Solutions in this chapter:
The contests described in other chapters are more specific to those where each competitor has its turn, and the results compare the individual performances. In this chapter, we’ll talk about competitions where the rival robots fight face to face in a more spectacular way.
In our experience, sumo is one of the most suitable kinds of competition for small robots, offering the opportunity to test an incredible range of techniques that may prove useful in all your projects, not just during contests. We will take a look at variations on some familiar solutions—such as bumpers and the use of the ultrasonic sensor—and we will introduce some new ones. For example, we will illustrate a transmission, which behaves like a sort of automatic gear switch.
Although the technical aspects of building a successful sumo robot are important, the design requires much more than simply putting together a few mechanical solutions: It requires a strategy. Will your robot be very aggressive, or do you prefer a defensive approach? It could be robust and slow, or lightweight and fast. It could be designed to actively search out its opponent, or to react when it’s under attack. You cannot work at the mechanical configuration and decide how the robot should behave after it’s finished. On the contrary, you have to pick up a strategy and design both the mechanics and the program according to it. This principle applies to any robot, but it is particularly important for sumo robots, and it is the key to understanding this chapter: We want you to devote the proper attention to the connections between the planned behavior of your robot and the solutions you can adopt to effectively implement it.
We explained in this chapter’s introduction that when you start building a robot for a sumo contest, you must have a strategy in mind. The process starts before you build your robot. It begins by examining the rules carefully, understanding what you can and cannot do, and deciding your line of action. You must try to imagine what the opponents’ strategies can be, and plan your robot to be able to resist their attacks and take advantage of their weak points. Obviously, you cannot really know how the other competitors will strategize and behave, but this exercise helps you to focus on a well-defined strategy. Remember that any strategy is better than no strategy at all! Figure 20.1 shows a simple sumo robot ready for action.
This section starts by describing a typical set of rules, which will help you in framing what a sumo contest is, and provide a starting point in case you want to organize your own. Then we’ll describe how you can tune your robot to produce maximum force, which is undoubtedly a very important component in a sumo competition. We will also explain how to configure your robot to take advantage of some important offensive and defensive behavioral strategies.
In a typical sumo competition, you will receive two sets of similar rules. The first set of rules states that the robots can be made out of any original LEGO pieces, in any desired quantity, but that they must be within a maximum size of 32 x 32 studs and a maximum weight of 1.5kg (3 lbs). In the alternative set of rules, which we call Mini Sumo, each robot may be built using only parts from a single MINDSTORMS NXT set; there is therefore no need for size and weight constraints.
For most other aspects the two sets of rules are almost the same:
The field is a circular or square pad with a contrasting external strip of 20 cm (8 inches). Usually the pad is white and the strip black, or vice versa. There is also an optional 2-inch warning line that warns the robot as it approaches the edge of the ring.
Only two robots can fight on the field at a time. Should one robot for any reason find itself outside the field boundaries, that is, any portion of it touches a point beyond the external strip, the robot loses the round. If neither robot is eliminated within a chosen time limit (e.g., three minutes), the match ends in a draw.
A robot may also be eliminated if it is overturned by its opponent or it finds itself in a situation where it can no longer maneuver.
No “violent” behaviors are allowed. A robot can only push or lift its opponent. It is in no way allowed to damage its opponent’s structure or parts. This will be left to the judge’s discretion.
A robot cannot drop any part or subsystem in the field either deliberately or involuntarily. Any part found loose on the field will be removed by a member of the panel.
The robots must be fully autonomous; any kind of remote control is for bidden.
Every robot must comply with the limits in size and weight at the beginning of a match, but once the match starts, it can modify its own structure, perhaps extending parts itself so its dimensions become larger than the initial specified size limits.
There are many other, less important, rules covering items such as batteries, the composition of the panel, the prematch test time, and more. Some sumo competitions require that your robot pass an admissions test: It should be able to push a block of wood out of the fighting ring. If it can’t beat a block of wood, it has little chance against another robot, and this rule is meant to screen out robots that are too weak to enter the contest. This rule is not that commonly enforced, but it is a good exercise, and it may help in filtering out weaker robots to make large tournaments quicker.
The making of a strong sumo robot requires much more than just brute force, but we cannot deny that maximizing the generated push will increase your chances of winning some matches and, maybe, the tournament.
When optimizing the pushing power of your robot, the first thing you need is an objective way to measure it. Without measuring the force, the improvements you make are subjective and, as a result, are very inaccurate. During preparation for one of the first robotic sumo contests, a friend and robot builder suggested a simple trick based on a very common object: scales, such as those used in many kitchens to weigh flour, sugar, and other ingredients.
You have to place the scale on its side, on the table or the floor, possibly removing the upper tray, and hold it firmly while your robot pushes against it. You’re not interested in the absolute value that the scale indicates, but rather in comparing the push produced by different setups.
Many factors affect this force; you can imagine a sort of path of power that goes from the batteries to the wheels, passing through the motors and the gearing, decreasing in accordance with the variables that affect each part along the path (see Figure 20.2).
The rules will hopefully specify that all competitors use the same kind of commercial batteries. Between the batteries and the motors, there’s the NXT. It’s worth reminding you once again that the NXT incorporates a current-limiting device to protect the motors connected to its output ports. If the rules allow the use of custom parts and you have them or are willing to make them, you can consider the use of a Motor Multiplexer from Mindsensors or HiTechnic, or the use of a homemade motor hub made by Philo (Appendix A).
The number of motors influences the generated power. Simply use the maximum allowed by the rules and by your own inventory. As for the mobility configuration, the differential drive allows for the highest combination of maneuverability and simplicity. The fact that it doesn’t go perfectly straight is not relevant to sumo fighting, and the dual differential drive has no advantages in this case. On the contrary, the capability to use one motor to turn and the other to move reduces the maximum generated force.
The optimal gearing is, as always, easier to determine by experiments than by calculations. Generally speaking, the higher the reduction ratio, the higher the push, but this doesn’t mean you should gear down too much. Speed has its importance (we’ll explain why later in the chapter), and very high reduction ratios introduce too much friction, which uses up precious power.
Now we come to the part where you have to convert the produced torque into the actual push. The wheels are a critical component: If they don’t grip the pad well, the rest of your efforts will prove fruitless. This is when the scales we mentioned earlier prove to be an enormous benefit. By testing different kinds of LEGO wheels, you’ll discover that there are significant variations in grip. The ones from the 8462 Tow Truck work particularly well, as well as the large spoke wheels from the W979648 Education Resource Set and many others. On no account should you use tracks. They offer extremely low grip, and almost no grip at all in the direction perpendicular to its motion. You’d have little hope at all if your opponent broadsided you—an eventuality more probable than a head-on collision.
If possible, try to test your robot on a surface similar to the contest’s official pad. Different materials require different wheels. For example, the wheel having the best grip on a smooth tabletop is not necessarily the one with the best grip on a rough plywood surface.
The position of the center of gravity is also very important when it comes to friction and your wheels. Keep the center of gravity (COG) as close as possible to the main drive axles.
We anticipated that force wouldn’t always make the difference in a robotic sumo contest. Many different strategies can affect the result and cause a robot to win out against a more powerful competitor. These include finding the enemy first, using speed as a force, using a gear switch for maximum speed and push, and other offensive tricks.
A very important rule is to find your enemy before he finds you. This basic military principle applies to sumo robots as well, for the simple fact that the first one to engage the other has a good chance of attacking it on a weak side. Sumo robots are generally designed to push forward, and they offer much less resistance when attacked from the side or rear. In fact, they often don’t even realize they’re under attack, because often they’re not designed to detect the enemy from behind or from the side. In such cases, you can say that three sides out of four are generally weak.
When you start planning, the most logical sensor to use to find the enemy is the ultrasonic sensor. Spin, go toward the closest object, and repeat. It seems perfect, but there are some loopholes. Among them is the fact that they interfere with each other. If both you and your opponent used the ultrasonic sensor at the same time, your robot would start reading your opponent’s signals and get confused. You could solve this problem by using Guy Ziv’s Ultrasonic ping method (see Appendix A).
Another issue is that you’re trying to find a pile of LEGO pieces, and the ultrasonic sensor is meant to find large solid objects. Possibly, you could mount the ultrasonic sensor on its side so that you have a more narrowed range but a better resolution. This can be left as an exercise to the reader.
A technique that is simpler but just as effective employs contact sensors, in the form of either bumpers or antennas. Bumpers don’t require any particular trick. You simply program your robot to turn toward the obstacle instead of avoiding it. Design compact and smooth bumpers devoid of any unnecessary protrusion, to reduce the chances of getting caught on an enemy robot and dragged off the playing surface. With antennas you can use either touch or rotation sensors, the latter being able to tell you more about the direction of the opponent. But with the latter, you will have to use the legacy rotation sensor because the NXT motors have too much internal gearing to move from a light push. And you would have wasted an entire motor!
Speed is an extremely important factor in the search for the enemy. Imagine two robots running freely on the sumo field, simply going straight until they find the border and change direction randomly. Supposing that they have different speeds, the faster of the two has a much greater chance of intercepting the other. For this reason, it’s important not to have too slow a robot. Find a compromise between pushing capability and speed.
The robots built around speed all have a common strategy: Crash into the opponent repeatedly and use its momentum instead of its strength. The resulting energy makes the opposing robot lose contact with the ground, which gives your robot time to rear up and assault again. One charge later, the enemy is often found helpless. Though these robots may seem cheap, you have to appreciate how much experimentation it takes so that they do not illegally destroy their opponents.
Note
Momentum is a physical quantity defined as the product of mass times velocity. You can understand what it means through an example: You face a person of your same weight and build that’s trying to knock you down. If you’re both stationary, you have a good chance to resist. If, on the other hand, you are stationary and the other person is running toward you, you will very likely go down.
Other robots use a transmission to get the best of both worlds: fast speed during the search phase, and maximum push after the engagement. Sure, some robots will use a special transmission ring included in some vehicle sets, but it was soon proven during a contest that it’s possible to make a sort of automatic gear shift even inside the strict rules of Mini Sumo. Look at the assembly in Figure 20.3. It’s not very solid, but it explains the principle: The wheel on the right in the picture is geared with a shorter ratio than the main one, and during normal motion it slips a bit because the robot is moving faster than the speed of the idler wheel. When the robot slows down or stops for any reason, the faster wheel slips, and at that point, the slowest one grips. Because it’s mounted on a short independent beam with a free end, part of the torque pushes the wheel down and consequently lifts the robot. This mechanism is very fascinating to watch, but it’s very difficult to understand just by looking at the picture, so we encourage you to build it and try it out. Just remember that you need two of these assemblies, one for each side, and a supporting wheel. The wheels in the NXT kit can replace the ones in the picture.
Many other tricks prove useful during a sumo contest. The ones most often used are meant to lift the opponent, thus getting two positive effects: reducing or canceling the grip of its wheel and transferring part of its weight on your robot. This class of method includes at least two large families, one based on inclined planes and the other on counter-rotating wheels.
An inclined plane works like a wedge that slips under the enemy robot. It can have the shape of some small slopes placed at the front side of the robot, or of a large inclined surface that covers the whole robot. In the latter case, a LEGO baseplate is the better choice: Mount it studs-down and you’ll have a very smooth top surface to wedge under your opponent.
Counter-rotating wheels are very effective too, but they require an additional motor to operate them. Be sure they don’t touch the ground, though; otherwise, they’ll counteract the forward motion of your own robot! The combined effects of the front wheels with the push of the robot may even overturn the opponent, a spectacular but rare event.
So far we have discussed attack strategies, but protecting the weak sides of your robots is important as well.
Every active defense system relies on the fact that you know what’s happening around you, and require some sensor to detect a possible attack. Depending on the rules of the tournament, you might find yourself dealing with a limited number of input ports, requiring that you carefully plan how to allocate them in regard to your navigation, attack, or defense subsystems. The simplest detecting system is a sort of large bumper that covers a whole side of the robot. With this “ring” bumper, one touch sensor will be able to detect an attack from any direction (see Chapter 9).
Another method of detecting the enemy is to check whether your motors have been stalled. This is easy to do with the built-in rotation sensor of the servo motors. If you run your motors, you can monitor their speed and rotation values. If the motors are under stress, through software you can detect a stall condition and assume you are jammed against something—in this case, the opponent. Then you can test each direction to know from which side the opponent got you.
When you detect that you’ve been tackled, you have the option of either escaping or facing your enemy. The first choice is best when fighting a slow, strong opponent, whereas the second works well when it’s your robot that has a strong push (though it’s not always easy to turn in place when being pushed). Some rules allow competitors to use more than one program. Take advantage of this opportunity by preparing different versions to implement different strategies, and then select the one most suitable when you know which robot you’ll be facing in a given match.
Also consider passive defense systems, the kind that don’t require any sensor or port. The more obvious defense mechanisms pertain to the shape and size of the robot itself. A smaller robot offers less surface area to an opponent than a larger one, and though a triangular shape is more difficult to build, it’s also more difficult to catch. Make the perimeter somehow convex if you can, so as not to offer any holds that will help your opponent. Clearance from the ground is important for the same reason: It reduces your enemy’s chances of wedging itself under your robot.
More sophisticated passive defenses include protruding beams or axles meant to keep the enemy away from your robot’s vital organs, freewheeling vertical wheels on the sides to neutralize lifting wheels, and free horizontal wheels to allow your robot to slip away when engaged on one side.
This phase is crucial to a good result. Start testing your robot on a pad similar to the tournament’s to make sure it doesn’t do senseless things in the most common situations. It should detect the edge of the field when reaching it from any angle: You can’t imagine how many robots won a match because their opponents killed themselves!
When everything works well, you can start more advanced testing. You really need a sparring partner, but it need not be a second robot. Many reasons suggest you use a fake robot as a sparring partner, something you can move by hand to create any situation you want. (Using a real robot, you’d end up testing both instead; plus, you risk not being able to control specific scenarios.) A simple box does the trick, or a heavy book. Start by leaving the fake robot still and in the middle of the field, and see what happens. Your robot should find it, sooner or later, and push it off the pad. When this works, move the fake robot yourself to test the defensive strategy of your robot, and its behavior at the edge of the pad, the most dangerous area.
Remember that the perfect robot doesn’t exist. For any winner of a contest, it’s possible to design an “antidote” robot capable of beating it. You just have to accept some compromises in your project and make some assumptions about your opponents, hoping they won’t prove too far from reality.
If you have no previous experience in robotic sumo, you may think of it as a competition based solely on brute force. We must confess that we also had many preconceptions our first time out at a competition of this kind, but we had to change our minds. Force is indeed important, but it typically proves useless when you’re up against a good deal of intelligence.
These competitions have nothing in common with the kinds of events that feature radio-controlled machines, called “robots,” that try to destroy each other. These are not robots, simply because they totally lack a distinctive robot property: autonomy.
The first important lesson that this chapter taught is that you must design your robot with a strategy in mind, choosing the configuration that best suits your goal. Start examining the rules, and then make a hypothesis about your opponents and devise a strategy to beat them. Your opponents may be very different from how you imagined them, but this is not important—what’s important is that you build and program your robot to be consistent with the strategy you chose. A perfect robot doesn’t exist; in fact, situations in which Robot A beats Robot B, which then beats Robot C, which in turn actually beats Robot A, are very common in contests. And they’re what make contests so interesting and instructive.
We hope you also understand the second important message of this chapter: When building and programming your robot, make reliability your first priority. If you can beat a block of wood in a sumo match, you’re halfway to success!