The programmer was the best programming student I have seen in years. He used two main strategies. 1. Most important was PID. This is the most efficient way I know to keep a robot on the line and smooth. 2. Notice the long straightaway in the middle of the board. He computed the time it would to take to run the line at the robots fastest possible speed. The when he hit the straightaway, he went full speed until just before the end of the straight section. This alone cut his time by three or four seconds.
surely he hardcoded the robot to go a certain path? if not how can the robot tell exactly where to turn? the fastest robot won't work in a random maze, unlike some of the slower robots.
The student used the left hand rule and the algorithm I posted as a PDF last year. He probably spent in excess of 100 hours improving the standard algorithm. The robot keeps track of all the long straight portions of the maze, so that the second time through, it can be speeded up considerably. The algorithm predicts when a turn is coming and slows down just before the intersection.
Following the left hand path puts the robot in league with the devil. This algorithm is also sub optimal. It might work well for this maze but it won't be best for all possible mazes. If you watch the learning run it does not explore the whole maze. It could therefore overlook a shorter route than the one discovered. A better algorithm would retrace and explore missed turnings and the whole maze before determining the best and shortest route.
@@nathanaelsmith3553 A better algorithm for a different case, yes. This is an optimization problem, not a generalization problem. What you do in engineering is entirely dependent on your goals. Yes, generally, you want the most generally useful system you can make, but sometimes you have a highly specific use case and you want to make a device that optimally fills it, regardless of what that costs in other domains of performance. If, for example, I were designing a hammer, I could increase its general utility by adding things like bottle openers and screwdrivers, but every feature I add that does not help it put nails in better limits its narrow use as a hammer. In making the tool more general, its fitness in each domain of performance became more limited. If I need a hammer that opens bottles and turns screws, that's the hammer I'll make, but if I need to put nails in at peak efficiency that hammer is worthless to me.
@@2010ngojo Would be funny if they programmed one to have a cheat switch, and when the judges laugh and say "WTF was that?!" proclaim, "Oh, I thought this was any%?"
The one that messed up did so for the first time at 4:39, but someone touched that spot at 3:34 in this video. This is enough evidence to warrant further investigation into tampering with the track. The person might otherwise be smoothing out an air bubble or observing some other inconsistency with the line in that spot.
Oils from the finger made the line too reflective for the camera to pick-up correctly? I guess, on the second run, the robot is allowed to use just the correct turn direction at each intersection (from the first run), but is still required to scan the path, right? What about cutting corners? Max motor power? Weight?
They seem to do the search by hugging the left wall of the maze. You could troll them by creating a circular maze where turning to the right on the first intersection gets to the finish in just another turn, but turning to the left on the first intersection finds a really long twisted path that eventually ends up at the finish line as well. (Of course if the rules are that the maze will have no loops, then the only "trolling" you can do is to make the left-wall-hugging maximally long, while the actual path to the finish is very shortly after turning right in the first intersection.)
How come the track was tamped with during the last run? look at around the time 3:33 someone put a finger over the exit place that the confusion happens.
A voltage doubler or tripler and then regulated back down? I've only seen either a battery/motor combo slightly overrated for the task and then simply regulated down, or using a regulated switching boost converter, which are used in just about everything else ever built that's regulated.
I don't think it's good for the professor to show bias by hyping up one of the students, talking about how fast his robot is and speaking enthusiastically about everything it does. But then when the other student goes, he hardly says anything at all. It could be discouraging for the other students to see the professor playing favorites.
@Dino Sauro These are not children. The professor shouldnt spoonfeed you the answer on what to code. You have to learn it yourself. They shouldnt give a shit if someone did better than them. If look closely some of the students are even smiling because at least one of them passed the test. It got nothing to do with ego thing. Either you pass or not. You also didnt notice that the professor is actually explaining to the other students how did the guy coded his one. If youre one of the students there, you should be taking notes.
@Dino Sauro Happy and excited to see what the students who work hard achieve. => Showing bias and favouritism. How can some people still think big ass grown students actually need teacher to mandatory praise them to do good. Your study, your efforts it's not the teacher fault that he excited to see someone can do better. Professor know dude good and give praise, not good no praise it's that simple. Imagine hyping sb up for them to fail miserably, how does that sound to you. What worse, fail or fail when everyone except you to do good? He know what his students is like and react accordingly, and also give out lesions on how dude achieve that speed. If you need someone attention and praise to do good, you need to grownups.
Could you describe the robotos movement knowing the mass and friction constant as single transfer function? Then you could calculate the break off frequency and give an estimated break distance for any given speed. From there you could calculate speed from the pre measured distance and the average acceleration.
3:34 SABOTEUR! He rubs something on the track to mess up the speed run. It's the exact same spot where the bot messed up. He thought nobody was watching.
All high schools and maybe even earlier should have competitions like this. They do very similar national competitions in Japan (they run in a maze instead of following a line) and the tiny robots move at lightning speed once the maze layout is learned. I'm not kidding. Check this out: ruclips.net/video/76blllun09Q/видео.html
There's someone standing between window and track casting shadow and light onto the track on the spot where the robot fails. Maybe it wouldn't have failed with more uniform lighting
@@peter9477 In order for that to happen, someone would have to make a new path after you've gone past that part, converting a straight path into a T shaped path, connecting to the infinite loop.
@@BeeEatingOrchid You're right about that... unless the goal lies somewhere inside the area formed by that loop, in a disconnected island. I don't know what terms are used to classify mazes but this would be one of several forms involving a cycle (loop) but it's true that not all mazes with loops would be impossible for always-turn-left.
I'm just getting into engineering, so I apologize if this is a dumb question: how is the voltage from the batteries being boosted and then regulated, and how does that make the power average out no matter the battery voltage? Transformer or Op Amp or what?
Probably a switching regulator. Given a pair of AA batteries, probably a boost converter. It'll run at a constant voltage till the batteries can't output any more current (then probably stop working suddenly)
@@l8dawn Boost converters? The diode prevents the high voltage side from backfeeding the supply - When the transistor is open, current flows source - inductor - transistor. When the transistor closes, the current flows source - inductor - load. The load usually includes a capacitor, and without the diode the capacitor would be at a higher voltage than the source and backfeed it. The diode prevents that reverse current flow (which is probably not harmful, in the case of batteries, but does mean that you can't actually, well, boost the voltage and makes the whole arrangement rather pointless.
This seems simple but its amazing probably took a good amount of work Seems like the first action it takes when it detects multiple possible direction is left, straight, then right.
Actually simpler. The rule is just to take the left-most path you haven't already been down. (When coming back out of dead-end, that means the next attempt will look to you like it's now trying "straight" relative to the original path, but it's still just left relative to the direction it's heading at that moment.)
Hello: ? about the 13 sec robot. The Professor said something about a circuit that maintained a constant 9 volts until the battery dies. Could some one please show me the circuit and the component requirements. I just bought a BOE-BOT a couple of weeks ago and have had trouble with getting the same function like turn left. They are not similar or the same. The clam is that a constant 9 volts will help this problem. I would like to try the circuit. Thank You Paul
Ecstacy of gold plays as everyone freezes in shock. Even the robots would turn to watch who stepped in. Out cries #1 "That's impossible! You've been gone for less then 5 minutes!!!" Even without a face, everyone could see the robot smirk as it was announced: "THE TWEAKER HAS RETURNED"
the dirty blonde hair in a long pony tail, the unshaven stubble, the white tee, and of course- the glasses... he is the chosen one, foretold since ancient times- destined to take the programming world by storm. his name? probably scott.
Really cool! That left side algorithm can be improved by having the robot skip all left turns until it hits a dead end/intersection, then have it turn left or turn around and turn left.
I would love to see a future version of this where just before the the end their is a orbital track around the end as that can be a issue with depth first graph search on any type of cyclic graph. I’m also assuming that they don’t get a Manhattan distance to make a heuristic out of either since both robots shown turn left down paths that increases the Manhattan distance before those those that decreases it , just I’m just curious if the robot is able to work out the nodes it’s expanded or is only capable of backtracking
Oils from the finger made the line too reflective for the camera to pick-up correctly? I guess, on the second run, the robot is allowed to use just the correct turn direction at each intersection (from the first run), but is still required to scan the path, right? What about cutting corners? Max motor power? Weigth?
@@maxsolo2652 my comment has nothing to do with any real world effects this is looking at a flaw that is in the depth first search on a cyclic graph without expanded node tracking as a orbital path around a end point can result in the search not finding a solution.
Why does the robot always check left first, v right? I get why randomized doesn't make sense, but I'd say anticipating a clockwise circuit would be a good idea if the problem track was likely to resemble real world preference. E.g. most F1 circuits run clockwise, not sure why, maybe "handedness" or another mental bias. I'm not sure about racing circuits in general, or whether that handedness could directly relate to the setting of these kind of robotics problems, but I'd start by turning right at the first corner, myself. Amazing video, the winner is a true champ.
I think mazes have to do with sword wielding in the past where if you hug the right wall so that your right shoulder touches the right wall you can hardly or just can't swing your right arm to the side but hugging the left wall allows if not maximizes the amount of the space to swing a sword with your right arm.
Heh. Pretty sure this has nothing to do with swords. It's not like having to run through mazes with swords was a common pastime, aside from in Greek mythology and Dungeons & Dragons...
This video is 13 years old, but the oldest comment that I can find as of the end of June 2022 is just three weeks old (beginning of June 2022), meaning the earlier comments were either deleted or comments were disabled up to then. So for the first time since 13 years ago we can discuss the cheater at 3:33.
you could improve it by predicting what turn would come next, in the case of the several left turns in a row, have it learn to expect such a thing. Some kind of pattern recognisition. It's amazing that whenever a human tries to design something random, like that track, it ends up resembling music
that is true in this scenario but not always the case which is why it is necessary to come up with an algorithm that can solve any maze not just a specific one. if you made a bot that could solve this maze by prioritizing straight lines over everything and you make a new maze suddenly the bot may not be as efficient.
@@milokiss8276 yes because prioritizing straight lines makes it so the bot has less information about the maze making it prone to getting lost easier. the left turn rule works amazing well in maze like these but when it comes to closed loop mazes always prioritizing left hand will have the bot going in a circle forever in which a simple exception of "make no more than 3 left turns in a row" would solve. in either case prioritizing straight lines is not always going to work so the left hand rule is still preferable for any maze.
@@godw1ll99 Do you happen to have an example of something that would have a bot get "Lost"? I'm having trouble understanding how it is that "Always turn left" or "Always turn right" or "Always go forward" could possibly differentiate. It seems like an arbitrary choice, As long as you stick to that rule.
Try "leftmost unvisited path" is an easy algorithm to describe and code. "Try first the straight path if there is one, then try something else (but what?), and try something else again if you have to go back" is rather more awkward to deal with.... Anyway for this particular maze left first is suboptimal, sure, but there are suboptimal mazes you can build for any algorithm. "Random pick of unvisited options" is possibly one with a better average outcome, for most mazes they're likely to design on that table, and at least it will behave differently if you have to reset it and start over.
Left hand rule, when you enter a maze, place your left hand on the wall and follow that wall. (or right, it doesn't matter) Eventually it will lead to the exit.
@@StarWarsMasters I was picturing three optical sensors underneath, pointing down, to detect lines, if any, to the left, right, and ahead. Could be other possibilities, but I'm not sure we're even on the same page here yet. Maybe you were asking about something else?
@@StarWarsMasters On the first (slow) run it doesn't know the way yet, so it follows the simple "follow the left path until you reach the goal" algorithm. From that it determines the most direct path, without the dead ends. On the second run it follows what it memorized so has no false turns. Does that help?
Nerds couldn't use : Tremaux's algorithm, invented by Charles Pierre Tremaux, is an efficient method to find the way out of a maze that requires drawing lines on the floor to mark a path, and is guaranteed to work for all mazes that have well-defined passages. A path is either unvisited, marked once or marked twice.
If the maze is the same every time you can create tailored algorithms that perform more optimally than the ones optimized to solve arbitrary mazes. That is what the people are doing here. A lot of people in the comment saying there are general algorithms for this and totally missing the point of the competition. It's an optimization competition, not a generality competition.
Der Roboter ist so programmiert das er immer erst nach links geht. Wenn man den linken jedoch mit dem hauptast verbindet geht er die ganze Zeit im Kreis oder zurrück
The programmer was the best programming student I have seen in years. He used two main strategies.
1. Most important was PID. This is the most efficient way I know to keep a robot on the line and smooth.
2. Notice the long straightaway in the middle of the board. He computed the time it would to take to run the line at the robots fastest possible speed. The when he hit the straightaway, he went full speed until just before the end of the straight section. This alone cut his time by three or four seconds.
surely he hardcoded the robot to go a certain path? if not how can the robot tell exactly where to turn? the fastest robot won't work in a random maze, unlike some of the slower robots.
Can u plz help me with the code?
Please help me with the same for my competition
@@mohsina9958 did you miss the first run? It's obviously slowly scanning the course and storing that path for the next run.
@@mohsina9958 nope, it works
anyone could tell he was going to win just by reading his shirt and seeing his sweet haircut
just by that head you already know it's the GOAT
While you were tweaking your bot, I was studying the maze.
Not even cut
Jealousy can be ugly too.
@@TheAllcreatorLiveArchives jealousy? Where
The student used the left hand rule and the algorithm I posted as a PDF last year.
He probably spent in excess of 100 hours improving the standard algorithm. The robot keeps track of all the long straight portions of the maze, so that the second time through, it can be speeded up considerably. The algorithm predicts when a turn is coming and slows down just before the intersection.
RoboticsProfessor can u plz mail me the code at sdas9587@gmail.com plz plz
where can I find the PDF ?
2:21
Following the left hand path puts the robot in league with the devil. This algorithm is also sub optimal. It might work well for this maze but it won't be best for all possible mazes. If you watch the learning run it does not explore the whole maze. It could therefore overlook a shorter route than the one discovered. A better algorithm would retrace and explore missed turnings and the whole maze before determining the best and shortest route.
@@nathanaelsmith3553 A better algorithm for a different case, yes. This is an optimization problem, not a generalization problem.
What you do in engineering is entirely dependent on your goals. Yes, generally, you want the most generally useful system you can make, but sometimes you have a highly specific use case and you want to make a device that optimally fills it, regardless of what that costs in other domains of performance.
If, for example, I were designing a hammer, I could increase its general utility by adding things like bottle openers and screwdrivers, but every feature I add that does not help it put nails in better limits its narrow use as a hammer. In making the tool more general, its fitness in each domain of performance became more limited. If I need a hammer that opens bottles and turns screws, that's the hammer I'll make, but if I need to put nails in at peak efficiency that hammer is worthless to me.
i think the robots would finish way faster if they skipped the black line and went straight to the end tbh
Line Speedrun Any% Cheats
That would be an interesting troll strategy for keks
This is a glitchless/no exploit run.
Wallhacks
@@2010ngojo Would be funny if they programmed one to have a cheat switch, and when the judges laugh and say "WTF was that?!" proclaim, "Oh, I thought this was any%?"
The one that messed up did so for the first time at 4:39, but someone touched that spot at 3:34 in this video. This is enough evidence to warrant further investigation into tampering with the track. The person might otherwise be smoothing out an air bubble or observing some other inconsistency with the line in that spot.
Great catch
Oils from the finger made the line too reflective for the camera to pick-up correctly?
I guess, on the second run, the robot is allowed to use just the correct turn direction at each intersection (from the first run),
but is still required to scan the path, right?
What about cutting corners?
Max motor power?
Weight?
Sabotage!
Brilliant scrutineering.
13 years too late
Congrats on getting algo'd over a decade later
They seem to do the search by hugging the left wall of the maze.
You could troll them by creating a circular maze where turning to the right on the first intersection gets to the finish in just another turn, but turning to the left on the first intersection finds a really long twisted path that eventually ends up at the finish line as well.
(Of course if the rules are that the maze will have no loops, then the only "trolling" you can do is to make the left-wall-hugging maximally long, while the actual path to the finish is very shortly after turning right in the first intersection.)
If there are no loops then the robot will still learn the shortest path by using the left hand rule.
If it remembers the distances and its recent turns, it could determine that it arrived at a previous coordinate.
maybe it's even dfs
If the maze has no loops and maximise left wall hugging as you said they can simply programme it to hug the right wall
@@songwriteropgt1196 If they don't know the maze in advance, how would they guess? (Also, you could maximize both branches.)
"The tweaker is back"
Seems like a really fun environment with awesome people!
How come the track was tamped with during the last run? look at around the time 3:33 someone put a finger over the exit place that the confusion happens.
@joshay Probably
Me getting recommend after 13 years🔥🔥still cool
same here lol
Same 😄
Me wondering if a black man can do this
The robot algorithm vs the RUclips algorithm. Congratulations, you’ve been RUclipsd.
A voltage doubler or tripler and then regulated back down? I've only seen either a battery/motor combo slightly overrated for the task and then simply regulated down, or using a regulated switching boost converter, which are used in just about everything else ever built that's regulated.
It's using a regulated boost converter. I don't know why he phrased it like that, although he is technically correct, it's done at the same time.
@@Bubu567 ah ok, thank you for the insight 👍. Yea his way would definitely work, but the above is a bit simpler.
@@Qui-9 I think it's for weight! Small capacity batteries weigh less! It's surely an ineffective way, but for that short race it's perfect!
I don't think it's good for the professor to show bias by hyping up one of the students, talking about how fast his robot is and speaking enthusiastically about everything it does. But then when the other student goes, he hardly says anything at all. It could be discouraging for the other students to see the professor playing favorites.
I agree, but also this video is 13 years old
@Dino Sauro it's a 13 year old video, of course it's behind times.
Lol I mean, if you aren't interesting you should probably learn to deal with it is all
@Dino Sauro These are not children. The professor shouldnt spoonfeed you the answer on what to code. You have to learn it yourself.
They shouldnt give a shit if someone did better than them. If look closely some of the students are even smiling because at least one of them passed the test. It got nothing to do with ego thing. Either you pass or not.
You also didnt notice that the professor is actually explaining to the other students how did the guy coded his one. If youre one of the students there, you should be taking notes.
@Dino Sauro Happy and excited to see what the students who work hard achieve. => Showing bias and favouritism.
How can some people still think big ass grown students actually need teacher to mandatory praise them to do good. Your study, your efforts it's not the teacher fault that he excited to see someone can do better. Professor know dude good and give praise, not good no praise it's that simple. Imagine hyping sb up for them to fail miserably, how does that sound to you. What worse, fail or fail when everyone except you to do good? He know what his students is like and react accordingly, and also give out lesions on how dude achieve that speed. If you need someone attention and praise to do good, you need to grownups.
Did that finger rub on the track at the exact spot the second racer messed up on have anything to do with the failure?
I love it !! It’s amazing to see wat software can do!
"Okay you're tweaking" LMAOOO
Could you describe the robotos movement knowing the mass and friction constant as single transfer function? Then you could calculate the break off frequency and give an estimated break distance for any given speed. From there you could calculate speed from the pre measured distance and the average acceleration.
You can just measure distance, my man.
I bet he just hard-coded the acceleration and deceleration after testing.
It took 13 years for me to see this 13 sec win.
3:34 SABOTEUR! He rubs something on the track to mess up the speed run. It's the exact same spot where the bot messed up. He thought nobody was watching.
Real
Huh
@joshay too much for a coincidence
If this was taught in my highschool,
I would've attended regularly. 😅
I doubt it. A good student attends regardless.
RUclips's recommendation algorithm has pulled up some real treasure from the depths of the past this time.
This brought on my PTSD. 🐁
Argh! The maze! NOOOOOO!
All high schools and maybe even earlier should have competitions like this. They do very similar national competitions in Japan (they run in a maze instead of following a line) and the tiny robots move at lightning speed once the maze layout is learned. I'm not kidding. Check this out: ruclips.net/video/76blllun09Q/видео.html
Theyre doing lines because the effort is going towards warehouses and the streets.
He made that 13 Years Ago !!!!!!!!!!!!!!
I'm Beyond Amazed !
There's someone standing between window and track casting shadow and light onto the track on the spot where the robot fails. Maybe it wouldn't have failed with more uniform lighting
i dont think that matters as the left side has some shadows too, the lines are dark enough
PROFISSIONAL.... 🇧🇷👏🇧🇷👏🇧🇷👏🇧🇷👏🇧🇷👏
If you are ever lost in a maze, keep turning left and you will find your way out.
I entered a 3-D maze over 2,964 years ago; I still haven't found my way out. 🙃
Always turn left works fine as long as nobody builds in a loop that you enter with a left turn... infinite looping!
@@peter9477 In order for that to happen, someone would have to make a new path after you've gone past that part, converting a straight path into a T shaped path, connecting to the infinite loop.
@@BeeEatingOrchid You're right about that... unless the goal lies somewhere inside the area formed by that loop, in a disconnected island. I don't know what terms are used to classify mazes but this would be one of several forms involving a cycle (loop) but it's true that not all mazes with loops would be impossible for always-turn-left.
I'm just getting into engineering, so I apologize if this is a dumb question: how is the voltage from the batteries being boosted and then regulated, and how does that make the power average out no matter the battery voltage? Transformer or Op Amp or what?
Probably a switching regulator. Given a pair of AA batteries, probably a boost converter. It'll run at a constant voltage till the batteries can't output any more current (then probably stop working suddenly)
@@jordanb722 it looks like just a LC circuit with a diode, how does the diode factor in?
@@l8dawn Boost converters? The diode prevents the high voltage side from backfeeding the supply - When the transistor is open, current flows source - inductor - transistor. When the transistor closes, the current flows source - inductor - load. The load usually includes a capacitor, and without the diode the capacitor would be at a higher voltage than the source and backfeed it. The diode prevents that reverse current flow (which is probably not harmful, in the case of batteries, but does mean that you can't actually, well, boost the voltage and makes the whole arrangement rather pointless.
@@jordanb722 I'm still kinda confused but thank you; I'll ask my professor later
3:37 interesting how that's the spot where it has issues on the second run... 🤨
may i know what algorithm he use? Thanks!
Following left wall? I was four when you made this comment lol
@@pyrenees2695 bro came back
Depth first search with backtracking when a dead-end is reached
@@pyrenees2695 haha
dfs search left direction first
That compare guy, I needed a guy like that at school. Great stuff
This seems simple but its amazing probably took a good amount of work
Seems like the first action it takes when it detects multiple possible direction is left, straight, then right.
Actually simpler. The rule is just to take the left-most path you haven't already been down. (When coming back out of dead-end, that means the next attempt will look to you like it's now trying "straight" relative to the original path, but it's still just left relative to the direction it's heading at that moment.)
Hello: ? about the 13 sec robot. The Professor said something about a circuit that maintained a constant 9 volts until the battery dies. Could some one please show me the circuit and the component requirements. I just bought a BOE-BOT a couple of weeks ago and have had trouble with getting the same function like turn left. They are not similar or the same. The clam is that a constant 9 volts will help this problem. I would like to try the circuit. Thank You Paul
No idea pal, but I hope you figured it out by now
@@cirogarcia8958 11 years later 😭🤣
@@usainbolt6186 Better late than never!
@@cirogarcia8958 yes 🗿
See you guys in 3 years when this pops up in the recommended again
Ecstacy of gold plays as everyone freezes in shock. Even the robots would turn to watch who stepped in. Out cries #1 "That's impossible! You've been gone for less then 5 minutes!!!" Even without a face, everyone could see the robot smirk as it was announced: "THE TWEAKER HAS RETURNED"
Great job!
That was so fricken cool
This is pretty cool!
the dirty blonde hair in a long pony tail, the unshaven stubble, the white tee, and of course- the glasses... he is the chosen one, foretold since ancient times- destined to take the programming world by storm. his name? probably scott.
nah he tweakin
Congrats!
Turns left at each intersection. Smart. That's how I do mazes.
The guy speaking sounds like Walter White! I love it!
Теперь я знаю как пройти любой лабиринт
i know if u cant send the code then please atlest can u tell me what logic have u used.. for eg. PID or anything elsE.???
neat, i need to get into robotics and more into coding.
The robot algoritm was like: left, straight, right blablablabla
Really cool! That left side algorithm can be improved by having the robot skip all left turns until it hits a dead end/intersection, then have it turn left or turn around and turn left.
I would love to see a future version of this where just before the the end their is a orbital track around the end as that can be a issue with depth first graph search on any type of cyclic graph. I’m also assuming that they don’t get a Manhattan distance to make a heuristic out of either since both robots shown turn left down paths that increases the Manhattan distance before those those that decreases it , just I’m just curious if the robot is able to work out the nodes it’s expanded or is only capable of backtracking
Oils from the finger made the line too reflective for the camera to pick-up correctly?
I guess, on the second run, the robot is allowed to use just the correct turn direction at each intersection (from the first run),
but is still required to scan the path, right?
What about cutting corners?
Max motor power?
Weigth?
@@maxsolo2652 my comment has nothing to do with any real world effects this is looking at a flaw that is in the depth first search on a cyclic graph without expanded node tracking as a orbital path around a end point can result in the search not finding a solution.
How do I sign up for this lol
Wow such a positive environment
Почему я смотрю это в 3 часа ночи (реально, уже пол 4)
Little robot: What is my purpose?
Me: You do the maze as fast as you can.
Little robot: Oh my God...
Why does the robot always check left first, v right? I get why randomized doesn't make sense, but I'd say anticipating a clockwise circuit would be a good idea if the problem track was likely to resemble real world preference.
E.g. most F1 circuits run clockwise, not sure why, maybe "handedness" or another mental bias. I'm not sure about racing circuits in general, or whether that handedness could directly relate to the setting of these kind of robotics problems, but I'd start by turning right at the first corner, myself.
Amazing video, the winner is a true champ.
Rule #1 in maze running, select a wall. It really doesnt matter which.
I think mazes have to do with sword wielding in the past where if you hug the right wall so that your right shoulder touches the right wall you can hardly or just can't swing your right arm to the side but hugging the left wall allows if not maximizes the amount of the space to swing a sword with your right arm.
@@meh6513 good point, I remember castle staircases would spiral clockwise upwards, so a defender heading down had the advantage
@@defenda1 yep, that was my inspiration for my psuedo history xd
Heh. Pretty sure this has nothing to do with swords. It's not like having to run through mazes with swords was a common pastime, aside from in Greek mythology and Dungeons & Dragons...
The teacher seems a little biased haha he's like this teams #1 fan haha!
приоритет налево,если сдеать отвод влево и там квадрат,там будет олений хоровод
It is something like micromouse competition for beginners
Didnt know FitMC was into tech. Awesome.
oh my god i hate you now i can't unhear him
Wait this is 13 years ago whoa
2:48 is that Peter Sripol?
Better acceleration and jerk than my printer. Good job.
Is it Jeb?
The team that couldn't finish also kept starting ahead of where the team that finished did... That would have seem to give them a time advantage.
I thought it was gonna *solve* the maze in 13 seconds
This video is 13 years old, but the oldest comment that I can find as of the end of June 2022 is just three weeks old (beginning of June 2022), meaning the earlier comments were either deleted or comments were disabled up to then.
So for the first time since 13 years ago we can discuss the cheater at 3:33.
you could improve it by predicting what turn would come next, in the case of the several left turns in a row, have it learn to expect such a thing. Some kind of pattern recognisition. It's amazing that whenever a human tries to design something random, like that track, it ends up resembling music
Now we go on team number oooooone Onii chann
Fact : That Robot Always Turns Left
Yeah man, it really does. It explores all the left paths but sometimes skips the right paths
@@sambitbasu6261 Left hand rule(or right if you like), always stick to the left and you eventually reached the end.
Nice
In the meantime in Japanese Robotics tournaments: “Omae wa mou shindeiru“
Apply some tire gripper on it and we will see.
Wow....... 13 seconds and it was 13 years ago....... 😅😅😅😅😅
左手の法則かー
It is very cool
sounds like bdoub from hermitcraft
So the winning discovery algorithm is to always try left first?
Right is an equally valid option. Just pick one or the other and stick with it.
"sheeeeeeeeesh.."
Simple left turn Clyde
Man where japan has advanced this thing were at 13 secs😂
It's impressive how students are able to do this, something i cannot accomplish lol
No doubt
Ah ain't no tweek'R ain't got no teef neever
So algorithm bring us here
Very good
the second robot always turns left instead of right lol this took a long time
Еслиб там был квадрат, никогда бы они не прошли
13 years ago, fr youtube
Using rollercoaster tycoon logic works every time.
the official time is going to be the timers-time ☝️
They all prefer the left side in the first run ^^
"Okay so you're tweeking"
That's a heavy accusation to throw around with such ease Mr.. you lift bruh?
Most of these obstacles WOULD be avoided if it preferred to go straight, Rather than turn... But as is said, The first time doesn't matter.
that is true in this scenario but not always the case which is why it is necessary to come up with an algorithm that can solve any maze not just a specific one. if you made a bot that could solve this maze by prioritizing straight lines over everything and you make a new maze suddenly the bot may not be as efficient.
@@godw1ll99 Well, Yes, But would there be a downside?
@@milokiss8276 yes because prioritizing straight lines makes it so the bot has less information about the maze making it prone to getting lost easier. the left turn rule works amazing well in maze like these but when it comes to closed loop mazes always prioritizing left hand will have the bot going in a circle forever in which a simple exception of "make no more than 3 left turns in a row" would solve. in either case prioritizing straight lines is not always going to work so the left hand rule is still preferable for any maze.
@@godw1ll99 Do you happen to have an example of something that would have a bot get "Lost"? I'm having trouble understanding how it is that "Always turn left" or "Always turn right" or "Always go forward" could possibly differentiate. It seems like an arbitrary choice, As long as you stick to that rule.
Try "leftmost unvisited path" is an easy algorithm to describe and code. "Try first the straight path if there is one, then try something else (but what?), and try something else again if you have to go back" is rather more awkward to deal with.... Anyway for this particular maze left first is suboptimal, sure, but there are suboptimal mazes you can build for any algorithm. "Random pick of unvisited options" is possibly one with a better average outcome, for most mazes they're likely to design on that table, and at least it will behave differently if you have to reset it and start over.
i want to learn how he stored the data
The second machine has a strong software bias to always try and go left.
Left hand rule, when you enter a maze, place your left hand on the wall and follow that wall. (or right, it doesn't matter) Eventually it will lead to the exit.
This video inspired Tunic.
Gordon Freeman ??
How does it know where to go
Optical sensors. And code...
@@peter9477 It just tracks the longest line? Or does it literally see the pattern. Of course it would be code.
@@StarWarsMasters I was picturing three optical sensors underneath, pointing down, to detect lines, if any, to the left, right, and ahead. Could be other possibilities, but I'm not sure we're even on the same page here yet. Maybe you were asking about something else?
@@peter9477 I mean how does it know the right way. That better?
@@StarWarsMasters On the first (slow) run it doesn't know the way yet, so it follows the simple "follow the left path until you reach the goal" algorithm. From that it determines the most direct path, without the dead ends. On the second run it follows what it memorized so has no false turns. Does that help?
Nerds couldn't use :
Tremaux's algorithm, invented by Charles Pierre Tremaux, is an efficient method to find the way out of a maze that requires drawing lines on the floor to mark a path, and is guaranteed to work for all mazes that have well-defined passages. A path is either unvisited, marked once or marked twice.
If the maze is the same every time you can create tailored algorithms that perform more optimally than the ones optimized to solve arbitrary mazes. That is what the people are doing here.
A lot of people in the comment saying there are general algorithms for this and totally missing the point of the competition. It's an optimization competition, not a generality competition.
Der Roboter ist so programmiert das er immer erst nach links geht. Wenn man den linken jedoch mit dem hauptast verbindet geht er die ganze Zeit im Kreis oder zurrück
Uow... Before Bitcoin are live...
Is it gas lol