In [13]:
# Import necessary tools
import pandas as pd
import matplotlib.pyplot as plt
import scipy as sp
import os
In [14]:
# Set environment variables for bash
WD = os.getcwd()
SRC_DIR = WD + "/.."
BUILD_DIR = WD + "/../../build"
GRAPH_DIR=BUILD_DIR + "/statistics/graphs" # Where we will place the graph plots
RES_IN=BUILD_DIR + "/experiment/res.csv"
%env SRC_DIR=$SRC_DIR
%env BUILD_DIR=$BUILD_DIR
%env GRAPH_DIR=$GRAPH_DIR
%env CSV_DIR=$CSV_DIR
%env RES_IN=$RES_IN
<git's root>/build/experiment instead, in which case the extracting will be ignored.
In [15]:
%%bash
if [ ! -e "$BUILD_DIR/experiment" ]
then
ARCHIVE="$SRC_DIR/statistics/results.tbz"
mkdir -p "$BUILD_DIR"
mkdir -p "$GRAPH_DIR"
tar -xjf "$ARCHIVE" -C "$BUILD_DIR"
fi
In [16]:
INDEX_NAMES = ["Protocol", "Topology", "Packet drop rate", "Num. robots"]
COLUMN_NAMES=["Consensus time", "Num. tx entries", "Num. rx entries", "Mean tx bandwidth", "Mean rx bandwidth"]
PROTOCOLS=["consensus"] # Our experiments only used the 'consensus' protocol, i.e.,
# placing all the robots and waiting for consensus to be reached.
TOPOLOGIES=["line", "cluster", "scalefree"]
DROP_RATES=[0, 0.25, 0.5, 0.75]
In [17]:
# Gets results' data and renames the axes
def readResults(filename):
df = pd.read_csv(filename, index_col=[0,1,2,3])
df.index.names = INDEX_NAMES
df.columns = COLUMN_NAMES
return df
In [18]:
# Gives statistical data for each {protocol, topology, drop rate, num robots}
# configuration about the specified column.
def crunchColumnByConfig(df, columnName):
# Get column's data
ret = df.xs(columnName, axis=1)
# Then group experiments by configuration
ret = ret.groupby(level=[0,1,2,3])
# Then do some pandas magic stuff
ret = ret.apply(pd.Series.reset_index, drop=True).unstack().transpose().describe().transpose()
return ret
In [19]:
data = readResults(RES_IN)
consensusData = crunchColumnByConfig(data, "Consensus time")
consensusData
Out[19]:
In [20]:
def plotGraph(df, topology, formats, deltas, yscale="linear", xlabel="Number of robots", ylabel="", savefileBaseName=None):
fig = plt.figure(figsize = (10,5))
axis = fig.add_subplot(111)
topologyDf = df.xs(topology, level=1)
plotNumber=0
for protocol in PROTOCOLS:
for dropRate in DROP_RATES:
currDf = topologyDf.xs((protocol, dropRate))
numsRobots = currDf.index.tolist()
numsRobots = [numsRobots[i] + deltas[plotNumber] for i in range(len(numsRobots))]
yPlot = currDf.xs("50%", axis=1)
yError = [(yPlot - currDf.xs("min", axis=1)),
(currDf.xs("max", axis=1) - yPlot)]
axis.errorbar(numsRobots, yPlot, yerr = yError, fmt=formats[plotNumber] + "-")
plotNumber += 1
axis.set_xlabel(xlabel)
axis.set_ylabel(ylabel)
axis.set_yscale(yscale)
axis.yaxis.grid()
axis.legend([str(drop*100)+"% drop" for drop in DROP_RATES], loc=0, ncol=1, title=(topology + " topology"))
if savefileBaseName != None:
plt.savefig(GRAPH_DIR+"/"+savefileBaseName+".png", dpi=600, format="png", transparent=False)
In [21]:
%matplotlib inline
# Set variables
CONSENSUS_YLABEL="Consensus Time (timesteps)"
DELTAS=[0, 0, 0, 0]
FORMATS=["ro", "go", "bo", "mo"]
# Plot graphs
plotGraph(consensusData, "line", FORMATS, DELTAS, ylabel=CONSENSUS_YLABEL, savefileBaseName="lineConsensus")
plotGraph(consensusData, "line", FORMATS, DELTAS, yscale="log", ylabel=CONSENSUS_YLABEL + " [log]", savefileBaseName="lineConsensus_log")
plotGraph(consensusData, "cluster", FORMATS, DELTAS, ylabel=CONSENSUS_YLABEL, savefileBaseName="clusterConsensus")
plotGraph(consensusData, "cluster", FORMATS, DELTAS, yscale="log", ylabel=CONSENSUS_YLABEL + " [log]", savefileBaseName="clusterConsensus_log")
plotGraph(consensusData, "scalefree", FORMATS, DELTAS, ylabel=CONSENSUS_YLABEL, savefileBaseName="scalefreeConsensus")
plotGraph(consensusData, "scalefree", FORMATS, DELTAS, yscale="log", ylabel=CONSENSUS_YLABEL + " [log]", savefileBaseName="scalefreeConsensus_log")
In [ ]: