Team 315: Cost Effective AI Integrated VR Drone Simulator

Team Members

• Morgan Skinner (CPE)
• Jackie Ou (CPE)
• Jimmy Lu (EE)
• Cameron Sayers (EE)
• Colby Hackett (ME)
• Jonathan Tooby (ME)

Sponsors

Dr. Shonda Bernadin

Dr. Migara Amarasinghe

Advisors

Dr. Marcos Vasconcelos

Abstract

This project created a low-cost drone simulator with artificial intelligence (AI) to teach K-12 students and drone enthusiasts about drone controls. Drones are often expensive, which makes it hard for schools and hobbyists to access this technology. The goal was to make learning about drone controls, programming, and AI fun, easy, and affordable, with students as the focus. The team used affordable hardware and open-source software to keep costs low. The AI helps users by guiding them to a given destination by providing the shortest distance from the start point. The design is simple for teachers to use as well as hobbyists who want a low-cost drone simulator. This simulator gives students, teachers, and enthusiasts a way to explore science, technology, engineering, and math (STEM) without needing expensive tools. This project shows learning tools can be affordable and exciting, inspiring interest in STEM and drone technology.

About

Project Overview

Welcome to our Senior Design Project at the FAMU-FSU College of Engineering! We are developing a cost-effective flight simulator that integrates artificial intelligence (AI) and virtual reality (VR) to provide realistic drone flight experiences. This innovative platform is designed to be affordable, portable, and functional, making it accessible for educational institutions and enthusiasts alike.

Collaborating Departments

This project is a collaboration between the Electrical and Computer Engineering (ECE) Department and the Mechanical Engineering (ME) Department, combining expertise in AI, control systems, and mechanical design.

Project Details

Project Scope

Develop a flight simulator that balances affordability with functional capability, enabling realistic flight experiences, with the ability to integrate A.I.

Customer Needs

• Simulate drone controls and behavior accurately.
• Ensure portability (must fit in a car).
• Integrate at least one AI model for drone operations.
• Support VR headset connectivity.
• Provide controller inputs with feedback.
• Maintain low cost for educational use.

Functional Decomposition

The functional decomposition of the drone simulator has four different functions in order to run the drone simulation.

• Controlling Drone Flight: This function is needed to allow the user to control the various aspects of drone flight. The process of the function goes as follows: the controller needs to process the user's input, then adjust thrust and direction of the drone.
• Simulate the Environment: This function is needed to place the user in an environment that applies real-world physics and obstacles a drone pilot can face. The process of the function goes as follows: Generate 3D terrain, then display this terrain.
• Integrate A.I.: This function is needed to implement the pathfinding algorithm created to show the shortest distance between two points in a maze. The process of the function goes as follows: collect the data of the maze (obstacles, start and end points), then train the A.I. model to find the shortest distance.

Concept Generation

100 concepts were generated using several techniques. The technique that generated the most concepts was the morphological chart method. The morphological chart method uses different solutions of satisfying a subsystem and combining one of each of these solutions for all subsystems gives a final design concept. The morphological chart below is the one used by Team 315 to generate 60 concepts.

The remaining 40 concepts generated used the crapshoot technique. The crapshoot technique has the members of the team come up with any method they can think of to create a concept. This method tends to have random results while also highlighting problems that may need to be solved with the design. Some concepts generated from this method were, surround users with 6 monitors to achieve VR, use a gaming-like controller in order to control the drone, and have an environment based on 3D Google Earth.

Concept Selection

After generating 100 concepts, five medium fidelity concepts and 3 high fidelity concepts were chosen. These concepts were chosen by the team based on what the team thought would be the best concepts to move forward with. Medium fidelity concepts were concepts the team had confidence in but were not the best out of the options. High fidelity concepts were concepts the team had high confidence being some of the best concepts out of the 100. After finding these eight concepts, several concept selection processes were used to determine the final design. These methods were Pugh charts, pairwise comparisons, and analytical hierarchy process used in that order.

In our Pugh analysis, we evaluated our top 5 design options for a VR-based gaming controller setup, each with different configurations. Our options included various controller types, VR settings, and development platforms, such as Unity and Unreal Engine. We established criteria based on the needs of our sponsor, including the ability to simulate a drone, provide a first-person view (FPV), ensure portability, incorporate AI functionality, offer a custom controller, and maintain low cost. Option 1: “a gaming controller with a VR display and an obstacle course created in Unity”, was set as the baseline. Each alternative was scored against Option 1 for each criterion, with scores of +1, 0, or -1 indicating whether an option was better, equal, or worse, respectively. Based on our Pugh Chart, we concluded that the best game engine to use is Unity. The best controller to use is a gaming controller, and the best environment to build is an obstacle course. All of the other options scored lower than option 1, indicating that the other design options are less desirable. With these conclusions, we decided option 1 met the needs most efficiently by balancing cost, functionality, and met requirements. Ultimately, no further Pugh charts were necessary.

The pairwise comparison compared design criteria against weighted priorities. Each criterion is compared against the others, helping the team quantify which factors should have the most influence in the concept selection process. Using these values, the team then used the analytical hierarchy process.

In our AHP process, we systematically evaluated three concepts against a set of six criteria: Simulate Performance, FPV, Portability, AI Functionality, Controller Functionality, and Low Cost. We started by creating a pairwise comparison matrix for the criteria, which allowed us to calculate their relative weights based on their importance to the project. Each criterion was rated in comparison to the others using the AHP scale, resulting in a normalized matrix where each criterion’s weight was determined. We then evaluated each concept (Concept 1, Concept 2, and Concept 3) against each criterion using pairwise comparisons, generating priority values for each alternative. These values were compiled into a final rating matrix, which was then multiplied by the criteria weights to yield an overall score for each concept. The final scores revealed that Concepts 1 and 3 were the top choices. This systematic approach ensured a consistent and objective evaluation process, highlighting Concept 1 as the preferred choice due to its high alignment with the criteria.

Our selected concept regarding our drone simulator involves using a gaming controller paired with a VR headset to enhance immersion. We will have the users navigate an obstacle course within a 3D urban environment developed in Unity. The course is intended to replicate real-life challenges a drone pilot might encounter, requiring precision and control to navigate through obstacles like rings, narrow passageways, and varying elevations. The VR setup allows users to experience first-person perspective (FPV) from the drone, giving them a realistic sense of depth, speed, and spatial awareness. This setup aims to provide an engaging, hands-on experience for users as they learn and practice piloting skills in a safe, simulated environment. Unity is a strong choice for this project, especially with its VR capabilities, versatile development tools, and robust 3D environment support. This setup in Unity should allow for a well-integrated, immersive experience that meets the project's needs for both realism and functionality in the simulated urban course.

Project Plan

After the final concept was selected, a spring project plan was created to give the deadlines for goals that needed to be completed. Below shows the plan used to create the final concept.

January:

• Code basic drone movement, environment, and test.

February:

• Create A.I. shortest path method
• Refine and detail environment

March:

• Integrate V.R. display
• Test and refine entire simulation (controls, environment, A.I., etc.)
• Complete final design

April:

• Senior design day
• Complete poster and final presentation
• Present poster and final presentation

Target Catalog

Key requirements include drone simulation, portability, AI integration, VR compatibility, controller usage, and cost-effectiveness.

Target Summary

The simulator aims to provide an immersive, affordable drone training tool with realistic physics, AI-driven pathfinding, and VR support, all while being compact enough to transport easily.

Future Work

Our next steps include:

• Spring 2025:
  • Create an obstacle course for skill testing.
  • Enhance the environment with additional details.
  • Refine and test the AI pathfinding system.
  • Implement VR headset compatibility.
  • Conduct final testing and make improvements.

Code of Conduct

Work Breakdown Structure

Preliminary Detailed Design