Reached the end of the main body

documentation/conclusion.tex

@@ -0,0 +1,72 @@
+\chapter{Conclusion}
+
+\section{Results}
+
+In this section we will summarize our most important achievements during the
+work on the project. First, we managed to implement robust detection
+algorithms, on which we could rely when we worked on higher-lever behaviors.
+During our tests, there were almost no false detections, i.e.\ foreign objects
+were not detected as a ball or a goal. Sometimes the ball and the goal were
+missed, even if they were in the field of view, which happened due to imprecise
+color calibration under changing lighting conditions. The goal detection was on
+of the most difficult project milestones, so we are particularly satisfied with
+the resulting performance. It is worth mentioning, that with the current
+algorithm, for successful detection, it is not even necessary to have the whole
+goal in the camera image.
+
+Another important achievement is the overall system robustness. In our tests
+the robot could successfully reach the ball, do the necessary alignments and
+kick the ball. When the robot decided that he should kick the ball, in the
+majority of cases the kick was successful and the ball reached the target. We
+performed these tests from many starting positions and assuming many relative
+position of the ball and the goal. Naturally, we put some constraints on the
+problem, but within th \todo{smth about constraints and such bullshit}.
+
+Furthermore, we managed not only to make the whole approach robust, but also
+worked on making the procedure fast, and the approach planing was a crucial
+element of this. In the project's early stages, the robot couldn't approach the
+ball from the side, depending on the goal position, and instead always walked
+towards the ball directly and aligned to the goal afterwards. The tests have
+shown, that in such configuration the goal alignment was actually the longest
+phase and could take over a minute. Then we introduced the approach planing,
+and as a result the goal alignment stage could in many scenarios be completely
+eliminated, which was greatly beneficial for the execution times.
+
+Finally, \todo{the kick was nice}.
+
+\section{Future Work}
+
+With our objective for this semester completed, there still remains a vast room
+for improvement. Some of the most interesting topics for future work will now
+presented.
+
+The first important topic is the self-localization. Currently our robot is
+completely unaware of his position on the field, but if such information could
+be obtained, then it could be leveraged to make path planning more effective
+and precise.
+
+Another important capability that our robot lacks for now is obstacle
+awareness, which would be unacceptable in the real RoboCup soccer game. Making
+the robot aware of the obstacles on the field would require the obstacle
+detection to be implemented, as well as some changes to the path planning
+algorithms to be made, which makes this task an interesting project on its own.
+
+A further capability that could be useful for the striker is the ability to
+perform different kicks depending on the situation. For example, if the robot
+could perform a sideways kick, then the goal alignment would in many situations
+be unnecessary, which would reduce the time needed to score a goal.
+
+In this semester we concentrated on the ``free-kick'' situation, so our robot
+can perform its tasks in the absence of other players, and only when the ball
+is not moving. Another constraint that we imposed on our problem is that the
+ball is relatively close to the goal, and that the ball is closer to the goal
+than the robot, so that the robot doesn't have to run away from the goal. To be
+useful in a real game the striker should be able to handle more complex
+situations. For example, the \textit{dribbling} skill could help the robot to
+avoid the opponents and to bring the ball into a convenient striking position.
+
+Finally, we realized that the built-in moving functions in NAOqi SDK produce
+fairly slow movements, and also don't allow to change the direction of movement
+fluently, which results in pauses when the robot needs to move in another
+direction. This realization brings us to thought, that the custom-implemented
+movement might result in much faster and smoother behavior.

documentation/overview.tex

@@ -0,0 +1,18 @@
+\section{Strategy Overview}
+
+Now that all of the milestones are completed, we will present a short overview
+of the whole goal scoring strategy, the block diagram of which can be found in
+\todo{learn to do figures and reference them}. At the very beginning the robot
+will detect the ball and turn to ball, as described in \todo{where}. After
+that, the distance to the ball will be calculated, the goal will be detected,
+and the direction to goal will be determined. If the ball is far away
+\textit{and} the ball and the goal are strongly misaligned, then the robot will
+try to approach the ball from the appropriate side, otherwise the robot will
+approach the ball directly. These approach steps will be repeated until the
+robot is close enough to the ball to start aligning to the goal, but in the
+practice one step of approach from the side followed by a short direct approach
+should suffice. When the ball is close, the robot will check if it is between
+the goalposts, and will perform necessary adjustments if that's not the case.
+After the ball and the goal are aligned, the robot will align its foot with
+the ball and kick the ball. For now, we assumed that the ball will reach the
+goal and so the robot can finish execution.

documentation/perception.tex

@@ -38,9 +38,61 @@ goal is white, and there are generally many white areas in the image from the
 robot camera, which have area larger than that of the image of the goal, for
 example the white field lines and the big white wall in the room with the
 field. To deal with the multitude of the possible goal candidates, we
-propose the following algorithm.
+propose the following heuristic algorithm.
 
 First, all contours around white areas are extracted by using a procedure
-similar to that described in the section on ball detection. Only $N$ contours
+similar to that described in the section on ball detection. Next, the
-with the largest areas are considered further (in our experiments it was
+\textit{candidate preselection} takes place. During this stage only $N$
-empirically determined that $N=5$ provides good results).
+contours with the largest areas are considered further (in our experiments it
+was empirically determined that $N=5$ provides good results). Furthermore, all
+convex contours are rejected, since the goal is a highly non-convex shape.
+After that, a check is performed, how many points are necessary to approximate
+the remaining contours. The motivation behind this is the following: it is
+clearly visible that the goal shape can be perfectly approximated by a line
+with 8 straight segments. On an image from the camera, the approximation is
+almost perfect when using only 6 line segments, and in some degenerate cases
+when the input image is noisy, it might be necessary to use 9 line segments to
+approximate the shape of the goal. Any contour that requires a different number
+of line segments to be approximated is probably not the goal. The preselection
+stage ends here, and the remaining candidates are passed to the scoring
+function.
+
+The scoring function calculates, how different are the properties of the
+candidates are from the properties, an idealized goal contour is expected to
+have. The evaluation is happening based on two properties. The first property
+is based on the observation, that the area of the goal contour is much smaller
+than the area of its \textit{enclosing convex hull}. The second observation is
+that all points of the goal contour must lie close to the enclosing convex
+hull. The mathematical formulation of the scoring function looks like the
+following \todo{mathematical formulation}:
+
+The contour, that minimizes the scoring function, while keeping its value under
+a certain threshold is considered the goal. If no contour scores below the
+threshold, then the algorithm assumes that no goal was found. Our tests have
+shown, that when the white color is calibrated correctly, the algorithm can
+detect the goal almost without mistakes, when the goal is present in the image.
+The downside of this algorithm, is that in some cases the field lines might
+appear the same properties, that the goal contour is expected to have,
+therefore the field lines can be mistaken for the goal. We will describe, how
+we dealt with this problem, in the following section.
+
+\section{Field detection}
+
+The algorithm for the field detection is very similar to the ball detection
+algorithm, but some concepts introduced in the previous section are also used
+here. This algorithm extracts the biggest green area in the image, finds its
+enclosing convex hull, and assumes everything inside the hull to be the field.
+
+Such rather simple field detection has two important applications. The first
+one is that the robot should be aware, where the field is, so that it doesn't
+try to walk away from the field. Due to time constraints, we didn't implement
+this part of the behavior. The second application of field detection is the
+improvement of the quality of goal and ball recognition. As was mentioned in
+the section on ball detection, the current algorithm might get confused, if
+there are any red objects in the robot's field of view. However, there
+shouldn't be any red objects on the field, except the ball itself. So, if
+everything that's not on the field is ignored, when trying to detect the ball,
+the probability of identifying a wrong object decreases. On the other hand, the
+problem with the goal detection algorithm was that it could be distracted by
+the field lines. So, if everything on the field is ignored for goal
+recognition, then the accuracy can be improved.

documentation/robotum_report.tex

@@ -7,22 +7,16 @@
 \usepackage[hidelinks]{hyperref}
 \usepackage{glossaries}
 
-
-
 \newcommand{\fig}{figures/}
 \usepackage{graphicx}
 \usepackage{tikz}
 \usetikzlibrary{quotes,angles}
 
-
-
 % lots of packages are included in the preamble, look there for more
 % information about how this.
 
 \include{robotum_report.preamble}
 
-
-
 % if you don't know where something can be found, click on the pdf, and
 % Overleaf will open the file where it is described
 
@@ -31,14 +25,12 @@
   \vspace*{6mm}
 }
 
-\author{Pavel Lutskov\\Jonas Bubenhagen\\Yuankai Wu\\Seif Ben Hamida\\ Ahmed Kamoun}
+\author{Pavel Lutskov\\Jonas Bubenhagen\\Yuankai Wu\\Seif Ben Hamida\\Ahmed Kamoun}
-\supervisors{Mohsen Kaboli\\ and the Tutor (insert name)}
+\supervisors{Mohsen Kaboli\\and the Tutor (insert name)}
 \submitdate{August 2018}
 
 \maketitle % this generates the title page. More in icthesis.sty
 
-
-
 \preface
 % \input{Acknowledgements/Acknowledgements}
 
@@ -46,31 +38,14 @@
 
 % \input{Introduction/Introduction}
 
-
 \setstretch{1.2} % set line spacing
 \input{introduction}
 \input{tools}
 \input{solintro}
 \input{perception}
 \input{jonas}
-
+\input{overview}
-% body of thesis comes here
+\input{conclusion}
-% \input{Body/SoftwareTools} %this loads the content of file SoftwareTools.tex
-% in the folder Body.
-% \todo{SoftwareTools} %this is how you add a todo, it will appear in the list
-% on page 2, and in orange in the margin where you add it.
-% \input{Body/Setup}
-% \todo{Setup}
-% \input{Body/Perception}
-% \todo{Perception}
-% \input{Body/Modeling}
-% \todo{Modeling}
-% \input{Body/Motion}
-
-% \input{Body/Behavior}
-
-
-% \input{Conclusion/Conclusion}
 
 \begin{appendices}
 %\input{Appendix/BehaviorImplementation}