The Ultimate Wingman for Cyclists

Cycling offers a number of benefits, like improved physical well-being, environmental preservation, and the sheer exhilaration of being outdoors and active. It is no wonder that many people select biking as their preferred mode of transportation. However, navigating through traffic on a bicycle can be hazardous at times, posing challenges even for seasoned cyclists.

When riding a bicycle in traffic, one of the most significant risks is the possibility of colliding with motor vehicles, which can result in severe injuries or even fatalities. Despite efforts to create dedicated bike lanes and other infrastructure improvements, accidents still occur frequently. According to the National Highway Traffic Safety Administration, in 2021, 966 cyclists were killed in traffic crashes in the United States alone, with thousands more suffering injuries.

Even when following traffic laws and sticking to designated bike lanes, cyclists still face a high risk of accidents. Factors contributing to this situation include distracted driving, speeding, and a lack of understanding of cyclists' rights. Poor road conditions, inadequate signage, and limited visibility further increase the dangers cyclists face when sharing the road with motor vehicles.

Clearly there is room for improvement in creating safer environments for cyclists and reducing the incidence of bike-related accidents. But the question is, how do we do that? A startup called Velo AI believes that their new product, named Copilot, could provide some meaningful help in this regard. Copilot is an AI-powered device that attaches to the back of a bicycle to keep an eye out for potentially dangerous situations. If anything of concern is detected, the system has mechanisms that are designed to get the attention of both the cyclist and the driver in hopes of avoiding an accident.

Built into the form factor of a traditional bicycle light, Copilot also comes equipped with a camera and some computing components that are capable of running computer vision algorithms. At present, this system can identify nearby vehicles and determine what sort of behavior they are engaged in. For example, it can determine when a car is following, approaching, or overtaking a cyclist. Copilot also has the potential to distinguish between types of vehicles (e.g.: a car vs. a city bus), and estimate the distance a car is going to pass by at. If these conditions appear to be dangerous, a pattern of lights flashes to get the attention of the driver, while an audible alert notifies the cyclist of impending trouble.

Computer vision algorithms can chew up a lot of compute cycles and rapidly drain batteries, so the hardware choice was crucial in building a practical device. The team chose to leverage a Raspberry Pi Compute Module 4 because of its combination of processing power and minimal energy consumption. A Hailo AI co-processor was also included in the design to accelerate the machine learning inference speeds.

In addition to the alarms, Copilot also works with a companion smartphone app. By mounting the smartphone to the handlebars, the rider can get a view of what is happening behind them.

The devices are still being tested in cities around the country at this time. If they prove that they can deliver on their promises, Velo AI’s Copilot could help to reduce bicycle accidents and give more people the confidence that they need to ride their bikes in high-traffic areas. The first shipments are due to go out soon, so we should know how well it works in the real world very shortly.