Currently, when a pedestrian approaches a crosswalk, he or she naturally looks both ways for incoming traffic. If there is a vehicle, the pedestrian makes eye contact with the driver and a face-to-face, human-to-human interaction occurs, signaling to the driver when to stop and the pedestrian when to cross. In a society of autonomous vehicles, this face-to-face communication system is difficult due to the replacement of human drivers with faceless machines. Autonomous vehicles must have the capacity to communicate with pedestrians in order to safely navigate urban and suburban roadways.

This project focuses on designing a vehicle-to-pedestrian (V2P) communication system that allows for the detection of pedestrians by autonomous cars and the communication of a pedestrian's intent to cross a street to an autonomous through the utilization of an iPhone and Apple Watch application.

An iOS and watchOS application was developed using Xcode & Swift. The application contained a specific mode for the vehicle and for the pedestrian, both of which communicated with a realtime database on the cloud, GPS coordinates and other location-related information, as well as cross status, was synced in real-time through the database. The pedestrian mode contained a single button that allowed the pedestrian to request to cross a street. When the button was tapped, a cross request was sent to the vehicular phone through the database, along with the pedestrian's current location information. The vehicular phone then determined whether a cross in the specific scenario was safe or unsafe, based on the breaking distance of the vehicle, calculated by the application. After this calculation was completed, a message was displayed on the vehicular phone to indicate whether to stop or to continue through the crosswalk. This message was also immediately relayed to the pedestrian's phone through the realtime database, and the pedestrian was directed either to cross or not cross. The watchOS version of the application extended the system to a pedestrian's Apple Watch. A notification was delivered to the watch when the cross was approved or rejected, and, depending on the decision, a custom vibration alert was delivered as well.

In order to simulate an autonomous car scenario, a golf cart with a human driver on a track was used for experimental testing, A pedestrian phone was placed on a music stand on the side of the track. A vehicular phone was placed on the golf cart. The driver of the cart was instructed to drive in a straight path through the 100 meter test zone and around the rest of the track several times. Data points from the phones were recorded into a CSV file on a laptop. The latitude and longitude data points were compared to their actual values through Google Maps and were determined to be accurate down to a meter. The velocity calculated by the phone was compared to the actual velocity and an average 3% difference was reported. This difference was likely due to human errors in data collection, as the speed of the golf cart varied slightly over the 100 meter course because of the human driver. Through manual pressing of the 'cross' button, the communication between the devices was determined to be working in close to real-time, which proved the concept of a phone-based V2P communication system.

In a real-life autonomous car scenario, several improvements could be made to increase the efficiency of the system. The vehicular phone could be embedded in the vehicle to increase the accuracy of the velocity & GPS coordinates and allow direct vehicular control from the application. The pedestrian's phone could potentially be substituted for a built-in device at the crosswalk, though this would require an infrastructure renovation. A WiFi-based DSRC system could be implemented. The breaking distance could be calculated directly by the specific vehicle manufacturer, as breaking distances may vary between different vehicle types. The pedestrian-side system could be extended to include support for the visually impaired through audio or haptic feedback. In addition, the vehicle could automatically recognize when the pedestrian had finished a cross through vehicular sensors, thus eliminating the need for a "Done Crossing" signal from the pedestrian. Despite all of these potential improvements, the system designed in this project proved the concept of a realtime vehicle-to-pedestrian communication system that could be implemented in autonomous vehicles in the future.