Student Works


Samriddho Ghosh

MDes 2023

A novel crowdsourced alternative navigation system for safer pedestrian travel.


Pedestrian safety in Berkeley has been a concern for the last few years especially post- pandemic when the city has seen a surge in unsafe activities. While the city maintains relatively low violent crime and homicide rates compared to national averages, there have been increases in certain types of crimes and assaults.

Continued emphasis on community engagement, targeted crime prevention strategies, and addressing underlying socioeconomic factors is crucial in further enhancing pedestrian safety in Berkeley. Hence, the first part of the project explores data collection and processing where pedestrian subjects are invited to objectively rate various points within the city based on their perception of safety, clubbed with synthesis of WarnMe crime data and streetlight data of the city that would give a safety-rated map of Berkeley. A custom algorithm is designed to provide alternative routes optimizing distance and safety would provide the user with routes for safer navigation.

This work focuses on an implementational approach to pedestrian safety navigation that shall help users get a better understanding of the routes they take, enforcing collective responsibility and safer experiences.


This thesis pioneers a more tailored approach to pedestrian routing. The author has developed “Godspeed”, an web based navigation platform that suggests alternative walking routes aiming to balance pedestrian safety and trip distance. It works by tapping into both objective crime stats and subjective crowd-sourced insights around locations’ based on the concept of perceived safety.

What sets Godspeed apart is how it quantifies “safety.” The system combines three core data sources: 1) Crowdsourced ratings where residents score locations from 1 ̧10 based on how safe they perceive them to be, 2) O3⁄4cial crime reports sent through UC Berkeley’s WarnMe system for incident locations and timing, and 3) City streetlight data to favor better-lit paths at night. These factors produce an overall “safety score”.

Godspeed’s backbone: an ever-evolving database where Berkeley residents anonymously tag locations from 1 ̧10 based on personal safety comfort. This granular block-by-block sentiment mapping provides a dynamic community pulse on perceived security risks impossible through fixed crime analytics alone. Yet reducing complicated psychosocial impressions
into discrete data points has risks. To keep the complex “safety quotient” honest, two additional datasets balance out crowdsourced biases:

First, researchers tap real-time incident reports from UC Berkeley’s WarnMe urgent alert system, extracting geocoordinates using Large Language Models. Machine learning digests information from raw warning emails to pinpoint recent crime locations and timing within just the past 48 hours. Second, the team incorporates impartial streetlight data showing poorly illuminated areas to avoid once sunset hits. Together these three data channels encompass both shared wisdom and tangible statistics for the algorithm to weigh.

Final Design

The tool’s launch represents an exciting step toward pedestrian planning centered on human experience over pure efficiency. Still, precisely quantifying the many subjective variables influencing perceived safety in diverse communities remains an ongoing challenge.

But by proving the concept of safety-conscious navigation, researchers have set the stage for future iterations to build on the approach using advanced perceptual computing. Features like personalized threat thresholds are now imaginable. Godspeed ultimately signifies a small leap toward technology in better harmony with users’ holistic priorities – helping to make everyday urban walking less nerve-wracking one route at a time. The implications likely extend far beyond this single city too.

Full listing