Sedar Olmez

Sedar is a PhD student at the University of Leeds, department of Geography. He graduated in Computer Science at King’s College London 2018. From a very early stage of his degree, he focused on artificial intelligence planning implementations on drones in a search and rescue domain, and this was his first formal attempt to study artificial intelligence. He participated in summer school at Boğaziçi University in Istanbul working on programming techniques to reduce execution time. During his final year, he concentrated on how argumentation theory with natural language processing can be used to optimise political influence. In the midst of completing his degree, he applied to Professor Alison Heppenstall’s research proposal focusing on data analytics and society, a joint endeavour with the Alan Turing Institute and the Economic and Social Research Council. From 2018 - 2023 he will be working on his PhD at the Alan Turing Institute and Leeds Institute for Data Analytics.

Research Interests

Sedar will be focusing on data analytics and smart cities, developing a programming library to try simulate how policies can impact a small world of autonomous intelligent agents to try deduce positive or negative impact in the long run. If the impact is positive and this is conveyed collectively taking into consideration the agent’s health, happiness and other social characteristics then the policy can be considered. Furthermore, he will work on agent based modelling to solve and provide faster solutions to economic and social elements of society, establishing applied and theoretical answers. Some other interests are:

  • Multi-agent systems
  • Intelligent agents
  • Natural language processing
  • Artificial intelligence planning
  • Machine learning
  • Neural networks
  • Genetic programming
  • Geocomputation
  • Argumentation theory
  • Smart cities

The model

This repository contains the agent-based model outlined in the “Investigating the emergence of complex behaviours in an agent-based model using reinforcement learning” paper. The model consists of two types of agents, the prey and predator. The environment contains positive points and negative points which prey agents try to collect and avoid. Prey agents are trained using the novel PPO algorithm while the Predator agent uses condition-action rules. The model aims to simulate the realistic behaviours that emerge among prey agents that experience novel situations. The model contains three trained neural networks known as brains, each brain was trained under a unique set of conditions known as experiments and can be switched depending on the experiment being run. The model is developed in Unity version 2018.4.15

Running the model

  1. Once the UnitySDK folder is launched in Unity, you can navigate to predator-prey/../Scenes to launch the initial state of the model.
  2. The Unity Prefabs are GameObjects as reusable Assets. We can access the PreyAgent object by double clicking on the PreyAgent prefab. Once the prey agent is accessed, the components window on the right side allows you to swap out the Brain which initially is set to neural_network_1.nn.

The Urban Traffic Simulator is an agent-based model developed in the Unity platform. The model allows the user to simulate several autonomous vehicles (AVs) and tune granular parameters such as vehicle downforce, adherence to speed limits, top speed in mph and mass. The model allows researchers to tune these parameters, run the simulator for a given period and export data from the model for analysis (an example is provided in Jupyter Notebook).

The data the model is currently able to output are the following:

Under development.

This website uses cookies and Google Analytics to help us track user engagement and improve our site. If you'd like to know more information about what data we collect and why, please see our data privacy policy. If you continue to use this site, you consent to our use of cookies.