November 16, 2017

Computational Geometry - Part 1

0 comments
I've spent some time studying an area called Computational Geometry, which according to Wikipedia is "a branch of computer science devoted to the study of algorithms which can be stated in terms of geometry." Examples include: Are two lines intersecting? How can you triangulate random points? If you are interested in the area, the best book I found was Computational Geometry: Algorithms and Applications. The algorithms I've studied are:

Find the convex hull of random points. If you have some points and you want to find the smallest possible area that includes these points, then you need to find the convex hull of those points.



Triangulate polygons. If you have a convex polygon or a concave polygon, how can you triangulate them? 



Triangulate points. If you have random points, how can you triangulate those?



Delaunay triangulation. If you look at the images above, the triangles look kinda odd. To improve the triangulation you can use an algorithm called Delaunay triangulation.


Voronoi diagram. If you run a stores in different cities, then you can create a Voronoi diagram to find out where those customers live that live the closest to a specific store.


That's it for part 1. Next part will include boolean operations on polygons!

November 11, 2017

Books about engineering you can read on the beach

0 comments
Engineering is fun and it's also fun to read about other engineers so you can learn something from their mistakes and find inspiration. Engineering is also about math, physics, etc, but this is a collection of books without any math that you can read on the beach. So leave your calculator behind!


A Thread Across the Ocean: The Heroic Story of the Transatlantic Cable. One of the big engineering projects that took place in the 1800's was to connect Europe with America by dragging a 2500 tonnes heavy telegraph cable from one continent to the next. If that was possible, it would take just a minute to send information between the continents, compared to the weeks it took to send the same information with a ship. How this was possible is explained in the book.

Structures: Or Why Things Don't Fall Down. I've always been fascinated by those engineers who worked without calculators. How could they build a ship like Titanic without computers? This book will tell you the history of how to engineer buildings, bridges, ships, and much more.

Fermat's Enigma: The Epic Quest to Solve the World's Greatest Mathematical Problem. Math and engineering are deeply connected. This book will tell you the story of when one guy tried to prove Fermat's last theorem. It will also tell you the history of mathematics.

Clouds of Glory: The Life and Legend of Robert E. Lee. This book tells the story of one of the famous generals from the American Civil War. But little is known that Robert E. Lee was also an engineer, which is maybe the reason he was successful as a general? The book includes a lot about the war, but also about the engineering challenges to build a fort and change the flow of a river.

Wizard: The Life and Times of Nikola Tesla: Biography of a Genius. Nikola Tesla was a famous inventor, and one of Elon Musk's role models, but the problem was that he never made money from his inventions so he died poor. This book will teach you why great engineering is not the end goal - make money is always the goal.

The Martian. What if you are on Mars, then an accident happens which forces your team to leave you behind, so you have to survive with what you have until someone can rescue you? This is of course a made up story, but the author worked as an engineer before he began writing books so the book is realistic.

Alan Turing: The Enigma. One of the great engineering challenges during the Second World War was to break the German codes. He wasn't alone, but one of the persons who accomplished this was Alan Turing, and this is the best biography on him.

Moon Lander: How We Developed the Apollo Lunar Module. It might have been difficult to drag a telegraph cable across the Atlantic Ocean without the help of computers, but what if you want to land on the Moon with only primitive computers? This book is written by the engineer who was responsible for designing the vehicle that actually landed on the Moon, so you will learn all about it.

The Quest for Artificial Intelligence. Artificial Intelligence or smart computers are becoming an increasingly larger part of our life, such as self-driving cars. The area is actually not new, people have always tried to build smart machines, and this book will tell you all about it.

The Innovators: How a Group of Hackers, Geniuses, and Geeks Created the Digital Revolution. This book will tell you the history of the computer - from the first attempt to build a calculator up to the guys who designed Google. It is written by the same author who wrote the best Steve Jobs biography, so you should read both!

The Boy Who Harnessed the Wind: Creating Currents of Electricity and Hope. What if you grew up in a poor place in Africa and decided you wanted to help the community to get electricity. This book will tell you how William Kamkwamba built a wind turbine from scrap.

Skunk Works: A Personal Memoir of My Years at Lockheed. One of the most recognizable planes is the F-117 which was the first plane that was hard to detect on a radar - a so called stealth fighter. This book is written by one of the engineers who built that plane and other innovative aircraft at the company Skunk Works.

Masters of Doom: How Two Guys Created an Empire and Transformed Pop Culture. One of the most famous computer games is Doom, and this book will tell you the story of how Doom was created.

And finally some shameless self-promotion. If you have read all books about engineering you should also read my book about engineering, which is a biography on Elon Musk: The Engineer: Follow Elon Musk on a journey from South Africa to Mars.

November 4, 2017

Would Leonardo da Vinci have streamed his work online?

0 comments
Television is 20th century technology and the new way to consume entertainment is to watch streamers on services like Twitch. A streamer is someone who is showing him/herself through a web camera while the person is doing something like playing a game, drawing a picture, or even looking for zebras in South Africa. Why? Traditional television has a broader audience, so the television shows tend to become kinda dull. Watching a stream on Twitch is more personal and you can even interact with the person through a chat. Streamers are also calming background noise, and I was listening to an art streamer while writing this article.

But watching someone else doing art is not something new. I've recently read a biography on Leonardo Da Vinci, written by Walter Isaacson - the same author who wrote the official Steve Jobs biography. He has also written a biography on Albert Einstein, and I've read all three of those books. In the book on Leonardo da Vinci, it's revealed that people were watching him when he was doing art, while he was saying "Drawing in company is much better than alone."
When Leonardo da Vinci was painting The Last Supper, spectators would visit and sit quietly just so they could watch him work. The creation of art, like the discussion of science, had become at times a public event. According to the account of a priest, Leonardo would "come here in the early hours of the morning and mount the scaffolding," and then "remain there brush in hand from sunrise to sunset, forgetting to eat and drink, painting continually." On other days, however, nothing would be painted. "He would remain in front of it for one or two hours and contemplate it in solitude, examining and criticizing to himself the figures he had created." Then there were dramatic days that combines his obsessiveness and his penchant for procrastination. As if caught by whim or passion, he would arrive suddenly in the middle of the day, "climb the scaffolding, seize a brush, apply a brush stroke or two to one of the figures, and suddenly depart."

Here are some other lessons learned from the book on Leonardo da Vinci:
  • You have to observe the real world and this is more important than learning from someone else:
    • If you look around you, you will not see any sharp lines, so why should you paint using sharp lines? "Paint so that a smokey finish can be seen, rather than contours and profiles that are distinct and crude."
    • Use shadows, not lines, and this is the secret to modeling 3d objects on a 2d surface. Leonardo da Vinci spent more time studying shadows (and thus light) than he did on any other artistic topic.   
    • Leonardo da Vinci took this one step further by dissecting human bodies to really learn how to make better paintings. Art and science is interwoven. If you paint someone, you should begin with the skeleton, then the skin, and finally add the clothing.
    • Study the movement of bodies. Many say that Leonardo da Vinci's characters are moving, and he said that "movements should announce the motions of the mind." If he met someone on the street with an appearance he wanted to study closer, he invited the person over for supper. 
    • To remember all observations, he always brought a notebook with him. In these books he wrote what he observed and what he wanted to observe, such as "Describe the tongue of the woodpecker."
  • He failed a lot. In some cases it could take 20 years before he finished a painting, and in some cases he never finished a painting he was working on. Leonardo da Vinci wanted is art to be flawless, so he could never finish them because they were so complicated if you took the lighting into account. In one case he had to dissect a body to learn how a muscle worked before he was satisfied with the painting.
  • Combine fantasy with observation. Leonardo da Vinci is famous for coming up with futuristic machines, like helicopters and tanks. But it turned out most of these machines were not meant to function because he created those machines for theatrical plays. If you want to draw a dragon, it's easier to combine parts from other animals, like the head from a dog, the eyes from a cat, the ears from a pig, and so on.
  • Procrastinate is not always bad. Leonardo da Vinci was having a discussion on how creativity occurs. Sometimes it requires going slowly, pausing, even procrastinating. "Men of lofty genius sometimes accomplish the most when they work least for their minds are occupied with their ideas and the perfection of their conceptions, to which they afterwards give form."

October 1, 2017

Evil pumpkins - or how to simulate subsurface scattering

0 comments
If you are making a game, it's really important that the game looks good and that the game runs as fast as possible on the computer. It's often difficult to achieve both good looks and speed, so you have to cheat by using various tricks. One of the cheats I read about was how to achieve fast subsurface scattering in Unity. The idea behind subsurface scattering is that light tends to shine through some materials. For example, if you are holding up your hand against a strong light source, you will see that some of the light shines through you the edges of your fingers. To implement this effect I decided to make a Halloween pumpkin.

I began by making a pumpkin sculpture in Blender.


Making a sculpture in a 3D software is not that far from using physical clay. This is why the technique has become popular because it's easier to model something in clay than by adding triangles one after the other. The most popular sculpt software is ZBrush, but Blender has a sculpt part which is also working fine. The problem with a sculpture is that it consists of far too many triangles so it will fail if you put it in a game:


The solutions is that we once again have to cheat by making a less complicated model on the top of the more complicated model. This process is called retopo. The less complicated version looks like this:


But this less complicated version is kinda ugly. To make it look better we can use the more complicated version and "bake" normals and ambient occlusion, which is again a cheat to make an ugly model look better:


...and now it's just a matter of adding the subsurface scattering materials to the pumpkin, and it will look like this:


July 24, 2017

This self-driving car is now faster with the flow field algorithm

0 comments
So I've made a self-driving car in Unity. The car is using the Hybrid A* pathfinding algorithm along with other algorithms to find its way around a confined area, such as a parking lot. You can test it here or here. The old version was working fine, but one can always improve. The updates in this version are:
  • Added lines showing the waypoints 
  • Fixed a bug where the wrong value was added to the sum or errors in the PID controller, so the car is now following the path much better 
  • Added a cost for switching driving direction, like forward -> reverse, because it would be annoying to sit in a car that constantly is switching driving directions
  • Fixed a bug where the smooth path algorithm optimized the distance to obstacle in 3d space and not 2d space 
  • Added an improved obstacle-intersection algorithm which is faster and more accurate. It's now using rectangle-rectangle intersection instead of circle-circle intersection
  • Added UI where you can display the grid and the flow field 
  • When optimizing the final path, it's now optimizing the distance to the average of the surrounding obstacles and not just the closest obstacle 
  • Cleaned up the code
  • Added flow fields as heuristic and for obstacle detection, which makes the pathfinding much faster 
  • Made so that the Hybrid A star can expand to a cell if this expansion has a lower cost than previously
  • Improved path following so the car is slowing down gradually before it reaches the end or a turning point such as "reverse -> forward". Previously you could often see how the car crashed into walls because it was driving too fast

This is the flow field the car is using to faster detect if it's colliding with an obstacle:


The area is divided into cells, and the darker the cell is, the further away it is from an obstacle. So if you want to check if a car is colliding with an obstacle while generating the path, you can see in which cell it is and see how far away from an obstacle it is. If it can't possible be colliding with an obstacle, then it's not colliding with an obstacle and you don't need to use a slow rectangle-rectangle intersection algorithm to check if it's colliding with an obstacle.

Update 2017-07-26
I realized that a good idea might be to close cells depending on the heading the car had when entering the cell. So the car can drive to the same cell even if it is closed if it hasn't arrived to that cell before with the same heading. The result looks like some abstract art:


...but the algorithm is now finding more solutions and the car can even turn around on its spot. The problem is that sometimes this solution is much slower, so maybe a combination between the slow and the faster algorithm can be used?

July 16, 2017

What happened to John Carmack after the book Masters of Doom?

1 comments
John Carmack is famous for creating the games Doom and Quake. In the book Masters of Doom, you will learn how he created those games and the company id software. That book ended with the game Quake 3, but what has he been up to since 1997? The answer comes in a free 300 pages pdf, which consists of interviews with John Carmack, beginning with Quake 3 and ending with Rage, so from year 1997-2008.

Here are some key lessons learned from the book
  • "The assumption that hiring more people gives a better product is often incorrect. It might (but not always) give you more product, but not necessarily better product." [He argued that the 11 people id software had at the time is the right number of people to do a project and having 50 people would hurt the project]
  • You should listen to the community, but this will also take time, so you have to prioritize. Carmack argued that John Romero (who was forced to leave the company) was talking to the community when he should have been working.
  • "I will always take an aggressively simple approach to things" was his answer to the question if Quake's level should be more interactive.
  • Carmack was working 70 hours a week for six years.
  • "I had this really bizarre conversation with a couple of lawyers and they were talking about 'How do you pick your target market? Do you use focus groups and poll people and all this?' It's like 'No, we just write games that we think are cool.' They're from such a different world that they fundamentally did not get that."
  • "There's plenty of starving artists that are probably talented and hard working, they're just doing something nobody cares about."
  • "The graphics technology everybody looks at is only a quarter of Quake's code, and it's not even the hard part of it. Look at the things that are really unglamorous but really important, like the file extension architecture in Quake."
  • "I tried just a couple of features, like transparency or shadows and reflections. And they just dropped in - it was so wonderful. And that was what really got me crusading for OpenGL. It was like, "This helps me. This is something that's letting me be more creative and try out more things to produce a better architecture than we would have had otherwise." A good API can help you produce better products." [So using a game engine like Unity is not cheating!]
  • How do you handle third-party developers when you license your engines? Carmack: It's, "Here's a CD. Thank you for the half-million dollars. Have a nice life."
  • "All of science and technology is built standing on the shoulders of the people that come before you. Did Newton patent calculus and screw the hell out of the other guy working in it? It's just so wrong, but it's what the business world does to things, and certainly the world is controlled by business interests." [Why software patents is a bad idea]
  • You hear it from the fan base a lot. "Do it right. We'll still be here. We'll wait," and it's tempting to just let things slip. But that's really not OK. If you're doing something cutting edge, you're making fundamental decisions about your architecture, and if you let it slide for a year or two, then it's just not the right decision anymore. Even if you pile on all these extras, it's not optimal. It's not targeted at what you're doing. [Why it's important to ship a game. They planned it would take 12 months to make Quake, but it took 18, which is fine] If we shipped on time, we probably weren't ambitious enough. Quake 2 would take 12 months because "a lot of things are functioning better at the company."
  • "Programming is really just the mundane aspect of expressing a solution to a problem. There are talents that are specifically related to actually coding, but the real issue is being able to grasp problems and devise solutions that are detailed enough to actually be coded." 
  • "I feel bad for some companies out there. The founders, who are these incredible engineers, are now directors of their departments doing management rather than engineering. At the same time most of the people they are managing are nowhere near as good as they were at doing the actual work. That's what I hope never happens to me. I want to stay in the trenches working on the things all the time."
  • "In the last two projects, my time has been split. I'd have about 3 months och pure research, playing around with different stuff. And then after that it's about 16 months of work on the project. It would be nice to shift that more towards research, but I would never want to devote a majority of my time to research." 
  • "It's not the one brilliant decision, it's the 500 smart decisions that really make things good. Even at the end of Quake 3, I had a to-do list of a thousand things that could potentially be improved on. So it's a matter of going through and knowing all these things that could be done, and prioritizing what the "sweet spots" are."
  • "I actually learn from almost anyone. Maybe I was smarter than the [college] professor but it didn't mean that there weren't things I could learn from him. So it doesn't take someone that's necessarily a "better" developer or programmer for them to have things that you can learn from. So almost all the programmers I've worked with, I've learned something from."
  • "There have been flight simulators where you just jump in, fly around and shoot things, and those are fun and interesting. Then there are these serious simulators where you have to convince yourself that you're being entertained. I still like playing a simple, fast game, where you jump in and have a good time, and I think there are five times as many people in the game buying world that also feel that way." 
  • "Sometimes I wind up feeling guilty that I'm not doing more. I'm willing to just ignore a whole lot of things. And that's pretty important, because so many things come in that are potential demands on my time, and it's just easy to see how people that are in a similar situation, wind up just getting their entire days spoken for, and not being able to do any work because every day, there's phone calls and emails from people that want to do something, they want to have an interview, or pitch a business proposal." 
  • On market research: "I don't do a really through canvassing, play everything that's out there. Usually I watch over someone's shoulder when they're playing the hot new game. But I don't spend that much time actually playing other games. I certainly play more Quake 3 than any other games out there."
  • "Too many programmers agree to random feature requests without thoroughly considering the impacts."
  • "There are a lot of arguments that can be made about game design, and I prefer simplicity and elegance. I'm always the one saying we want the minimum number of everything, because I want it to be simple and fun to play." He would later say that he stepped away from id software because "I'm really no representative of what most of our market is now. I did realize that my very simplified game design ethic is not what the market is demanding. [Quake 3] was the id game I probably spent the most time playing and enjoyed the most, but it was actually one of our less successful titles." 
  • "Usually when I set out making the technical decisions I don't know how it's going to turn out. A lot of it is working out what works, and what ideas come to you. We commonly switch gears during our development process when a really good opportunity comes up." 
  • "Once or twice a year I go on "working retreats", where I lock myself in a hotel room for two weeks with no internet connection for completely focused work." 
  • "We try to pick directions that a good number of people will enjoy, then just do a quality job implementing it. Some people will love it, and some people will hate it. The people that hate it usually scream louder."
  • "Tessellation has been one of those things up there with procedural content generation where it's been five generations that we've been having people tell us it's going to be the next big thing and it never does turn out to be the case. What we want is something that you can carve up the world as continuously as you want without any respect to underlying geometry." 
  • "A lot of people don't seem to really appreciate how the vertex fragment rasterization approach to computer graphics has been unquestionable the most successful multi-processing solution ever."
  • "'Oh just thread your application.' Anyone that says that is basically an idiot, not appreciating the problems." 

Update! John Carmack replied to this article on twitter: 

July 11, 2017

Why you need to learn the Flow Field algorithm

0 comments
As said in the last post, I've read a lot about pathfinding. One of the pathfinding algorithms I found was the Flow Field. A Flow Field covers the entire grid (if a cell is not an obstacle and if a cell can be found from the start position), so while A* is finding the shortest path from the start position to a goal position, a Flow Field is inspired by water and is finding a path from each cell to the start position. It looks like this:


In the image above, the darker a cell is, the further away it is from the start position, which is the red sphere. The blue lines is pointing in the direction of the shortest path. But you can also add more start positions, which is actually speeding up the algorithm. So if you want to find the shortest path to the nearest door, then you can use the Flow Field.

A Flow Field is useful if you have a lot of units trying to find their way through your map, and was used in the game Planetary Annihilation. What they were doing was to divide the area into sectors, then use A* to find the shortest path between the sectors, and then they used a Flow Field to find the shortest paths through a sector. It looks like this:


But that's not it. A few years ago I made a self-driving car in Unity after reading a report on the Hybrid A* algorithm. One of the heuristics used by the Hybrid A* algorithm was something they called Dynamic Programming. I realized that Dynamic Programming is actually a Flow Field, so I updated my self-driving car with that algorithm. The 160x160 Flow Field looks like this and is generated in around 0.05 seconds:


You can also use the Flow Field algorithm to help you with obstacle detection. In the Hybrid A* algorithm, it's obviously important to quickly find the distance to the closest obstacle, and you can generate these distances fast once in the beginning. That Flow Field looks like this:


July 4, 2017

Pathfinding best practices and surprising uses

0 comments
Weather is crap, so I've studied a lot of resources related to pathfinding, which is the art of finding a path between two or several points. I've already spent a lot of time studying the Hybrid A* pathfinding algorithm used by self-driving cars, but one can't study enough of pathfinding! This is a summary of what I want to remember and what you also will most likely want to know:
  • A good pathfinding engine can be used for more purposes than just moving units around the world. In the real-time strategy game Empire Earth, pathfinding was used for tasks such as terrain analysis, choke-point detection, AI military route planning, weather creation/movement, AI wall building, animal migration routes, and pathfinding. 
  • Pathfinding can also be used as a flood-filling algorithm. The closed list will store all nodes searched, so you can use it to find which nodes were filled with the flood. In this case you can return the top open list node without searching for the cheapest node to process next because in this case, tile costs mean nothing.
  • Pathfinding can be time-consuming, so it may freeze the game. A solution to this problem is to split the paths up over time. This works well for both static and dynamic maps, and was used successfully in the game Empire Earth, which used paths that supported thousands of units. The solution is to first generate a quick path to get the unit moving because the unit should move the moment the player tells it to move. This quick path uses a pathfinder algorithm that stops after maybe 10 iterations and picks the node closest to the goal-node as the path's end-node. Then you should generate a full path, which is a path from the end of the quick path to the goal node we wanted to reach from the beginning. But this full path may be ugly and look wrong because it begins at the end of the quick path, so you need to add a splice path, which uses the same idea as the quick path by using the unit's current position as starting position and a point on the full path as end position. This point should be experimented with but can be like eight nodes in on the full path.
  • In Empire Earth, if an AI-controlled unit failed to find a path more than 10 times in 30 seconds (perhaps because the unit was walled in), they simply killed that unit.
  • Finding the cheapest node in the open list is often time consuming so you need to optimize the list. To optimize this you can use a sorted heap structure that allows for fast removal, but slow insertion. The second method is to have an unsorted list where insertion is fast, but removal is slow. A third alternative is to have a "cheap" list with roughly the 15 cheapest nodes (sorted). So each time you want to add a new node to the open list, you first check if it's cheaper than any nodes in the cheap list, then add it to that list and sort it, so it can have more than 15 nodes. Otherwise, add it to the list with all the other nodes in the open list, which is unsorted. If the cheap list is empty, add 15 of the cheapest nodes from the expensive list.
  • Iterative deepening is the art of restricting the A* algorithm by limiting the number of iterations allowed or limiting the maximum path length. If A* fails to find a path, the restriction can be increased. This can be used if the map is dynamic. Let´s say you have two agents that want to go through a door. The first agent will find a path, but the second will fail. If you had not used iterative deepening, then the second search would have taken a long time before failure. With iterative deepening, the second search will fail fast, but then it may succeed the next time loop because the first agent may have passed the door.
  • In some environments it might be a good idea to return the failed path if pathfinding fails. If an agent follows a failed path instead of just standing still, it might appear as the agent is exploring the environment. 
  • If everything else fails, let the player add waypoints, and the pathfinding algorithm can find paths between these waypoints. 
  • You don't always have to use A*. What if a straight line to the goal is possible? 
  • If the player can't see the AI controlled enemy units, then you can ignore pathfinding and just teleport the units, so pathfinding is not always needed. 
  • It's possible to pre-compute every single path in a search space and store it in a look-up table. For a 100x100 grid, this will need 200 MB of space. An alternative is to calculate a few paths, store them, and see if the path you create while the game is running can use those paths. A unit will most likely pass through a door, so calculate paths from the door.
  • It's important to optimize the search space, the fewer nodes you have to search through, the better. Maybe you don't have to divide the are into small squares, what if you divide the area into waypoints? Or you can combine both techniques: first find the shortest path through the waypoints, and then find the shortest path between the waypoints by dividing the area into squares. In the game Company of Heroes, the high-level search space representation was a hex-grid, and the low-level representation was a square grid. One way to improve search space is to use Quadtrees, which means merging cells that don't have an obstacle in them into one large cell. It works like this: Pathfinding in an Entity Cluttered 3D Virtual Environment
  • A* demands that you use a heuristic which i admissible, meaning that the h-cost is never larger than the true cost. This will result in an optimal path. But if you don't follow this rule, you can get a faster algorithm but not an optimal path. But why would you need an optimal path, will anyone really notice? You can add this to your game by simply multiplying the h-cost with a constant such as 1.5.
  • The best heuristic to use on a grid is the octile heuristic. It looks like this: max(deltax, deltay) + 0.41 * min(deltax, deltay), assuming that diagonal movement cost 1.41. The problem with the euclidean distance is that it underestimates the cost, while manhattan distance overestimates the cost.
  • When should you use a grid, waypoints, or a navigation mesh to represent the search space?
    • Grids are most useful when the terrain is 2D, when implementation time is limited, when the world is dynamic, and when sufficient memory is available. Don't use them when you have a large open world or when you need accuracy (like when you have a house with an angle that doesn't fit the grid)
    • Waypoints are useful when implementation time is limited, when fast path planning is needed, and when you don't need high accuracy.
    • Navigation meshes should be used when you have time to implement them. But there's no best technique to create an optimal navigation mesh. 
  • No one solution is useful all the time. Navigating open terrain requires a mix of different techniques. Use local collision avoidance techniques for nearby areas and pre-processed data for longer distances. Using multiple techniques that complement each other yields better results than any one technique alone.
  • It's kinda boring if all AI controlled units follow the exact same path. To solve this, each edge between two nodes can have a width, and you need to make sure this width doesn't collide with an obstacle. Then each unit follows the edge but with a certain distance from the edge.
  • While A* is finding the shortest path to a single point from a points, Dijkstra's algorithm will find the shortest path to all points from a point.
  • Pathfinding algorithms tend to terminate when the goal node is the node with the smallest cost, so the algorithm doesn't terminate when it reaches the goal node. But when it first reaches the goal node, then that path is in many cases also the best path, so you could stop the algorithm when it first sees the goal node and not when the goal node is the node with the lowest cost. 

Sources

July 1, 2017

The secrets of Artificial Intelligence in classic games

0 comments
The old classic computer games were fun and often challenging because of the Artificial Intelligence that controlled the enemies. But how could the AI in those old computers be challenging? This will be a short description of how the AI in classic games was working and the tricks they used.
  • Warcraft. The AI was waiting for some fixed amount of time, and then started spawning a wave of units to attack you. It would spawn these units at the edge of the gray fog and not at the base as you first may have thought. It continued to spawn these units until your defenses were nearly overwhelmed - and then it would stop, leaving you to kill the remaining enemy units and win the fight. So the AI didn't have to worry about building buildings, saving money, or building units. 
  • Pac-Man. The AI had two states: normal, which is when the player is collecting dots, and another state when the player has eaten the power-up. In their normal state, each of the four ghosts moves in a straight line until they reach a junction. At a junction, they randomly choose a route to move to next. Each ghost chooses either to take the route that's in the direction of the player (calculated by using an offset, so no pathfinding), or to take a random route. The choice depends on the ghost: each has a different probability of doing one or the other. This sounds like stupid AI, but those who played the game thought the AI was smarter, arguing the ghosts tried to set a trap or group up to attack the player.
  • The Sims. This game put the intelligence in the world and not in the character. The objects in the world, like a fridge or a television, tell nearby sims what they offer, such as food or entertainment. The object will also tell the sims how to perform the action to get the offer. 
  • Starcraft. The game had several pathfinding problems. One of these problems was that the units who collected minerals jammed. The solution was: "whenever harvesters are on their way to get minerals, or when they're on the way back carrying those minerals, they ignore collisions with other units."

Sources

June 27, 2017

Finding inspiration from other games

0 comments
In April this year I participated in the Ludum Dare competition where you are making a game in 48 hours. I made an Aircraft Carrier Simulator, looking like this:



Today, I saw on steam that someone had released a game called Carrier Deck. It looks like this:


In the game you "manage the deck of the Nimitz-class nuclear powered supercarrier, the USS Ronald Regan (CVN-76)." So the game has the same game play, the same view angle, the same carrier type, and the same F18 aircraft as my Ludum Dare game had. I wonder if they found inspiration from my game? That would be cool because I don't mind anyone finding inspiration from my ideas. I had plans to make a complete game out of my Ludum Dare game but decided not to because I thought it wasn't fun enough.

June 13, 2017

Resources if you want to learn more about architecture

0 comments
I've set out to learn more about architecture for a project I'm working on. This will be a collection of the resources I thought were the best. I've learned there is not one "bible" you can read to learn everything about architecture, so you have to use several resources.

The timeless way of building. This book is written by the architect Christopher Alexander, and is the first book in a series of three books that summarizes his ideas on architecture. You might ask yourself why some buildings are "better" than other buildings? Why do you prefer to live in a house but hate to live in another house? Christopher Alexander argues that the reason one building is better than the other is that the building you like was designed by using certain patterns. The main idea is that building a building is an iterative process, and those who will live in the building should be involved in the process. If everyone is aware of this "pattern language" then it will be easier for everyone to participate in the building process, and the result will be a building which feels alive.

Ted talks. Ted is an organization which posts talks, and these talks are supposed to be something extraordinary. It has a search function where you can find several talks on architecture. All of them are not extraordinary, but some are, including one guy who had an idea to stop the spreading of the Sahara desert by using clay buildings.

The Pruitt Igoe Myth. Pruitt Igoe was a housing project in St. Louis. The main idea was to give poor people an opportunity to live in better buildings. But the project failed because the houses were not maintained, so they began to fall apart, which resulted in that the people living in them stopped caring, so they fell apart even more. So the key point from the documentary is that a house on its own is not important, but what's happening with the house when the people have moved in.

June 8, 2017

Youtubers playing Tesla Simulator

0 comments
So I uploaded my Tesla Simulator to itch.io, and then something strange happened: people began uploading videos to YouTube of themselves playing it. No-one has ever uploaded any videos of themselves playing my games, and it's kinda strange to see someone playing your game online. No big youtubers are playing it, but smaller youtubers are, which is better than if no-one is playing it! Anyway, this is a small collection of them:







People are also wondering if I'm going to update it. And the answer is yes. But I have to figure out with what. Someone suggested you should be able to run over people, but that's not going to happen, and someone else suggested a big city, but that will take a looong time to build, so we will see...

June 5, 2017

The ups and downs of itch

0 comments
A few weeks ago I found a site called itch.io where you can upload games, books, and assets. You can either give away the content for free or you can sell it. It's also possible to use the combination by asking people to pay what they want for what you've uploaded. I uploaded my (free) Tesla Simulator as an experiment, and this is the result (May 18th - June 4th):
  • 1956 views and 1240 downloads
  • 307 visitors came from my website, while 99 percent of the rest came from itch
  • 839 downloaded the PC version, 335 the Mac version, and 66 the Linux version. This was really interesting to know because I've previously ignored the Mac and Linux users, but they are also interested in your products!
  • A few people have made YouTube videos showing when they play the game. No-one made a YouTube video of the game before I uploaded it to itch
But what if you also want to make money? As usual it depends on the game you have, but someone on Twitter shared their revenue and it looked like this:


Itch is less crowded than the other channels where you can sell your game, but that's also good because it's easier for people to discover your game. It's also easier to upload your game to itch if it's not finished so you can maybe use itch to get feedback while developing the game and then release it on Steam?

Lessons learned from the development of Minecraft

0 comments
A few days ago I wanted to learn how the team behind Minecraft marketed the game. I made a Google-search and found an article which referenced the blog written by Notch, who came up with the original idea behind the game. I realized it might be interesting to read the entire blog and see if I could learn anything from it. But I wanted to read it from the beginning and Tumblr didn't have that option, so I downloaded the xml file with all blog posts and converted it to pdf with the help of some PHP magic. The result is a 508 pages pdf, which you can find here: The Word of Notch.

It turned out it was really interesting to read the blog from the beginning of the end. Here are some lessons learned:
  • The "blocky" main character was from another game Notch had protoyped, but he thought the blocky design would fit into his new game. That design hasn't changed since much 2009. 
  • The name of the game changed from Minecraft: Order of the Stone. Why Minecraft? Because Notch thought it was a good name.
  • Notch published a lot of blog posts during the development of the game, sometimes two a day or more. Once he published nine smaller posts in one day. Not all post were related to Minecraft - some were movie reviews and other were game reviews. He also posted random ramblings about various topics like the singularity. Sometimes it took 5 days between each post.
  • Notch has always been a fan of the game Team Fortress, so he had a plan to add a similar "Fortress mode" to Minecraft as well as something he called "Zombie Siege." 
  • Notch was programming what he thought was fun to program. Once he had planned to implement multiplayer but got bored and decided to play around with water. This happened more than once. He added blocks like gold and trees instead of any multiplayer code. "I’m getting stressed out over the multiplayer release. It’s not going as well as I’d hope, and as a result, I’ve lost motivation, causing me to work really slow. This is a bit of a vicious circle, and I think I need to relax for a bit."
  • Some posts included stuff like videos and images made by people who were playing the game.
  • He was often struggling with performance problems, that Minecraft wasn't running fast enough. So if you are struggling with that yourself, then don't feel bad!
  • In 2009 Notch explained he had "another job" and didn't have enough time to update Minecraft. He asked for donations: "Also, donate!" 
  • He was asking for feedback. "Please remember to give me some feedback!" And when he got feedback, he showed that he appreciated it. "For those of you who have sent me feedback (ever), thank you very much. You’re the best players a game developer could hope for. You’re honest with your opinions and still understand that I may feel different from you. I listen to all your feedback. Thank you for helping me develop this game."
  • He outsourced the sound and music to a guy called C418.
  • In 2009, he wanted to sell the alpha version of Minecraft for 9.95 Euro, 14.95 for the Beta, and 19.95 for the final version. He also wanted a free demo, which I remember Minecraft had an online version of where you could just place and remove blocks. If you purchased an early version, you would get "to help fund the development of Minecraft," "custom skins," "all future updates". He would later add that those who bought the game got early access to new updates, such as survival mode.  
  • He used a forum while developing the game, and he got comments on the blog. "I think the minecraftforums.net forums are most excellent, and a LOT of good suggestions come out of that place." The forum was run by the community and Notch didn't have anything to do with it. Emails worked fine in the beginning and he replied to all of them. But as the game became more popular he got 50 emails a day and decided to reply to just some of them, but he read all of them.  
  • Most blog posts were short, just a few lines (so much for the SEO talk of long well-written blog posts). Some consisted of just images of someone playing the game or a YouTube video, or a YouTube videos where Notch showed an update to the game.
  • 587 people had purchased the game in June 2009, which is about 2 months since he began blogging, in a later post he said it had sold 990 copies between July 19 and June 13.
  • He was feeling the pressure if being behind schedule (because of an illness, most likely flu) and was "freaking out about making survival mode as fun as I can."
  • He wasn't using a design document, but he had two text files: 
    • bugs.txt - bugs that he couldn't fix immediately but want to fix
    • features.txt - a list of ideas he liked but is afraid to forget if he didn't write them down
  • When coding, his goal was to make what he wanted to make as fast as possible with whatever "ugly" code, and then he improved the code. It "helps motivation to actually see things happening in the game."
  • He spent a lot of the time playing the game. "It's very important to me that the game is actually fun and the only way to make sure they are is to play with it." And if the game is fun, it's also fun to play the game. 
  • When he finished a feature, he tried to release it to the public as soon as possible. Before releasing, he was checking the bugs.txt if there's a bug he should fix. 
  • Version numbering can be tricky to figure out, be he said he would just "bump the version number" when releasing a new version. 
  • He didn't keep a schedule when he wanted to have stuff completed, but "rather keep striving towards my mental image of what I want the game to be like." 
  • If he didn't know what to build next to the game, he played the game as if it was completed, and then "add the first thing that doesn't work yet."
  • "My main method for delivering updates about Minecraft and so will still be this blog" was a quote after he got a twitter account. He was also using irc to chat with people who was playing the game. Irc was one of the software he started at the beginning of the work day and kept it on all day.   
  • "Mostly my free time is fairly normal. I try not to get burned out while doing this." so except for some 12 hour coding marathons, he didn't work himself to death. He was trying not to work on weekends "as much as possible." He also took a three week holiday in Indonesia in the first year of development (no Internet available). While in Indonesia, his service provider almost deleted his database (including information about everyone who bought the game) because his account ran out of money. But he managed to find a wifi on McDonalds and could fix the issue. 
  • When making the game he had a narrow scope, and had plans to add modding support and extra expansions to increase the life span of the game. When the finals version was finished, his plan was to actually make two mobile games (unrelated to Minecraft) and then make another unrelated game, and the continue developing Minecraft, "or whatever else is most requested by the community."
  • At specific points, the game was free. "The reason it's free is to get as much feedback as possible on it, so I can fix bugs and tweak gameplay as good as possible. Once the game is done, it will not be free, except in the form of a limited demo." This upset people who paid for the game, but Notch promised they would get early access to other content (1-3 weeks before other players) and of course the final game. "That's what you've really paid for"
  • A few months into the game development, Notch realized he wanted to hire an artist who could work on the textures and on the models. He also wanted to pay this artist even though he could have found someone who wanted to do it for free. This was the reason: "Some people asked me why I said I wanted to pay, and it’s primarily because I want to make sure it’s a proper business deal so there’s no drama later on. I’ve had that happen before with donated content. But also, I want people to dare charge for their work. I want to be able to do indie games for a living, and I definitely want to help support others to do the same. If someone absolutely don’t want to get paid and make awesome graphics, I can live with getting it for free. ;) And, yes, you will of course get full credits for your art. I’d own it, but I wouldn’t claim to have made it myself." But the first artist he hired would leave after a few months. "However, I went from doing Minecraft on my own to suddenly having to wait for someone else before I could move on. We had repeated problems getting anything animated to load into the game, and eventually I decided to just go ahead on my own."
  • He was experimenting with ads, both on the blog and on the Minecraft homepage but only for those who played the game for free. "In the past, I’ve been dead set on not having any ads at all, but surely there must be a reason everyone else has ads? Thus this experiment." But he realized he would remove the ads because the ads were advertising other game: "But I’m encouraging users to pay for other games, when I would rather have them pay for my game, which is the reason they come there in the first place."
  • He was live-streaming his competition in Ludum Dare live on Twitch. He was scared to develop Minecraft live, because someone could take the source code, but he was doing public play-testing of the game.
  • Minecraft had during the first Christmas a "free copy to a friend deal." During the three weeks this campaign progressed, 1436 copies were given away. Also 361 copies of the game were sold. He also got 2 donations, a total of 67.58 Euro. So it might be a good idea to be transparent with how many copies are sold of the game, because people have this psychological principle saying that they buy what's popular. When Notch married, everyone who bought the game also got a gift code of the game they could give away to someone.
  • He was experimenting with a public TODO list, so people know what's planned and what's going to happen. If he wasn't sure what to do next, he asked the community by using polls. "I put up a poll on what to do next, and posted it on irc and on twitter. Farming got 39% of the votes, so that’s what I’m doing now!" He was also asking how long the day/night cycle should be by using a poll. He was using polldaddy.com. He began using scrum when he hired more people. 
  • People could register for free and registered user #100000 got a free copy of the game and a personal email. Notice that some of those #100000 users were computer bots who registered accounts and thus not real humans. Maybe it helped to sell the game because everyone didn't know that even though he wrote about it in the blog?
  • "I like it when several small but relatively simple individual components [like fire and water] kinda gang up on you and make the game a really complex environment. There’s no need in making each component super complex."
  • It took about a year before Notch realized that Minecraft should have an endless map.
  • He was playing other computer games while developing the game. 
  • He started to sell the game a month after the first engine test.
  • He had server issues and at one point he had to give away the game for free because the server died.
  • "Basically, any developer working on the game [Minecraft] (two people at the moment) can just come up with something they’d want to add on a daytoday basis, as long as the rest of the team thinks it’s a decent idea. If it ends up being fun, it gets added."
  • January 12, 2011: one million sales. April 25, 2011: two million sales.
  • "We stepped on the plane to New York, and the Minecraft server died. While flying from New York to San Francisco, Tobias managed to get the servers back online via the amazing wonder of inflight WIFI."
  • "One reason why Minecraft has managed to get as much personality as it does it that it’s only been a couple of fairly nerdy game developers working on it."
  • They didn't analyze how players played the game: "Right now, the only way we can figure out roughly what people are doing with the game is to track logins. Once you’re logged in, we have no idea what happens."

April 28, 2017

Short April Updates - from forest fires to foam skirts

0 comments

Forest Fire

Many years ago I took a class in Differential Equations. One of the assignments was to simulate the evolution of a forest fire and the result was displayed on a graph. The problem was that you only got the end result and couldn't see how the fire evolved to get to the final result. A few years later I learned Unity, and decided to see if I could simulate a forest fire while at the same time see how it evolved. And it turned out I could.
The forest fire worked, but the visuals didn't. The first problem I had was smoke. A forest fire is large and you need a lot of smoke to cover the entire area - and that is not good from a performance perspective, and the resulting simulation was running in like 3 fps (30 fps is the goal). Luckily, both Unity and I have improved our skills.
The first thing I improved was to replace the fire particles with a glowing-ground-shader. Most of the time, you can't see the fire because of all the smoke, so by having just flat fire instead of fire particles makes no difference. I also decided to use animated smoke particles from Unity's free library of particles: Free VFX Image Sequences & Flipbooks. By using animated particles, the smoke will look more thick and you can use less particles to achieve the same result.
The second problem I had when I created the original forest fire was similar to the smoke problem: a lot of trees will decrease the performance of the simulation. To solve that problem when I created the original version was to combine all trees into a single mesh. This is also a complicated solution because it will make it much more difficult to remove the trees as they burn down. Luckily I could now replace all that code with a single shader called GPU instancing, which is something Unity recently added. If you add one of those materials to objects with the same mesh, Unity will automatically combine them, and the performance will improve. Anyway, this was the result:


Shaders

As said in the forest fire update I added an animated glow shader to simulate the fire. The reason was that I've also improved my shader skills. The first shader skill I learned was Interior Mapping, which is a technique used to simulate floors in buildings. The second technique was Parallax Mapping, which is a technique not that different from interior mapping, but you can simulate more "organic" deformations, such as this:


Tesla Simulator

I few years ago I made a Tesla Simulator to market a book I wrote about a guy called Elon Musk. It was available to the public through Unity's webplayer, which you could run directly in a browser like Chrome or Firefox. The problem now is that most browsers have stopped supporting the webplayer, so I had to make an offline version. At the same time I've cleaned up a lot of silly coding mistakes I made, because the Tesla Simulator was the first project I made in Unity - and I've apparently learned a lot since then.

"Make a boat" tutorial

One of my most popular Unity tutorial I've made is Make a realistic boat in Unity. What the tutorial didn't have was how to add foam to the boat, and I've thought a lot about what the best way is to add foam to a boat. Then I found this tweet:


...which was a cool idea on how to add foam, so I decided to implement it: Add foam. To make that work, I had to learn a technique called Convex Hull, which I wrote a separate tutorial on how to implement: Find the convex hull of random points.

April 24, 2017

Aircraft Carrier Simulator - or how to make a game in 48 hours

0 comments
Ludum Dare was on this weekend. It's a competition where you are supposed to make a game during 48 hours, and you may win fame and glory if you make a good game, so no money is involved. But people have come up with good ideas by participating in the competition, and they have later modified the game, sold it, and thus made money from it. One example is the game Broforce.

There are two different competitions going on at the same time: Jam and Compo. If you are participating in the Compo, you have to make everything yourself, including the textures, models, and sounds. Not a single piece can be found by googling. This is a however a gray-zone. Are you allowed to steal an engine sound from a YouTube video, or do you have to record it yourself? While making my game, I watched a Twitch streamer who was also participating in the competition, and he had a really hard time making sounds for his game. You could see on the stream how he squeezed a coca cola can to make a specific sound, so you have to be a little creative. I usually get my sounds from this site: Bfxr, which is a site you officially are allowed to use. I usually click a million times on the "generate-random-sound-button" until I find something useful and then I modify the sound in Audacity.

Each Ludum Dare competition has a theme. The theme this time was "A small world," so I decided to make an Aircraft Carrier Simulator. You can say that an aircraft carrier is like a small world because you can find everything you need on the ship. A few years ago I made an airport simulator for the same competition. It failed miserably, and I have hopefully improved since then.

This was my 9th Ludum Dare competition. One thing I have learned is that sleep is important. The competition begins at 3 am local time and ends 48 hours later. I have learned that sleeping until like 7 am, work the entire day until like 10 pm, and then go do something else like watching a streamer while thinking about the game until 12 pm which is bedtime, and then repeat the process the next day is the way to go. Anyway, this was the result:





The game idea is that you are in command of this aircraft carrier's aircraft operations. It's your responsibility to steer the aircraft so they can refuel, land, and launch from the catapult so the enemy is not coming closer than zero meters from the aircraft carrier. If the enemy get any closer, you fail! The more aircraft you have in the air, the more difficult it will be for the enemy to come closer. If you think it looks and sounds interesting, you can test the game (and download the source file) here: Aircraft Carrier Simulator.

March 10, 2017

Evolution of helicopters in the Vietnam War

0 comments
The Vietnam War was the first conflict where helicopters played an important role. While the future Cold War battlefield in Europe would have consisted of great tank battles, the jungles of Vietnam prevented tanks from moving efficiently. There were American tanks in the Vietnam War, but their main purpose was to keep roads clear from the enemy by using a tactic called Thunder Run, which is the same tactic that years later would be used by American tanks to capture Baghdad.

Tanks in the Vietnam War were always staying close to roads because a tank in a jungle is a big target for the enemy. But the tanks would clear jungle close to roads if needed. There's an example of a helicopter that crashed in the jungle. To rescue it, tanks on a nearby road were ordered to move through the jungle and secure the crash site. And if the enemy had built a bunker, a tank would drive up to the bunker, stick the barrel into the bunker, and fire. But again, these bunkers had to be close to roads or the tank would simply not be able to get there.    

With tanks out of the picture, the US came up with the idea to use helicopters. In the Korean War, the main purpose of the helicopter was to transport wounded to the rear two at a time. But when the Vietnam War happened, technological improvements had transformed the helicopter to a fighting machine, so they were no longer just transportation vehicles. A guy called Jim Gavin wrote an article called "Cavalry - And I don't mean horses" where his vision was that bigger, faster helicopters could carry infantry into the battle and make it a three-dimensional nightmare for an enemy commander. US decided to add helicopter transported infantry to it weapons of choice, and the recently developed Bell UH-1 Iroquois (nicknamed Huey) was chosen as the main helicopter for this unit:

The Huey

The helicopter transported infantry was sent to Vietnam. But if you've seen the movie We Were Soldiers, you know that helicopter transported infantry wasn't a three-dimensional nightmare for an enemy commander. If the enemy could cut off the few landing zones available in the Vietnamese jungle, where the helicopters could land, the infantry would be on its own. Each Huey has two machine guns operated by relatively unprotected soldiers, so the damage they could make was limited. The solders on the ground could get support in the form of artillery and air support from winged aircraft, but this support wasn't accurate enough to rescue the infantry in all situations. So something else was needed.

The Huey was a great helicopter, so someone realized that it was possible to take the basic engine and rotor, but add a new body to get a helicopter that could support the infantry. So the Bell AH-1 Cobra was quickly developed and would arrive to the Vietnam War.

The Cobra

The Cobra could kill the enemy but it couldn't find it. So someone came up with idea idea that you could combine the Cobra with the smaller, but much more agile helicopter Hughes OH-6 Cayuse (nicknamed Loach). Now you would get a so-called hunter-killer team, where the Loach was finding the enemy and the Cobra was killing it. The Loach could seat five: two pilots and three passengers. But in the Vietnam War, the Loach had either two pilots and one gunner, or one pilot, one gunner, and one mini-gun fired by the pilot. Why not four in the helicopter and a mini-gun? Because that would make it too heavy!

The Loach

One of the pilots who flew a Loach as part of a hunter-killer team in the Vietnam War was Hugh Mills. He wrote about his experiences in the book Low Level Hell - A scout pilot in the Big Red One. A helicopter scout in the Vietnam War had the following jobs:
  1. Find enemy base camps, fighting positions, supply caches, trails, and other signs of enemy movements.
  2. Assess damage made by high-altitude bombers, such as the B-52.
  3. Find landing zones for the infantry carried by the Huey helicopters.
  4. Help the infantry and tanks on the ground by giving them information, such as which terrain is the most advantageous, and if the are moving in the correct direction.

If you are a scout pilot (the hunter), the easiest way to find out where the enemy is hiding is to fly as close as possible to where you think the enemy is and hope that the enemy will fire at you. If you are fired at, then you drop smoke, so the Cobra (the killer) know where the enemy is and can fire the rockets. The result of this suicide tactic was that Mills was shot down no less than sixteen times and wounded three times.
Though I was getting shot at almost every day, I never got used to it. But getting shot at was usually the way a scout found the enemy, and finding the enemy was our basic job.

It might sound strange to sacrifice one helicopter, while the other helicopter is waiting for the scout the be shot at. But the pilot in a Loach has a better view of the surroundings, and the Loach is also a smaller helicopter making it more difficult to hit, so the tactic makes sense. To help the Loach, the Cobra was always staying within sight of it and did everything the pilot in the Loach didn't have time to do. The Loach pilot had to fly the helicopter while having his eyes constantly focused on the ground, so the Cobra read the map and transmitted radio messages.
The good scout pilot never stops talking to his gun [the Cobra] from the moment he goes down out of altitude until he comes back up again. It not only keeps the Cobra happy and informed, but it tends to keep your own guts stabilized when you're down low working and, at any instant, could catch a bellyful of AK-47 fire.