Group Projects In Fall 2018
Heist
Members: Jason Scot, Jared Entin, Patrycja Miga
Abstract:
Heist is a trap-building game. It simulates unpredictable encounters using predictable actors. The player takes on the role of museum chief security officer, and must place obstacles to hinder would-be thieves. After obstacle placement, a robber appears. The robber knows how to solve each obstacle individually, but the challenge–and interesting AI behavior–comes when the robber must solve multiple obstacles in some order. Additional layers of randomness are introduced with museum guards, who will detect and pursue the robber, as well as supply closets where the robber may hide and wait for guards to pass. The game’s most important mechanic is the iterative property it begins to take on as the player replays the level. Whether the player has won or lost, they take the lessons they have learned–about the game’s rules, actors, quirks, etc.–and refine their own approach. In this critical step, they build up a significant advantage over the non-learning robber, but curiously, the game still manages to be sufficiently challenging. Please see our project’s public/eden webpage for more information, photos, videos, and a web playable version of our game (http://eden.rutgers.edu/~jss398/).
Video Demonstration:
https://www.youtube.com/watch?v=Ip_0UTMQgOI&feature=youtu.be
GitHub Release: https://github.com/Onsco/Group9_GameScience/releases
Food Analysis Game
Members: Jiatong Li, Amogh Kulkarni
Abstract:
We designed a game to encourage users to maintain a healthy diet. In the game, the character eats whatever the user eats and grows according to the nutrients of the food. The game is mainly developed in Unity and machine learning techniques are used to recognize the foods in the image. The game can successfully recognize foods in real life and images of foods.
Video Demonstration:
GitHub Release: https://github.com/ak1132/Food-Nutrition-AR-Game
Manuscript: Food Analysis Game Paper
I-See-U Game Simulation
Members: Zetao Yu, Chenjie Wu
Abstract:
I-See-U proposes an approach to train a mobile robot to self-navigate in the grid maze, while in this process, the robot would be able to predict its opponent’s behavior and avoid them. The novelty of our work consists in combining adversarial reaction, path planning and reinforcement learning. The reward and punishment function are based on the information each agent can observe, the fixed obstacles/boundaries and heuristic predictions (i.e. approximate distance between two agents; approximate distance from current position to goal). Moreover, by setting up the curriculum learning, we are capable of training the agents’ policy step by step efficiently by seeing any improvement precisely.
Video Demonstration:
GitHub Release: https://github.com/suiboli314/ISeeU-Game
Manuscript: I-See-U Game Simulation Paper
SyDEVS Framework using Behavior Trees
Members: Nikhil Jiju
Abstract:
In this project, I look into modeling how a group of people move in a given situation. In order to accomplish this, the Steersuite library with the tree behavior implemented will be combined in a framework taking advantage of Sydevs. This framework allows to compartmentalize the simulation in a way that agent information is taken from user input and then passed into the simulation to be simulated. The result is an crowd simulation that can dynamically change the agents in it.
Video Demonstration:
Release: https://sydevsbehavior.wordpress.com/sydevs-framework-using-behavior-trees/
Manuscript: SyDEVS Framework using Behavior Trees Paper
Floor Layout Generator
Members: Pawan Alur
Abstract:
There are many games with indoor levels that make use of randomized floor layouts. Alongside that, there are a lot of research that focuses on fields like evacuation, path-finding etc. that assume that they have a large data set of floor layouts. In this project, my goal was to build a floor layout generator that creates layouts following certain restrictions.
Video Demonstration:
GitHub Release: https://github.com/alurpawan/FloorLayoutGenerator/releases
Manuscript: Floor Layout Generator Paper
Horde Shooter: Powered By Reinforcement Learning
Members: Haris Sahmanovic, Danny Abbo, Justin Guadagno
Abstract:
Our game is a top down horde shooter where the player is pitted against ai who were previously trained using reinforcement learning. The game is simple: destroy as many enemies as you can before dying. The player starts with health equaling 100, which is reduced by 1 each time they are hit by an enemy’s attack. The player receives 1 point for every enemy destroyed and the game ends when the player’s health reaches 0. At the end, the game displays the score the user received for the trial and is encouraged to play again. Our AI works by utilizing 2 brains: the walking brain and the aiming brain. Each brain records episodes in increments of 500 steps. At each step, the agent determines a reward value. At the end of each episode, the agent uses the information it has learned in order to help optimize its strategies. In the end we have a powerful AI that really tests the users limits.
Video Demonstration:
GitHub Release: https://github.com/dannyabbo/GameScienceGroup13/releases
Office Space Manager
Members: Andrew Rears
Abstract:
Office Space Manager is a game that utilizes behavior trees in Unity to create an office space which the player must design to run with optimal performance. Every virtual worker has needs that must be satisfied, such as hunger and thirst. These needs will interrupt their productivity until it is taken care of, and the workers utilize various smart objects (such as water coolers) placed around the office to satisfy their desires. Workers in this office will dynamically adjust their behavior to address these needs, and will prioritize certain needs over others. The player is in charge of placing these smart objects so that the workers will be able to work as long as possible, earning the player the most amount of money. The player can also hire more workers to increase money earned, but this will affect how crowded the office becomes and how workers will move from one area to another. The challenge of the game comes from balancing what workers need to be productive while also increasing the office workforce at a steady rate.
Video Demonstration:
GitHub Release: https://github.com/CommanderJersey/office-space-manager/releases
Lockdown
Members: Andrew Hernandez, Behnaz Ghouchani, John Cheng
Abstract:
The game we have created is called Lockdown. It is inspired by the plans that are utilized by buildings to protect its occupants from hostile agents, such as in an active shooter situation. These plans aim to direct occupants to areas in which they will be hidden or safe from any hostile agent in the building. The player can influence civilian movement by placing signs and alarms. Additionally, distractions can be placed to lure hostile agents away from their targets. We believe our game can be a helpful tool in simulating this kind of scenario and more. Other scenarios such as general evacuations or just traffic direction can be analyzed by directing crowds to and from locations in our game.
Video Demonstration:
GitHub Release: https://github.com/bg452/cs598-lockdown/releases
Traffic Simulator
Members: Troy Rubenstein
Abstract:
This project presents a framework for using publicly available street data and autonomous agents to simulate traffic on existing road networks. We use the public OpenStreetMaps project to retrieve road map data and couple it with autonomous cars in order to represent traffic that the region may encounter. We use a directed graph to represent the lanes and intersections associated with each road, and use stoplights to create simple traffic patterns. We implement our cars as composite agents, coupling both physical limitations for our cars, and simple pathfinding.
Video Demonstration:
GitHub Release: https://github.com/truben5/TrafficSimulator/releases
Manuscript: Traffic Simulator Paper
City Escape Game
Members: Roman Stashchyshyn, Jack LoCasto
Abstract:
This game was created as a final project for a computer science elective at Rutgers University. The idea for the game was started with interest in learning about crowds and how they could react in a stressful environment. Through immersive gameplay and continued testing, we were able to study how different environmental and stress factors could cause a computer driven NPC to react to danger. Our game takes the dynamic results from everyone’s gameplay and allows us to study the behavioral patterns that had higher success rates for survival. Through computer simulations, the psychology of crowd dynamics can be used to simulate these scenarios and give feedback in order to come up with better escape and behavioral plans for civilians in hazardous environments. The reality is, situations like mass shootings cannot be replicated in real life. Through the use of software such as Unity and other types of data analytics, we could do our best to study these types of environments and how people may react when presented with harm. The statistical data gained from these simulations could then potentially be used to teach others for best practices when in danger. Give our game a try!
Video Demonstration:
GitHub Release: https://github.com/jal472/CityEscapeFinalProject/releases
Building a Chatbot using Reddit
Members: Lawrence Jiang, Da Huo, Zhitong Zhao
Abstract:
A chatbot should be able to make a conversation with a human. The model has been exposed to a set of training data and learns to respond to an input from a human. This task is classified into to approaches: a generative one and a retrieval based one. Reddit is a very popular forum where conversations of all kinds of topics take place. It’s naturally a very good database for a task like building a chatbot. From a generative approach, directly extracting the comment-reply pair loses context relation. On the other hand, not many literature reported work on a retrieval based approach using the Reddit data. We try to build a dataset that can capture the context relationship. We also made an attempt to utilize the ranker in a retrieval-based approach to rank the candidates in the response pool of a generative approach.
Video Demonstration:
Manuscript: Building a Chatbot using Reddit Paper
Symbol Detection for Interior Design in Virtual Reality
Members: Mason Grosset
Abstract:
Virtual reality application that lets users decorate a level. Uses a convolution neural network to detect symbols, which are used as shortcuts to tools and settings.
Video Demonstration:
GitHub Release: https://github.com/MnMaxon/VR-Gesture-Detection/releases
Dungeons, Dragons, and Multi-Agent Reactive Planner
Members: Chris Hartley, Dean Keinan, Ryan Morey
Abstract:
We introduce the design of an expressive multi-agent system relying on the rules and mechanics of theDungeons and Dragons tabletop role playing game to drive dynamic planning and an emergent narrative. Furthermore, we document the process of implementing the system and incorporating it into a belief injection core game mechanic.
Video Demonstration:
https://www.youtube.com/watch?v=8Kad1w0ltv0&feature=youtu.be
Manuscript: Dungeons, Dragons, and Multi-Agent Reactive Planner Paper
Lecture Simulator
Members: Andrew Li
Abstract:
To improve the lecture experience a Unity built program designed to capture data regarding student knowledge retention. Additionally, there may be potential for it’s use as a more general purpose lecture supplement for use to be more convenient by student and instructor alike.
Video Demonstration:
https://www.youtube.com/watch?v=wPBxLzx46rQ&feature=youtu.be
GitHub Release: https://github.com/LithiumCore/Lecture-Simulator/releases
Manuscript: Lecture Simulator Paper
A Twist on Social Deduction
Members: Bailey Chen, Chris Gong
Abstract:
Our game is based on the well-known “accusation and deception” games such as Town of Salem, Mafia, Trouble in Terrorist Town, and many others in which one player or a group of players out of all the players are assigned the objective to eliminate all the non-murdering players. The difference between our game and the major titles just mentioned is that our game will only involve one player, who is always assigned the role of murderer, and a group of AI, whose job is to determine who the murderer is. We used knowledge representation techniques to map the agent’s knowledge base in the acquisition and filtering of new or existing information. The primary information the agents will use to solve the murder is a series of location and time clues. Using these location and time clues, a graph algorithm will be run on the map of the game to replicate the most likely path each agent (including the player) went on around the time of the murder. We believe that these techniques can be further extended to security systems where incidents happen with many people around. For example, if there is a robbery in a museum with multiple bystanders, we can use location and time clues from cameras in each room of the museum to replicate the path of a supposed robber and match that with one of the civilians present at the time. Note that the path of a robber would look different than that of a murderer and would have to be adjusted accordingly.
Video Demonstration:
https://www.youtube.com/watch?v=OpkrSVF2cTI
GitHub Release: https://github.com/chris-gong/knowledge-representation/releases
Crowd Escape Simulation – Communication
Members: Kevin Wu, Cedric Abano
Abstract:
The impact of communication on maze solving is a complex feature to analyze, but it carries significant weight in developing evacuation procedures and safe building design. Using agents generated with limited knowledge, human heuristics, and communication mechanisms, this research attempts to analyze the effects of communication on escape efficiency. The results indicate some of the challenges regarding analyzing a complex mechanism such as this, and propose viable solutions and significant factors to consider for future work.
Video Demonstration:
GitHub Release: https://github.com/dodecaphobia/CS442-Group-17/releases/tag/1.0.0
Manuscript: Crowd Escape Simulation – Communication Paper
Public Building Simulation
Members: Mo Hadaya, Stanislav Potapov
Abstract:
You are the owner of a very large room. You are able to add various items to improve the room, such as microwaves and televisions. Your goal is to think about how many people you want in your building and to devise a strategy on where to place your items. Are you okay with spending a lot to ensure happiness, or do you want to be on a low budget while making sure no one is so unhappy that they leave.
Video Demonstration:
GitHub Release: https://github.com/Yoshiyoyo/NPC-MVP/releases