Innovative Bioluminescent Proteins: A Breakthrough in Non-Invasive, Multifunctional Biological Imaging

Completely artificial proteins that produce bioluminescence can serve as a non-invasive method for bioimaging, diagnostics, drug discovery, and more. Bioluminescence is the natural chemical process of light creation in some living creatures that makes fireflies flicker and some jellyfish glow. Scientists have long been interested in borrowing the secrets of these animals' light-producing genes to
HomeEnvironmentUnlocking the Impact of Protected Bike Lanes: A Machine Learning Insight

Unlocking the Impact of Protected Bike Lanes: A Machine Learning Insight

A recent study utilizes machine learning to tackle a challenging issue: determining the optimal locations for new protected bike lanes to maximize their benefits. The research team has incorporated advanced machine learning techniques and optimization methods to guide such decisions, which called for innovative computational strategies.

A recent study from researchers at the University of Toronto Engineering employs machine learning to tackle a complex question: where is the best location for new protected bike lanes to achieve the greatest benefits?

“Currently, some individuals have excellent access to protected biking paths, allowing them to cycle to work, shops, or recreational spots,” explains Madeleine Bonsma-Fisher, a postdoctoral fellow in the Department of Civil & Mineral Engineering and the lead author of a paper published in the Journal of Transport Geography.

“Increasing the number of lanes could enhance the variety of destinations reachable, and earlier studies indicate that this could lead to an uptick in cycling trips.

“Nonetheless, many individuals lack sufficient access to safe cycling routes, which restricts their mobility. This situation prompts the question: should we prioritize the number of connected destinations and potential trips overall, or should we focus on maximizing access for the highest number of people?”

Bonsma-Fisher and her colleagues — including co-supervisors Professors Shoshanna Saxe and Timothy Chan, along with PhD student Bo Lin — utilize machine learning and optimization to provide clarity on such decisions. Addressing this issue required novel computational techniques.

“This specific optimization challenge is known as an NP-hard problem, signifying that the computational requirements to solve it increase dramatically with the network’s size,” notes Saxe.

“Employing traditional optimization methods on a city the size of Toronto would lead to failures. However, Bo Lin developed an innovative machine learning model capable of analyzing millions of combinations of over 1,000 different infrastructure projects to identify where to build impactful new cycling routes.”

Using Toronto as a representative example for large, car-centric cities in North America, the researchers created maps of potential bike lane networks along major roads, optimized using two general strategies.

The first strategy, termed the utilitarian approach, aimed at maximizing the number of trips achievable via protected bike lanes in under 30 minutes, without consideration for who benefits from those trips.

The second strategy, referred to as equity-focused, sought to increase the number of individuals with at least some access to the cycling network.

“Optimizing for equity results in a broader, more dispersed map that doesn’t concentrate heavily on downtown areas,” Bonsma-Fisher points out.

“While this approach does enhance accessibility in more parts of the city, it may yield a slightly reduced overall increase in average accessibility.”

“There exists a trade-off,” Saxe explains.

“This trade-off is temporary, assuming that an extensive cycling network will eventually cover the city, but it is significant for our current actions and could persist due to ongoing challenges in establishing cycling infrastructure.”

Another vital discovery was the identification of certain routes that seemed crucial regardless of the strategy employed.

“For instance, the bike lanes along Bloor West appear in every scenario,” Saxe adds.

“These bike lanes benefit not just nearby residents but serve as a critical link for enhancing both equity and utility within the bike network. Their consistent importance across all models challenges the notion that bike lanes are merely localized issues affecting only nearby individuals. Our optimized infrastructure consistently shows benefits for neighborhoods that are quite far away.”

The research team is already sharing their findings with Toronto’s city planners to assist in making informed infrastructure investment decisions. They also aspire to apply their analytical methods to other cities in the future.

“Regardless of your local challenges or the decisions you make, it’s essential to have a clear understanding of your objectives and to evaluate whether you are achieving them,” advises Bonsma-Fisher.

“This type of analysis can offer an evidence-based, data-informed way to tackle these difficult questions.”