Creating An "Autopilot" For Lander in Perl - perl
I'm using Perl to create a simple Lunar Lander game. All of the elements work (i.e. graphical interface, user implemented controls, etc), but I cannot seem to get the "AutoPilot" function to work. This function should fly the lander to a spot that it can land (or a spot designated as a target for landing), and then safely land there. The restrictions placed on landing are the slope of the place the lander lands and the velocity that the lander has when landing. The only file I can change is AutoPilot.pm. I will post the code I am allowed to work with:
package AutoPilot;
use strict;
use warnings;
# use diagnostics;
=head1 Lunar Lander Autopilot
The autopilot is called on every step of the lunar lander simulation.
It is passed state information as an argument and returns a set of course
correction commands.
The lander world takes the surface of the moon (a circle!)
and maps it onto a rectangular region.
On the x-axis, the lander will wrap around when it hits either the
left or right edge of the region. If the lander goes above the maximum
height of the world, it escapes into the space and thus fails.
Similarly, if the lander's position goes below 0 without ever landing
on some solid surface, it "sinks" and thus fails again.
The simulation is simple in the respect that if the langer goes at a high speed
it may pass through the terrain boundary.
The y-axis has normal gravitational physics.
The goal of the autopilot is to land the craft at (or near) the landing
zone without crashing it (or failing by leaving the world).
=head2 Interface in a nutshell
When the simulation is initialized, AutoPilot::Initialize() is called.
Every clock tick, AutoPilot::ComputeLanding() is called by the simulator.
For more explanation, see below.
=cut
# if you want to keep data between invocations of ComputeLanding, put
# the data in this part of the code. Use Initialize() to handle simulation
# resets.
my $call_count = 0;
my $gravity;
my ($x_min, $y_min, $x_max, $y_max);
my ($lander_width, $lander_height, $center_x, $center_y);
my $target_x;
my ($thrust, $left_right_thrust);
my ($max_rotation, $max_left_right_thrusters, $max_main_thruster);
my $ascend_height = 980;
=head1 AutoPilot::Initialize()
This method is called when a new simulation is started.
The following parameters are passed to initialize:
$gravity, a number, describing the gravity in the world
$space_boundaries, a reference to an array with 4 numerical
elements, ($x_min, $y_min, $x_max, $y_max), describing
the world boundaries
$target_x, a number representing the target landing position
$lander_capabilities, a reference to an array with
5 elements,
($thrust, $left_right_thrust, $max_rotation, $max_left_right_thrusters, $max_main_thruster),
describing the capabilities of the lander.
$lander_dimensions, a reference to an array with
4 elements,
($lander_width, $lander_height, $center_x, $center_y),
describing the dimensions of the lander.
=head2 Details
=head3 Dimensions
The dimensions are given in 'units' (you can think of 'units' as meters).
The actual numbers can take any real value, not only integers.
=head4 World dimensions
The lander world is a square region with a lower left corner at
($x_min,$y_min) and an upper right corner at ($x_max, $y_max).
The measurement units of these dimensions will just be called units
(think about units as meters). By definition, $x_max>$x_min and
$y_max>$y_min.
The default values for the lower left and upper right corners
are (-800,0), and (800,1600), respectively.
=head4 Lander dimensions
The lander is $lander_width units wide and $lander_height high.
The coordinates of the lander are always specified with respect to its center.
The center of the lander relative to the lower left corner of the lander bounding box
is given by $center_x, $center_y. Thus, if ($x,$y) are the coordinates of the lander,
($x-$center_x,$y-$center_y) and ($x-$center_x+$lander_width,$y-$center_y+$lander_height)
specify the corners of the bounding box of the lander. (Think of the lander as completely
filling this box.) The significance of the bounding box of the lander is that a collision
occurs if the bounding box intersects with the terrain or the upper/lower edges of the world.
If a collision occurs, as described earlier, the lander might have just landed,
crashed or 'escaped' (and thus the lander failed).
The constraints on these values are: $lander_width>0, $lander_height>0,
$center_x>0, $center_y>0.
The default value for the width is 60 units, for the height it is 50,
for $center_x it is 30, for $center_y it is 25.
=head4 Forces
The gravitational force is:
$g
The thrust exerted by the engine when fired is:
$thrust
The thrust exerted by the left/right thrusters when fired is:
$left_right_thrust
=head4 Limits to the controls
Within a single timestep there are limits to how many degrees the
lander may rotate in a timestep, and how many times the side thrusters,
and main thruster, can fire. These are stored in:
$max_rotation, $max_left_right_thrusters, $max_main_thruster
=head4 Target
The target landing zone that the lander is supposed to land at:
$target_x
which returns
the string "any" if any safe landing site will do, or
a number giving the x-coordinate of the desired landing site.
Note: there is no guarantee that this is actually a safe spot to land!
For more details about how the lander is controlled, see AutoPilot::ComputeLanding.
=cut
sub Initialize {
my ($space_boundaries, $lander_capabilities,$lander_dimensions);
($gravity, $space_boundaries, $target_x, $lander_capabilities, $lander_dimensions) = #_;
($x_min, $y_min, $x_max, $y_max) = #{$space_boundaries};
( $thrust, $left_right_thrust, $max_rotation,
$max_left_right_thrusters, $max_main_thruster) = #{$lander_capabilities};
($lander_width, $lander_height, $center_x, $center_y) = #{$lander_dimensions};
$call_count = 0;
}
=head1 AutoPilot::ComputeLanding()
This method is called for every clock tick of the simulation.
It is passed the necessary information about the current state
and it must return an array with elements, describing the
actions that the lander should execute in the current tick.
The parameters passed to the method describe the actual state
of the lander, the current terrain below the lander and some
extra information. In particular, the parameters are:
$fuel, a nonnegative integer describing the remaining amount of fuel.
When the fuel runs out, the lander becomes uncontrolled.
$terrain_info, an array describing the terrain below the lander (see below).
$lander_state, an array which contains information about the lander's state.
For more information, see below.
$debug, an integer encoding whether the autopilot should output any debug information.
Effectively, the value supplied on the command line after "-D",
or if this value is not supplied, the value of the variable $autopilot_debug
in the main program.
$time, the time elapsed from the beginning of the simulation.
If the simulation is reset, time is also reset to 0.
=head2 Details of the parameters
=head3 The terrain information
The array referred to by $terrain_info is either empty, or
it describes the terrain element which is just (vertically) below the lander.
It is empty, when there is no terrain element below the lander.
When it is non-empty, it has the following elements:
($x0, $y0, $x1, $y1, $slope, $crashSpeed, $crashSlope)
where
($x0, $y0) is the left coordinate of the terrain segment,
($x1, $y1) is the right coordinate of the terrain segment,
$slope is the left to right slope of the segment (rise/run),
$crashSpeed is the maximum landing speed to avoid a crash,
$crashSlope is the maximum ground slope to avoid a crash.
=head3 The state of the lander
The array referred to by $lander_state contains
the current position, attitude, and velocity of the lander:
($px, $py, $attitude, $vx, $vy, $speed)
where
$px is its x position in the world, in the range [-800, 800],
$py is its y position in the world, in the range [0, 1600],
$attitude is its current attitude angle in unit degrees,
from y axis, where
0 is vertical,
> 0 is to the left (counter clockwise),
< 0 is to the right (clockwise),
$vx is the x velocity in m/s (< 0 is to left, > 0 is to right),
$vy is the y velocity in m/s (< 0 is down, > 0 is up),
$speed is the speed in m/s, where $speed == sqrt($vx*$vx + $vy*$vy)
=head2 The array to be returned
To control the lander you must return an array with 3 values:
($rotation, $left_right_thruster, $main_thruster)
$rotation instructs the lander to rotate the given number of degrees.
A value of 5 will cause the lander to rotate 5 degrees counter clockwise,
-5 will rotate 5 degrees clockwise.
$left_right_thruster instructs the lander to fire either the left or
right thruster. Negative value fire the right thruster, pushing the
lander to the left, positive fire the left thruster, pushing to the right.
The absolute value of the value given is the number of pushes,
so a value of -5 will fire the right thruster 5 times.
$main_thruster instructs the lander to fire the main engine,
a value of 5 will fire the main engine 5 times.
Each firing of either the main engine or a side engine consumes
one unit of fuel.
When the fuel runs out, the lander becomes uncontrolled.
Note that your instructions will only be executed up until the
limits denoted in $max_rotation, $max_side_thrusters, and $max_main_thruster.
If you return a value larger than one of these maximums than the
lander will only execute the value of the maximum.
=cut
sub ComputeLanding {
my ($fuel, $terrain_info, $lander_state, $debug, $time) = #_;
my $rotation = 0;
my $left_right_thruster = 0;
my $main_thruster = 0;
# fetch and update historical information
$call_count++;
if ( ! $terrain_info ) {
# hmm, we are not above any terrain! So do nothing.
return;
}
my ($x0, $y0, $x1, $y1, $slope, $crashSpeed, $crashSlope) =
#{$terrain_info};
my ($px, $py, $attitude, $vx, $vy, $speed) =
#{$lander_state};
if ( $debug ) {
printf "%5d ", $call_count;
printf "%5s ", $target_x;
printf "%4d, (%6.1f, %6.1f), %4d, ",
$fuel, $px, $py, $attitude;
printf "(%5.2f, %5.2f), %5.2f, ",
$vx, $vy, $speed;
printf "(%d %d %d %d, %5.2f), %5.2f, %5.2f\n",
$x0, $y0, $x1, $y1, $slope, $crashSpeed, $crashSlope;
}
# reduce horizontal velocity
if ( $vx < -1 && $attitude > -90 ) {
# going to the left, rotate clockwise, but not past -90!
$rotation = -1;
}
elsif ( 1 < $vx && $attitude < 90 ) {
# going to the right, rotate counterclockwise, but not past 90
$rotation = +1;
}
else {
# we're stable horizontally so make sure we are vertical
$rotation = -$attitude;
}
# reduce vertical velocity
if ($target_x eq "any"){
if (abs($slope) < $crashSlope){
if ($vy < -$crashSpeed + 6){
$main_thruster = 1;
if (int($vx) < 1 && int ($vx) > -1){
$left_right_thruster = 0;
}
if (int($vx) < -1){
$left_right_thruster = 1;
}
if (int($vx) > 1){
$left_right_thruster = -1;
}
}
}
else{
if ( $py < $ascend_height) {
if ($vy < 5){
$main_thruster=2;
}
}
if ($py > $ascend_height){
$left_right_thruster = 1;
if ($vx > 18){
$left_right_thruster = 0;
}
}
}
}
if ($target_x ne "any"){
if ($target_x < $px + 5 && $target_x > $px - 5){
print "I made it here";
if (abs($slope) < $crashSlope){
if ($vy < -$crashSpeed + 1){
$main_thruster = 1;
if (int($vx) < 1 && int ($vx) > -1){
$left_right_thruster = 0;
}
if (int($vx) < -1){
$left_right_thruster = 1;
}
if (int($vx) > 1){
$left_right_thruster = -1;
}
}
}
}
if ($target_x != $px){
if ( $py < $ascend_height) {
if ($vy < 5){
$main_thruster=2;
}
}
if ($py > $ascend_height){
$left_right_thruster = 1;
if ($vx > 10){
$left_right_thruster = 0;
}
}
}
}
return ($rotation, $left_right_thruster, $main_thruster);
}
1; # package ends
Sorry about the length of the code...
So, there are a few things I want this autopilot program to do. In order they are:
Stabilize the lander (reduce attitude and horizontal drift to zero if they are nonzero). Once stabilized:
If above a target and the target's segment is safe to land on then descend on it.
Otherwise ascend to the safe height, which is above 1200 units. You can safely assume that there are no objects at this height or higher and also that during straight ascends from its initial position, the lander will not hit anything.
Once at the safe height, the lander can start going horizontally towards to its target, if a target is given, otherwise it should target the first safe landing spot that is can sense by scanning the terrain in one direction. It is important that the lander maintains its altitude while it moves horizontally, because it cannot sense objects next to it and there could be objects anywhere below this height.
Once the target x coordinate is reached and is found to be safe to land on, start a descend.
If the target x coordinate is is reached, but the terrain is unsafe, if a good spot has been seen while moving towards the target, go back to it, otherwise continue searching for a good spot.
Once a good spot is seen, just land on it nice and safe.
Ok, so, I've updated the code. My code is now able to land the lander in all tests (except one, got fed up, the code works close enough) where there is no target. However, I am having huge troubles figuring out how to get the lander to land at a target. Any ideas with my code so far? (actual used code is found in the ComputeLanding subroutine)
Here's a hint: try approaching the problem from the other end.
Landing is almost equivalent to takeoff with time reversed. The only thing that doesn't get reversed is fuel consumption. (Whether that matters depends on whether the simulation counts fuel as part of the lander's mass. In a realistic sim, it should, but at a glance, it looks like yours might not.)
The optimal way (in terms of fuel efficiency) to take off in a rocket is to fire the engines at maximum power until you're going fast enough, then turn them off. The slower you climb, the more fuel you waste hovering.
Thus, the optimal way to land a rocket is to freefall (after a possible initial burn to correct heading) until the last possible instant, and then fire the engines at full power so that you come to a stop just above the landing pad (or hit the pad at whatever velocity you consider acceptable, if that's greater than zero).
Can you calculate what the right moment to turn on the engines would be? (Even if you can't do it directly, you could always approximate it by binary search. Just pick a point and simulate what happens: if you crash, start the burn earlier; if you stop before hitting the surface, start it later.)
(Ps. This seems like a rather silly exercise for a Perl programming course. Yes, you can certainly solve this in Perl, but there's nothing about Perl that would be particularly well suited for this exercise. Indeed, this isn't even fundamentally a programming problem, but a mathematical one — the only programming aspect to it is translating the mathematical solution, once found, into a working program.)
You could use a genetic algorithm for the lander implementation check out this book AI Techniques for game programming. It has exactly what you need with code examples. However, those examples are in c++.
Related
When releasing two buttons in the new Input System it detects the differences
I've been creating a game that uses the input system for moving the player _playerInput.actions["Move"].performed += x => { // We're moving _walking = true; // Set the movement input to the value passed by the player _movementInput = x.ReadValue<Vector2>(); // The angle that we need the player to rotate towards so the player look where he's moving and we change it to be in degree _targetAngle = Mathf.Atan2(_movementInput.x, _movementInput.y) * Mathf.Rad2Deg + cam.eulerAngles.y; }; Here is my code but when I use the Arrows to move like the right and up Arrow and I want to release them it detects the smallest difference between the release of the two buttons which is near impossible to avoid (difference between release) before releasing "_movementInput" contains (0.71, 0.71) but after release it becomes (0, 1) or (1, 0). I want it to stay (0.71, 0.71) as it was before releasing the two buttons
Confused about get_rays function in NeRF
I've been trying to understand NeRF. I finished reading the paper(Tancik) and watched some of the videos. I have been looking at some parts of the code. However, I can't quite wrap my head around what the get_rays function does in terms of the code. Could anybody just run through line-by-line about what each line in the the get_rays function is supposed to do? def get_rays(H,W , focal, c2w): #c2w is pose i, j = tf.meshgrid(tf.range(W, dtype=tf.float32), tf.range(H, dtype=tf.float32), indexing='xy') dirs = tf.stack([(i-W*.5)/focal, -(j-H*.5)/focal, -tf.ones_like(i)], -1) rays_d = tf.reduce_sum(dirs[..., np.newaxis, :] * c2w[:3,:3], -1) rays_o = tf.broadcast_to(c2w[:3,-1], tf.shape(rays_d)) return rays_o, rays_d
It creates two lists, rays_o represents points where rays originate (camera centre) and rays_d represents direction vectors of each ray casting through the centre of every pixel of the camera. In this case, all values in the rays_o are the same because the function gets rays from a single camera.
Multiple for loops executing at the same time? Swift
Basically I am creating a game that spawns a ball after a certain amount of time has passed. However, as time goes on and the player score increases, balls that are more difficult to kill will spawn. Prior to this, all the spawn methods I have set in place function as I would like them to, but they were contained in a switch statement that calculated the probability a ball would spawn. //pseudo: let spawnProbability = arc4random_uniform(100) switch spawnProbability { case 0…60: //60% spawn chance //spawn small ball case 61…90: //~30% spawn chance //spawn medium ball case 91…100: //~9% chance //spawn large ball default: break } This worked as a plan B, but I'd like to implement what I originally had designed which was the method described above. My thought process to accomplish what I originally had in mind is: for loop that spawns small balls until the score counter reached a certain amount after the score counter reached, say, 100pts, another for loop would be executed in addition to the first loop. So, medium and small balls spawn after a delay. after the score counter reached another mark, large balls would begin to spawn on top of medium balls and small balls. Here's what I mean: (pseudocode again) for(var i = 0; i < 10; i++) { //delay //spawn small ball } if score >= 100 { for (var x = 0; x < 10; x++) { //delay //spawn medium ball } } // . . . and so on I've put alot of thought into how I could accomplish what I want but I can't come up with a better approach to this problem. I understand why this isn't working and why multiple for loops can't be executed with delay in the same class at the same time (or maybe they can and I'm just going about it in the wrong way) but I'm stuck in the sense that I don't know a better way to implement the spawn technique that I want.. Any help is greatly appreciated!!
Physicsjs Screen wrap
I am currently having a bit of trouble making objects in my world wrap. It sort of works, but very often objets appear to get stuck on the boundaries. My wrap code is as follows: // Wrap our position if we are outside of the world bounds if (this.state.pos.get(0) > 860) { this.state.pos.set(0, this.state.pos.get(1)); } else if (this.state.pos.get(0) < 0) { this.state.pos.set(860, this.state.pos.get(1)); } if (this.state.pos.get(1) > 640) { this.state.pos.set(this.state.pos.get(0), 0); } else if (this.state.pos.get(1) < 0) { this.state.pos.set(this.state.pos.get(0), 640); } Is there a better way of doing this? Should I use a translation on the object's position vector rather than simply setting it?
Without a jsfiddle it's a bit hard to diagnose, however this might be due to the this.state.old.pos not being set too. If the position (only) is set, then the velocity will be calculated as the difference between the current and the previous positions (in accordance with verlet integration). In that case, you're implicitly giving the body a huge velocity. I'd recommend adding/subtracting an amount rather than setting, then you can do the same with the old position. Here's a working example: http://labs.minutelabs.io/Relativistic-Space-Sheep/ With the relevant line of code: https://github.com/minutelabsio/Relativistic-Space-Sheep/blob/master/library/js/mediators/boilerplate.js#L743
PyEphem Rise/Set Azimuths for Stars and Constellations
I can't seem to figure out how to calculate an azimuth of a star, let's say Sirius, when it is at alt=0 (star rise/star set). So essentially in the end I will have two values. Sirius.az # star rise Sirius.az # star set Is this possible? THANKS!
Simply check the .az attribute of the star once your observer object has determined the moment of rising or setting. For example: import ephem s = ephem.star('Sirius') boston = ephem.Observer() boston.lat = '42.37' boston.lon = '-71.03' boston.next_rising(s) print s.alt, s.az boston.next_setting(s) print s.alt, s.az Running this script at this moment gives me the encouraging output: 0:00:00.0 112:23:25.2 0:00:00.0 247:36:34.9 As you can see, the rising and setting routines have already established the moment of zero altitude, leaving the azimuth for you to read off and use!