Skip to content

Latest commit

 

History

History
466 lines (321 loc) · 24.9 KB

README.md

File metadata and controls

466 lines (321 loc) · 24.9 KB

The University of Melbourne

Project 2 for Graphics and Interaction (COMP30019), 2020

To play, download the latest release.

Gameplay video: https://www.youtube.com/watch?v=-2TXDMclXc4.

Table of Contents

Explanation of the Game

In the year 2077, basketball has been made obsolete by a new sport, Neo-Basketball.

Neo-Basketball looks a lot like our basketball, except there's more balls, more hoops, and only slam dunks. It features the most important invention of the 21st century, the Light Grappling Hook™.

Swing and dunk your way through several NNBA™ approved levels as you attempt to set a world record score in the most anticipated game of 2021.[citation needed]

You are Kobe Sakurai the Second, hot-headed and ready to make up for your father's defeat of the 13th Space Jam of 2068. Your family's honour rests upon the upcoming game. I hope you are ready, son. The world awaits... Now get out there and JAM!

How to Play

To start the game from Unity, open the mainmenu scene and press play. We recommend the aspect ratio is set to 16:9.

A tutorial will guide you through the key game mechanics. We have implemented familiar first-person shooter controls: WASD to move, space to jump, left click to shoot your grappling hook and escape to pause. By holding down left click, your grappling hook can be used to swing or pull balls towards you. The crosshair will change colour when you are aiming at a hoop or ball.

Once you're done with the tutorial, try to build up the highest score possible on each of our levels! Score points by picking up balls, flying and dunking before the time runs out. To achieve the highest score, build up a score multiplier by picking up multiple balls without touching the ground. Your goal is to dunk all the balls in level. Be warned, if the timer goes to zero you can no longer get any more score.

You can practice or just jam in your favourite levels in sandbox mode, where balls reset, hoops don't deactivate and the time limit is removed.

Design Pillars

At the beginning of the project, after brainstorming and coming up with the initial concept, we decided to establish several design pillars to refer back to when making design decisions over the course of the project. Formalising our ideas in this way let us stay focused on a core vision for how the game should play. The design pillars were as follows:

  • Movement is fun
  • Agile movement is rewarded
  • Enter the flow state
  • Room for mastery

This cemented our concept of the game as one about centred around fast-flowing movement and mastery of the mechanics. It also guided our later design decisions, including the way we set up the score system to motivate optimal play and the intended gameplay flow. We removed obstructions to the main gameplay flow such as the necessity for precision aiming with aim assist. We discuss this further in the section on evaluation feedback.

Technologies

The project was created with Unity 2019.4.3f1.

Modelling Objects and Entities

For development simplicity, the majority of our objects are based on Unity primitives. For example, the platforms are transformed cubes and the balls are spheres. These primitives come with robust meshes and colliders. We styled these assets with custom materials, shaders and post-processing effects to suit our needs.

In some cases, Unity's primitives were not sufficient. We needed a torus and high-vertex sphere for our hoop, so we created and exported these models from Blender. We further sourced the grappling gun and skybox from the Unity asset store.

Complex entities, such as the hoop and player body, are a combination of multiple objects and colliders, saved as a prefab for use across every scene.

Graphics Pipeline

Both of our shaders are primarily vertex shaders. These apply per-vertex position transformations and shading in model space.

We used Unity's post-processing stack in our rendering pipeline. It applies full-screen filters and effects before the image is displayed on screen. We use post-processing for ambient occlusion, bloom, vignette, colour grading & grain. We also made the menu system canvases render in world so that the post processing could affect it, but also had the Score UI just be an overlay, which appears to be rendered after the post processing as it is not affected.


Without post-processing.


With post-processing.

Camera Motion

In game, the camera is rigged to the player body (actually a child of its transform). The camera rotates to follow the mouse in the standard style of first-person games.

This camera style was chosen to be very intuitive, and more suited to gameplay involving precision aiming as compared to, say, a third-person perspective. The first-person perspective is also more immediate and conducive to immersion, which helps convey more direct experiences like the visceral experience of fast motion.

Shaders

Shader 1: Grappling Rope

Our initial implementation of the grappling hook had a simple line renderer being drawn between the player's gun and the hook.

However, this had a couple of negatives:

  • It looked uninteresting.
  • It didn't provide the user feedback when they had attached their hook to a surface.

We decided to add a custom vertex shader to displace each vertex in the liner renderer based on a function of the distance with the sine function. This would make the rope have a wave appearance.

The first implementation was done rudimentarily as follows:

float distanceFromGun  = abs(distance(v.vertex, _GunLocation));
float displacement = sin(lengthA) * _Amplitude;
v.vertex += displacement;

This along with increasing the number of vertecies from of the line from 2 to 1000 resulted in:

However, this made it so the rope did not attach to the hook or the gun. We wanted a function to modify the amplitude of the sine wave when it is closer to both ends. In particular, we were looking for a function that passed through the origin and approached an asymptote. The function we chose that ended up having these properties was of the hyperbola family:

float normaliseAmplitude(float d)
{
  return (-0.3f/(d + 1.0f)) + 0.3f;
}

We apply this function to the distance of the vertex from both the gun and the hook it shoots. The resultant values are 0 at minimum when the vertex is near their respective object. The values also max out at some constant, which proves useful as we do not want the displacement amount to grow forever with larger and larger distances.

float lengthA  = abs(distance(v.vertex, _GunLocation));
float hookDistance = abs(distance(v.vertex, _HookLocation));

float gunNorm = normaliseAmplitude(lengthA - 0.98f);
float hookNorm = normaliseAmplitude(hookDistance - 0.98f);

Finally we have our displacement function, where we apply the min function over the values we calculated previously. This clamps the displacement to 0 at both ends of the rope, allowing us to connect the rope exactly to the tips of the gun and the hook.

float4 displacement = min(gunNorm, hookNorm) * sin(lengthA - 0.98f + _Time[3]*10) * _Up * _Amplitude;

Breaking down each part of the displacement now:

  • min(gunNorm, hookNorm) modifies the magnitude of the Sine wave so that it is 0 close to hook or gun, but some constant c (determined at normaliseAmplitude) at maximum.
  • sin(lengthA - 0.98f + _Time[3]*10) * _Up This is the actual sine wave, which wobbles vertically with time.
  • _Amplitude This is the property we pass in from GrappleHook.cs to manage the animation of the hook becoming taut.

Finally, this displacement is calculated for each vertex along the line of the rope: v.vertex += displacement;. With this we had achieved a very satisfying rope whipping effect.

Finally to elaborate on the taut animation i.e. making the rope pull taut once the hook is attached, the _Amplitude property is adjusted. Once the hook is attached, inside theGrappleHook.cs script we linearly interpolate from 1 to 0 in 8 FixedUpdate frames, and pass it to the shader through the property. This eventually sets the displacement to 0, and thus make the shader render it as a simple straight line, signifying that the rope is taut.

Here is the final result!

Shader 2: Hoop Ripple

Our second shader is for a wave/ripple-like effect for the inside surface of each hoop. This is achieved by displacing the height of each of the vertices on the surface of the hoop by a sine wave which is a function of the distance from the centre and the current time, generating a radially symmetric rippling effect.

This is calculated in a vertex shader as follows:

// Apply a sine wave displacement based on distance to center
float d = length(float2(v.vertex.x, v.vertex.z));
float freq = 1.0 / _Wavelength;
float offset = _Time.y * _Speed;
float t = d * freq + offset;
float h = sin(t);
v.vertex.y = _Amplitude * h;

In order to allow for lighting to be calculated properly for the displaced vertices, the normals are also recalculated. If we consider the function we are applying to be f(x, z) mapping to an (x, y, z) point, we can get the following expression:

f(x, z) = (x, _Amplitude * sin(freq * d + offset), z)

In order to calculate the normal vector, we can take the cross product of the partial derivatives df/dx and df/dz. We have

df/dx = (1, _Amplitude * cos(t) * freq / d * x, 0)

and

df/dz = (0, _Amplitude * cos(t) * freq / d * z, 1)

where t = d * freq + offset as above.

Taking the cross product gives us the vector:

(_Amplitude * cos(t) * freq / d * x, -1, _Amplitude * cos(t) * freq / d * z)

In addition, in order to prevent artifacts occurring near the centre of the mesh, we enforce that normals very near to the centre face strictly downwards, creating a small flat spot in the middle. This issue appeared to be due to an irregularity with our mesh, which was created by squashing a sphere down to be very thin and treating it as a disc.

Overall, this leads us to the following:

// Calculate new normal
float3 normal;
if (d < 0.01) {
  // Keep the centre looking consistently flat
  normal = float3(0, -1, 0);
} else {
  // Normalized cross product of df/dx and df/dz (partial derivatives)
  normalize(float3(
    _Amplitude * cos(t) * freq / d * v.vertex.x,
    -1,
    _Amplitude * cos(t) * freq / d * v.vertex.z));
}

In order to accentuate the overall rippling effect, we also vary the color of each vertex based on the height calculated from the sine wave. This is done by linearly interpolating a coefficient to multiply the color by between some set limits.

// Interpolate color based on adjusted height
float c = lerp(_ColorMinValue, _ColorMaxValue, (h + 1) / 2);
o.color.rgb = c * _Color;

Here the value of h is sin(d * freq + offset) as calculated previously.

Finally, a Phong illumination model is applied in a pixel shader using code taken from the workshops to get some simple lighting effects.

Particle Systems


The dunk effect.

The explosion effect plays when balls are dunked and the player is thrown away from the hoop. It is composed of a smoke puff, shockwave and electric embers. It uses textures for the official Unity particle pack.

The smoke puff randomly emits a small number of puffs within a spherical region. We adjusted the color over lifetime to fade to lower hues of red with increasing transparency as the puffs dissipate.


The smoke puff's colour over lifetime.

The shockwave spawns a circular disk at the base of the explosion, increases in size over lifetime, then reduces back to the original state.

Finally, the movement of the electric embers is dictated by a velocity over lifetime curve. We emit a large number from a hemisphere, with positive radial velocity. To give the effect of the embers flying back and the hoop 'resetting', the velocity over lifetime curve is set to drop to negative, which brings them back towards the center.


The ember's velocity over lifetime.

The explosion effect on deactivated hoops is muted, with fewer emitted particles and lower velocity.

Querying and Observational Methods

Demographics

# Age Gender Occupation
1 21 Male University Student
2 21 Male University Student
3 21 Male University Student
4 12 Male School Student
5 18 Male University Student
6 20 Male University Student

Ideally, it would have been desirable to survey a larger number of people with much more diverse demographics in order to receive more varied feedback, but that was difficult to do given the current situation.

Methodology

We used two Querying/Observational methods:

  • Think Aloud
  • Post Walkthrough Interview

Think Aloud

For our observational method, we gave the users a small set of instructions on how to play the game as well as an instruction to voice out their thoughts.

From there, minimal input from the interviewer was given as users attempted to play through our levels. We recorded any issues or interesting points we noticed from observing their gameplay, as well as any frustrations or suggestions they voiced.

To allow us to follow the user's thought process, participants shared their screens on Discord as we took dot-point notes. Observing the participant helped identify pain points and common mistakes as they learnt the game.

Post Walkthrough Interview

Our interview was conducted in a similar manner, but discussion about the game was held afterwards. Participants were only instructed on how to set up the demo and use the controls if it blocked the playthrough.

The post walkthrough interview questions we used were:

  • Did the game feel too fast or too slow? How was the pace?
  • Was the game easy to understand?
  • What pain points did you encounter during the game?
    • Aiming?
    • Movement?
    • Ball dunking?
  • Would you change any other aspect of the game?
    • Visuals
    • Gameplay
  • What new features would you like?
  • How do the controls compare to other games you’ve played?
  • Do you have any other thoughts you would like to discuss?

To allow interviewees to warm up, we structured the interview to sandwich the most difficult questions in the middle of the session. These open-ended questions allowed us to ask follow-up questions and ask for further elaboration on their thoughts.

Feedback and Solutions

Through our observational/querying methodologies, we were able to pick up on common complaints and misunderstandings with the mechanics of our game. These included complaints about game feel as well as visual clarity in general. Here is an overview of the most common points of feedback, along with the solutions we came up with.

Difficulty with "Floaty" Controls

A common complaint with our gameplay was that it felt very "floaty", especially when jumping and moving through air. Many players seemed to struggle with moving the character in the way they wanted.

The main cause we identified for this was that the maximum speed we had set for travel through the air was much too high. Moving too quickly led to players being unable to aim and position themselves effectively.

Therefore, we decided to reduce the maximum air speed (enforced by a "braking force" applied to the player in the opposite direction of their movement if they travel too fast). Crucially, we set different speed limits depending on whether the player was currently grappling. This allowed us to set a high maximum speed for when the player is grappling, and lower it for when they are simply airborne. This has the side effect of significantly slowing down the player when they release a grapple, which provided an improved gameplay flow of fast grapples with small pauses in between to re-orient and grapple again.

In order to provide better information to the player about their velocity and accentuate the sensation of fast movement, we also added dynamic field-of-view relative to the player's speed, increasing slightly as the player accelerates.

Difficulties with Aiming

We observed that users often committed to a ball grapple, expecting it to hit, and then were often disappointed or frustrated when they missed the ball. Since the core gameplay is based around near-constant fast and dynamic movement, aiming at small targets like the balls is particularly difficult. Additionally, we didn't want difficulties with aiming to be prominent as we did not consider it to be a core driver behind the game concept, relating to skilful movement or flow.

To solve this, we implemented an "aim assist" feature to improve aiming usability and decrease the chance of a player missing a ball and interrupting the flow of gameplay.


Aim assist on a ball.

This was implemented by first conducting a sphere cast, covering a fairly sizeable area in the general direction the player is looking and checking if a ball is present. If this succeeds, the direction vector in which the player is looking is shifted a small number of degrees (2-3) in the direction of the detected ball. (This number was chosen to match the size of the crosshair in the player's view). A raycast is then done in this direction to determine if the player has a direct line of sight to the ball. If this succeeds, the player's crosshair changes color to orange to clearly indicate that a shot hook would hit the target. If the player then fires a hook, it is aimed towards that point.

After testing out this feature, we found that it greatly improved the general flow of the game, and extended it to apply to hoops as well.

Visual Noise from Skybox

The skybox presented some issues with clarity. Since the skybox featured a similar grid-like pattern of black and bright blue to the platforms, it often made it difficult for players to make out the platforms against the backdrop.


The previous skybox.

This was solved by changing the backdrop to one which is more contrasting with the platforms and other game objects. (Mostly black, as if in space, but above the Earth to avoid just being in a black void.)

Lack of Instruction

Players mentioned that without us telling them the controls, there would be no way to discover the mechanics of the game. In response, we implemented a simple tutorial which guided the user through the controls and objectives of the game in an interactive manner.

"Orbiting" Hoops

Several players encountered an issue where they would attempt to grapple towards a hoop while having significant lateral velocity, which caused them to overshoot. Since the force being applied by the rope was constant, they then entered into a stable elliptical orbit around the hoop, and kept swinging around in circles as long as they held down the button.


Demonstration of the orbiting effect.

We had also found this issue while testing manually and found that it greatly obstructed the gameplay, as it prevents you from easily reaching the hoop to dunk a ball, an essential part of the core gameplay loop.

To fix this, we applied a positive jerk (rate of change of acceleration) to the pulling force. More simply, we made the pulling force increase the longer the player has been grappling. We set this value low for when grappling onto regular surfaces but increased it greatly for the hoop specifically. This increasing force over time causes the player to "crash" out of orbit and hit the hoop very quickly in most instances. Even if they overshoot greatly, most of the time they will only complete a few revolutions before reaching the hoop, so we considered the problem to be solved.

Score & Intended Gameplay Ignored

One recurring point that we observed in players was that they completely ignored the score number in the top right, which we had intended to be a central motivator behind the gameplay. Some players also simply sat in place, pulling all of the balls to themselves without moving, which was the opposite of the intended gameplay style.

After asking players about this, we determined that they were ignoring the score number because they did not understand how it worked or what its significance was. To help remedy this, we added an explanation to the tutorial.

We also wanted to give better feedback for actions which grant score, drawing more attention to the score value, and hence, our intended gameplay style (fast movement in the air, which is rewarded by the score system). To do this we added attention-grabbing score popups whenever the player picks up or dunks a ball. This should make the score number more transparent, especially in cases where players get less score from certain actions e.g. dunking into deactivated hoops.

Individual Contributions

Gatlee Kaw

  • Sound design
  • Contributed to hook shader
  • Whiteout effect for falling too far
  • FOV changes when moving quickly
  • World assets

Steven Nguyen

  • Score system + UI
  • Menu system + UI
  • Contributed to hook shader
  • Level design
  • Level system
  • Gameplay video

Matthew Lui

  • Hoop & balls
  • Tutorial level
  • Particle systems
  • Collision

James Dyer

  • Grappling hook (firing, pulling into walls, pulling balls towards you)
  • Spawning and collecting balls
  • Tuning movement, jumping and shooting mechanics
  • Crosshair and aim assist
  • Ripple effect shader for the hoop interior
  • Level design for the final level

External Resources

Sounds

Assets & Packages