Enhancing Urban Mobility via AR Location-based Parking Navigation System

In this repository, we provide:

• Codes to reproduce all of our results
• Talk about its limitations in real-time content attachment for navigation using ARCore Geospatial API
• Project report that describes all the methods used in the development of this project

Introduction

The rapid advancements in mobile technologies have generated significant interest in Location-Based Services (LBS). As the reliance on outdoor navigation systems within mobile applications continues to grow, conventional map navigation can be upgraded through the strategic integration of Augmented Reality (AR). Meanwhile, by leveraging landmarks and pathways, AR-based navigation systems hold the potential to enhance user experience to unprecedented levels. This work presents a novel approach combining an automatic camera-based parking space detection method with an AR routing tool, resulting in an interactive driving navigation system. The proposed solution combines Machine Learning methods for parking space detection with advanced navigation systems, offering benefits beyond time-saving for drivers.

Methodology

The proposed system is comprised of five major components. A Smart Parking Space Detector, a Coordinate System Mapping, a Map and Routing component, an AR-based Interactive Navigation component, and a Server. The system pipeline is demonstrated below.

• Smart Parking Space Detector

  • Detects the location of each parking spot in a parking lot
  • Classifies the occupancy of each spot
  • Uses camera-based Machine Learning methods
  • Images are captured from an overhead CCTV camera
  
  

• Coordinate System Mapping

  • A mapping component that creates a single coordinate system to work through
  • Establishes a connection between Smart Parking Space Detector's local coordinate system and the global geographic coordinate system (latitude and longitude)
  • Uses camera calibration method to find a transformation matrix
  • Obtains Google Earth's geographic information for parking spots in a parking lot
  
  

• Map and Routing

  • Determines the optimal route to the closest available parking spot
  • Relies on the Mapbox Navigation SDK, which offers a range of features including route calculation and turn-by-turn instructions
  • Leverages the waypoints associated with each step to incorporate visual cues and create an interactive real-time route guidance experience
  
  

• AR-based Interactive Navigation

  • Sends real-time directions and guidance through an immersive AR experience
  • Allow users to follow directional arrows, virtual signage, and dynamic path indicators projected onto their smartphone
  • Leverages the routing data result of the map component and integrates it with an AR interface
  • Places heading arrows along the way at every waypoint in the current step, which is close to the user’s location
  • Attaches a parking visual marker pointing to the parking spot when the driver reaches the destination
  
  

• Server

  • Facilitates data transfer between the parking space detection module and the user’s device
  • Is responsible for processing and mapping prediction results obtained from the detection component
  • Sends the global location of the closest available parking spot to the user’s device for downstream tasks
  • Leverages the Flask framework to ensure smooth communication
  • Uses AWS web service for module deployment
  • Uses Nginx web server to manage incoming HTTP requests and Unicorn3 for process handling

Results

Some screenshots of the application during interactive navigation are demonstrated below.

App Main Page	Curved Path Navigation	Straight Path Navigation	Destination Marker Placement

Limitations

The experiments conducted to evaluate the proposed system have provided valuable insights. While the system generally performs well in accurately placing visual cues and providing interactive driving guidance, there are challenges related to reliability and stability, particularly in localization and object placement. The limitations of VPS technology in highways and roads, as well as high-speed driving scenarios, also highlight the need for further improvement.

Citation

@misc{kimia2023arbased,
      title={Enhancing Urban Mobility via AR Location-based Parking Navigation System}, 
      authors={Kimia Afshari},
      year={2023}
}