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.02 -p 1000 -a 1 -r MDP -o 49234 -g 0.1 -u 0.05 -mean 0.005 -std 1.0

MDP Relative
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.02 -p 1000 -a 1 -r MDP -m relative -o 49234 -g 0.1 -u 0.05 -mean 0.005 -std 1.0

Human
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.02 -p 1000 -a 1 -r Human -o 49234 -g 0.1 -u 0.05 -mean 0.005 -std 1.0

MDP Constant and Human
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.02 -p 1000 -a 1 -r MDPandHuman -o 49234 -g 0.1 -u 0.05 -mean 0.005 -std 1.0


MDP Relative, Simulated EMG, Real Robot
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.002 -p 1000 -a 1 -r MDP -m relative -n -g 0.1 -u 0.05 -mean 0.005 -std 1.0

MDP Relative, Real EMG, Real Robot 
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.002 -p 1000 -a 1 -r MDP -m relative -n -e -g 0.1 -u 0.05 -mean 0.005 -std 1.0

MDP Constant + Human, Real EMG and Real Robot 
/usr/bin/python blueberry_move_and_emg.py -t 40000 -s 0.002 -p 1000 -a 1 -r MDPandHuman -n -e -g 0.1 -u 0.05 -mean 0.005 -std 1.0


This open call could be used to run a parameter sweep with the average reward of the last 5000 steps as a comparison.
Turn off live plotting all together? This can be done by changing the -p to the tmax! That is one solution. 

usage: blueberry_move_and_emg.py [-h] [-t TMAX] [-s SLEEP_TIME]
                                 [-p PLOT_EVERY_N] [-a PERSIST_TIME_STEPS]
                                 [-n] [-e] [-r REWARD_TYPE]
                                 [-m MDP_REWARD_TYPE] [-o PORT]
                                 [-g ANGLE_ERROR_THRESHOLD]
                                 [-u ANGLE_UPDATE_THRESHOLD]
                                 [-mean ALPHAU_CONST] [-std ALPHAS_CONST]

optional arguments:
  -h, --help            show this help message and exit
  -t TMAX               Maximum number of iterations. Default is 4000.
  -s SLEEP_TIME         How long does the agent rest for in ms, set speed of
                        algorithm. Default is 0.025, or about 30Hz.
  -p PLOT_EVERY_N       Update the plot every N iterations. Default is 10.
  -a PERSIST_TIME_STEPS
                        Actions should persist for N iterations. Default is 3.
  -n                    Use this flag to use the a physical robot. Default is
                        simulated robot.
  -e                    Use this flag to use the a physical EMG signal.
                        Default is simulated EMG signal.
  -r REWARD_TYPE        Define the reward type for the learning agent, can
                        choose MDP, Human, or MDPandHuman. Default is MDP.
  -m MDP_REWARD_TYPE    Define the reward calculation for the learning agent
                        in the MDP. Can be constant (default) or relative.
  -o PORT               Define the simulation robot port.
  -g ANGLE_ERROR_THRESHOLD
                        Define the angular error threshold. Default is 0.05.
  -u ANGLE_UPDATE_THRESHOLD
                        Define the angular update threshold. Default is 0.05.
  -mean ALPHAU_CONST    Define the mean step size of the actor. Default is
                        0.0025.
  -std ALPHAS_CONST     Define the step size fraction of the mean step size of
                        the standard deviation of the actor. Default is 0.5
                        
Running the new blueberry bot
cd ~/Dropbox/School/Graduate/CMPUT/projects/NAO/scripts_mac/blueberry/pynaoqi
/usr/bin/python blueberry_move_and_emg.py

Running with options:
/usr/bin/python blueberry_move_and_emg.py -t 10000 -s 0.025 -p 10 -a 1 -m relative -o 51737 -g 0.05 -u 0.05

cd /Applications/Choregraphe.app/Contents/Resources/bin/
./naoqi-bin

cd /Applications/Choregraphe.app/Contents/Resources/
./choregraphe

cd ~/Dropbox/School/Graduate/CMPUT/projects/NAO/scripts_mac
python kory_movement_live_plot.py

cd ~/Dropbox/School/Graduate/CMPUT/projects/NAO/scripts_mac
python kory_emg_good.py