Toward Wireless Mesh with WiFi-Direct Apps

A project using WiFi Direct and the Andoird platform.

Paul O'Neil and Steven Presser


    In a disaster, it is not unusual for many of the infrastructure-based networks (e.g.: cellular or internet) to be damaged or destroyed. However, wide-area first responder communication and information dissemination within a disaster zone could be greatly improved by the availability of a network. Repairing infrastructure may need to wait until after the disaster situation is alleviated and mobile replacements may be impractical or spotty. However, those in the affected ares are likely to still have any battery powered devices they had which used these networks - i.e.: cell phones. We looked at how cellular devices could be used in order to create a communication network to the outside world for those who could not reach infrastructure networks and how communication within a disaster zone could be accomplished using these devices.

    We chose to use Android devices. Some newer Android devices (for example, the Nexus 4 and 7) support a protocol called WiFi Direct. WiFi Direct is intended to allow two devices to communicate without infrastructure. The target use case for WiFi Direct is simple communication between two devices. For example, WiFi Direct could be used to stream media from your phone to your television.

    We were aiming to use WiFi Direct as a link-layer transport method and build a network on top of it. Unfortunately, we found that this had significant difficulty. For example, WiFi Direct devices sometimes have very divergent views of the world. For example, one device believes it is connected to another, while the second is unaware of the existence of the first. Additionally, the Android implementation of WiFi Direct appears to only allow one to one associations between devices. Additionally, disconnecting from one device and connecting to another without manual user intervention does not appear to function correctly. For this reason, though WiFi Direct as described online appears to be an ideal solution, we were unable to implement this due to the buggy nature of the Android Implementation. Should this improve sufficiently in the future to overcome these problems, we could revisit this project and turn it into a much broader project.

    However, we are able to demonstrate that, should WiFi Direct work as advertised, our solution would be feasible. In order to do so, we've built a relatively simple demo. We start with two devices. Device one is connected to a wider network - perhaps the cell network. In a disaster situation, this would be a device at the edge of the affected area, which is able to speak to undamaged infrastructure. Device two cannot see this infrastructure network, but can connect to device one over WiFi Direct. It is then able to transmit all of its traffic to device one via the WiFi Direct link. Device one then acts as a gateway and passes this traffic to the wider world.

    The most novel part of our implementation of this idea is the requirements. We took advantage of the VPNService interface, which appeared in Android API 14 to capture the systems traffic. In this way we achieve two important goals: First, the device does not need any modification or rooting to run our system - simply an application install. Second, applications do not need to be aware of our application to use it - traffic is transparently passed through our network.


Project Presentation