Scope

Pedestrians across the country are plugged in, listening to music or podcasts, hearing the news, and making phone calls as they walk and run.  While this can promote exercise and increase productivity, it decreases the pedestrian’s awareness of the environment around them, making them susceptible to accidents with cars, bikers, trains, and other pedestrians. The goal of Theia is to help these pedestrians by informing them about potentially dangerous entities approaching them outside of their visual field. Theia is a vest for a walker or runner to wear that communicates with their phone via Bluetooth, informing them through their headphones when there is something coming towards them from behind and what it is, allowing them to be safer and more aware of their surroundings without having to sacrifice the use of headphones.

Requirements

Theia must be a lightweight vest, weighing less than 5 pounds so it can be comfortably worn by walkers and runners.  Since it is a piece of wearable technology, it will be battery operated, although we do not know how much power we will require at this point and therefore we do not know the size of the batteries we will need. It needs to sense objects at least 8-10m away and determine if the object is moving towards the user at a speed faster than their own, as objects moving away, not moving, or moving slowly are not relevant obstacles to the user. At this point, Theia needs to capture an image of the object and categorize it as either a pedestrian, a biker, a car, or a train. Our goal is to have at least 70% accuracy in this process. The user must now be informed of the object approaching them through their headphones, and this entire process from initial detection to communication should ideally take less than 3 seconds.  The user should only be informed once on each approaching object.

We anticipate a number of challenges in implementing this. Since this product will be on the back of a moving pedestrian, capturing an image of high enough quality to effectively and accurately process it will be difficult. Because of this, we plan to start testing image captures in this environment as soon as possible to maximize time we have to develop alternative solutions if necessary. In addition, maintaining a good latency will be challenging yet important so the user can be informed of the approaching object before it passes. Keeping power consumption low such that Theia can be used for a moderate length walk or run without having to weigh the user down with excessive amounts of bulky batteries will also be a challenge.  So, as we plan our design and order parts, we will prioritize low power consumption, even if that causes more work on our part. Finally, figuring out how best to communicate with the user will be a challenging process, both because we don’t know the best technical way to do so quickly and efficiently and because we don’t know what will be ideal for the user. Because of this, we will investigate numerous ways to effectively communicate and maintain a list of alternative plans.

Implementation

Our current design is to build a wearable technology mounted on the back of a vest. We plan to use an ultrasonic sensor to detect when an object is approaching, and then use a camera to take a picture of the object (i.e. a person, biker, or car). We will probably attempt various machine learning and computer vision techniques for object recognition, to meet our time, power and accuracy requirements. We will use a Raspberry pi to interface with the sensors, and we plan to use bluetooth capabilities of the Raspberry pi to communicate with user’s phone. We plan to build a mobile application to sync with the Raspberry pi and communicate the information with the users headphones, and potentially other wearables (i.e a watch).