How to draw a hierarchical diagram in tikz?
Next time, please note that you should investigate solutions yourself and post an MWE which at least provides the framework of the document and, for example, the text for the nodes of your picture. (This is a lot easier than copying from an image for people who would like to help you.)
This answer uses forest to create the diagram as a tree. It uses shading to give a little depth to the nodes and the shadows library to enhance this effect. Two phantom children are used to help aligning the final nodes of the tree and the connecting lines to the first of these are added after the tree is complete since this node has four parents.
\documentclass[border=5pt]{standalone}
\usepackage{forest}
\usetikzlibrary{arrows.meta, shapes.geometric, calc, shadows}
\colorlet{mygreen}{green!75!black}
\colorlet{col1in}{red!30}
\colorlet{col1out}{red!40}
\colorlet{col2in}{mygreen!40}
\colorlet{col2out}{mygreen!50}
\colorlet{col3in}{blue!30}
\colorlet{col3out}{blue!40}
\colorlet{col4in}{mygreen!20}
\colorlet{col4out}{mygreen!30}
\colorlet{col5in}{blue!10}
\colorlet{col5out}{blue!20}
\colorlet{col6in}{blue!20}
\colorlet{col6out}{blue!30}
\colorlet{col7out}{orange}
\colorlet{col7in}{orange!50}
\colorlet{col8out}{orange!40}
\colorlet{col8in}{orange!20}
\colorlet{linecol}{blue!60}
\begin{document}
\pgfkeys{/forest,
rect/.append style={rectangle, rounded corners=2pt, inner color=col6in, outer color=col6out},
ellip/.append style={ellipse, inner color=col5in, outer color=col5out},
orect/.append style={rect, font=\sffamily\bfseries\LARGE, text width=325pt, text centered, minimum height=10pt, outer color=col7out, inner color=col7in},
oellip/.append style={ellip, inner color=col8in, outer color=col8out, font=\sffamily\bfseries\large, text centered},
}
\begin{forest}
for tree={
font=\sffamily\bfseries,
line width=1pt,
draw=linecol,
ellip,
align=center,
child anchor=north,
parent anchor=south,
drop shadow,
l sep+=12.5pt,
edge path={
\noexpand\path[color=linecol, rounded corners=5pt, >={Stealth[length=10pt]}, line width=1pt, ->, \forestoption{edge}]
(!u.parent anchor) -- +(0,-5pt) -|
(.child anchor)\forestoption{edge label};
},
where level={3}{tier=tier3}{},
where level={0}{l sep-=15pt}{},
where level={1}{
if n={1}{
edge path={
\noexpand\path[color=linecol, rounded corners=5pt, >={Stealth[length=10pt]}, line width=1pt, ->, \forestoption{edge}]
(!u.west) -| (.child anchor)\forestoption{edge label};
},
}{
edge path={
\noexpand\path[color=linecol, rounded corners=5pt, >={Stealth[length=10pt]}, line width=1pt, ->, \forestoption{edge}]
(!u.east) -| (.child anchor)\forestoption{edge label};
},
}
}{},
}
[Compressed Sensing\\Theory, inner color=col1in, outer color=col1out
[Projection Matrix\\Theory, inner color=col2in, outer color=col2out
[Optimise Projection\\Matrix, inner color=col4in, outer color=col4out]
[Reduce Number of\\Measurements, inner color=col4in, outer color=col4out]
]
[Reconstruction\\Algorithms, inner color=col3in, outer color=col3out
[Convex Relaxation
[Sparse Signal\\Estimate, rect, name=sse1
]
]
[Greedy Pursuits
[Sparse Signal\\Estimate, rect, name=sse2
]
]
[, phantom, calign with current
[A\\B, phantom
[Our Work, orect, name=us
[{Improved Sparse Signal Estimate!}, oellip
]
]
]
]
[Non-Convex\\Minimisation Methods
[Sparse Signal\\Estimate, rect, name=sse3
]
]
[Combinatorial\\Algorithms
[Sparse Signal\\Estimate, rect, name=sse4
]
]
]
]
\begin{scope}[color=linecol, rounded corners=5pt, >={Stealth[length=10pt]}, line width=1pt, ->]
\draw (sse2.south) -- (us.north -| sse2.south);
\draw (sse3.south) -- (us.north -| sse3.south);
\coordinate (c1) at ($(sse1.south)!2/5!(sse2.south)$);
\coordinate (c2) at ($(sse3.south)!2/5!(sse4.south)$);
\draw (sse1.south) -- +(0,-10pt) -| (us.north -| c1);
\draw (sse4.south) -- +(0,-10pt) -| (us.north -| c2);
\end{scope}
\end{forest}
\end{document}
Have a look at here: Texample - Simple Flow chart
Needed libraries, as in this example are \usetikzlibrary{shapes,arrows}.
Edit: Taken from Texample:
\usepackage{tikz}
\usetikzlibrary{shapes,arrows}
\begin{document}
% Define block styles
\tikzstyle{decision} = [diamond, draw, fill=blue!20, text width=4.5em, text badly centered, node distance=3cm, inner sep=0pt]
\tikzstyle{block} = [rectangle, draw, fill=blue!20, text width=5em, text centered, rounded corners, minimum height=4em]
\tikzstyle{line} = [draw, -latex']
\tikzstyle{cloud} = [draw, ellipse,fill=red!20, node distance=3cm, minimum height=2em]
\begin{tikzpicture}[node distance = 2cm, auto]
% Place nodes
\node [block] (init) {initialize model};
\node [cloud, left of=init] (expert) {expert};
\node [cloud, right of=init] (system) {system};
\node [block, below of=init] (identify) {identify candidate models};
\node [block, below of=identify] (evaluate) {evaluate candidate models};
\node [block, left of=evaluate, node distance=3cm] (update) {update model};
\node [decision, below of=evaluate] (decide) {is best candidate better?};
\node [block, below of=decide, node distance=3cm] (stop) {stop};
% Draw edges
\path [line] (init) -- (identify);
\path [line] (identify) -- (evaluate);
\path [line] (evaluate) -- (decide);
\path [line] (decide) -| node [near start] {yes} (update);
\path [line] (update) |- (identify);
\path [line] (decide) -- node {no}(stop);
\path [line,dashed] (expert) -- (init);
\path [line,dashed] (system) -- (init);
\path [line,dashed] (system) |- (evaluate);
\end{tikzpicture}
\end{document}
Edit 2: In the specific case my code looks like that.
\usepackage{tikz}
\usetikzlibrary{shapes,arrows,trees,calc}
\begin{document}
% Define block styles
\tikzstyle{block} = [rectangle, draw, text width=2.5cm, text centered, rounded corners, minimum height=4em,fill=blue!20]
\tikzstyle{line} = [draw,thick, -latex']
\tikzstyle{cloud} = [draw, ellipse, text width=2.5cm, text centered]
\tikzstyle{edge from parent}=[->,thick,draw]
\begin{tikzpicture}[auto,edge from parent fork down]
\tikzstyle{level 1}=[sibling distance=180mm,level distance=18ex]
\tikzstyle{level 2}=[sibling distance=55mm,level distance=24ex]
% Place nodes
\node [cloud,fill=red!40] (cst) {Compressed Sensing Theory}
child{node [cloud,fill=green!40] (pmt) {Projection Matrix Theory}
child{node [cloud,fill=green!20] (opm) {Optimize Projection Matrix}}
child{node [cloud,fill=green!20] (rnm) {Reduce Number of Measurements}}
}
child{node [cloud,fill=blue!40] (ra) {Reconstruction Algorithms}
child{node [cloud,fill=blue!20] (cr) {Convex Relaxation}}
child{node [cloud,fill=blue!20] (gp) {Greedy Pursuit}}
child{node [cloud,fill=blue!20] (ncmm) {Non-Convex Minimization Methods}}
child{node [cloud,fill=blue!20] (ca) {Combinatorial Algorithms}}
};
\node[block,yshift=-1cm,below of= cr](sse1){Sparse Signal Estimate};
\node[block,yshift=-1cm,below of= gp](sse2){Sparse Signal Estimate};
\node[block,yshift=-1cm,below of= ncmm](sse3){Sparse Signal Estimate};
\node[block,yshift=-1cm,below of= ca](sse4){Sparse Signal Estimate};
\node[below of= ra,yshift=-8cm,rectangle, draw, text width=16cm, text centered, rounded corners, minimum height=3em,fill=orange!40](ow){Our Work};
\node[below of= ow,yshift=-1cm,rectangle, draw, text width=8cm, text centered, rounded corners, minimum height=3em,fill=orange!20](isse){Improve Sparse Signal Estimate};
%% Draw edges
\path [line] (cr.south) -- (sse1.north);
\path [line] (gp.south) -- (sse2.north);
\path [line] (ncmm.south) -- (sse3.north);
\path [line] (ca.south) -- (sse4.north);
\path [line] (sse1)--($(sse1.south)-(0,0.5)$) -| (ow.175);
\path [line] (sse2)--($(sse2.south)-(0,0.5)$) -| (ow.150);
\path [line] (sse3)--($(sse3.south)-(0,0.5)$) -| (ow.30);
\path [line] (sse4)--($(sse4.south)-(0,0.5)$) -| (ow.5);
\path [line] (ow.south) -- (isse.north);
\end{tikzpicture}
\end{document}
