MSO – Safe Paths for All (Project 46)

Project Roadmap

Empathize

We conducted field observations across three locations — Pasir Ris, Tampines, and Kovan — during peak hours to study real behaviours of both pedestrians and AMD users. Sites were selected based on a 4:1 pedestrian-to-AMD ratio criterion.

Our observations revealed that 21% of AMD users at Pasir Ris did not ring their bell, 19% rode at excessive speeds, and up to 70% of pedestrians at Tampines did not react to a bell. We also ran an online survey and a statistical analysis, which showed that pedestrians feel just as safe as AMD users — confirming that complacency, not fear, is the root issue.

Define

From our research, we identified four key issues: lack of knowledge of the rules, lack of application in practice, lack of an avenue to report issues, and confusion over the definition of a “shared path.”

To evaluate our ideas, we developed the ALSAFE framework — scoring solutions across Adaptability, Low Maintenance, Safety, Ability to Scale, Feasibility, and Ease of Measuring Success. After scoring over 15 brainstormed concepts, we narrowed down to a two-pronged approach: outdoor physical interventions to subconsciously guide behaviour on-site, and an indoor educational experience to build lasting knowledge and measure impact.

Solution

Our two-pronged solution was deployed at 38 Commonwealth Ave and Queenstown Library.

Outdoors, research-backed tactile strips use Edge Rate theory and colour psychology to subconsciously slow AMD users, transitioning from yellow (slow down) to red (stop/dismount). Standees at key points along the path communicate three safety messages and direct users to the booth.

Indoors, a modular VR booth hosts a role-based game where players experience the shared path as either a pedestrian or cyclist, completing lesson rounds built around six key messages. A pre- and post-game survey with 17 scored questions measures knowledge transfer, targeting a 30% improvement in survey scores and a minimum score lift of 5 points per player.

The Problem: A False Sense of Safety

Our field observations across Pasir Ris, Tampines, and Kovan during peak hours uncovered behaviours that shaped every design decision. 21% of AMD users at Pasir Ris did not ring their bell, 19% rode at excessive speeds above 25 km/h, and up to 70% of pedestrians at Tampines did not react to a bell at all. At 38 Commonwealth Ave, 47% of pedestrians walked on the wrong path, rising to 61% among those using their phones.

Most strikingly, our statistical analysis showed no significant difference in how safe pedestrians and AMD users feel on shared paths (p = 0.34). Pedestrians feel just as safe as the riders zooming past them. Complacency, not fear, is the root issue, and that fundamentally shaped our approach towards education over enforcement.

From this, we identified four root causes: lack of knowledge of the rules, lack of application in practice, lack of an avenue to report issues, and confusion over what a “shared path” actually means.

How We Evaluated Our Ideas: The ALSAFE Framework

To move from research to solution, we developed the ALSAFE framework, a weighted scoring system used to evaluate over 15 brainstormed concepts across six criteria: Adaptability, Low Maintenance, Safety, Ability to Scale, Feasibility, and Ease of Measuring Success Metrics.

Safety was weighted highest (1.0), followed by Adaptability and Ability to Scale (0.8 each). This ensured our final solution could work across different path types, be replicated at other sites, and most importantly, not introduce new risks to path users.

We studied four local initiatives and international examples from Seville, Oregon, and Vancouver. A consistent finding across all cases: infrastructure alone is insufficient. Lasting behavioural change requires education, repetition, and habit formation.

On the Ground: Commonwealth Pilot

We brought our solutions to Commonwealth to bridge the gap between design and reality. Our weekend VR booth facilitated public outreach through surveys and direct resident dialogue, while the week-long tactile strip deployment served as our primary tool for sustained observation. This dual approach allowed us to monitor behavioral shifts and gather qualitative data.

A Closer Look at Our Solution

Our two-pronged solution targets both the on-site experience and off-site education of shared path users. The three tabs below break down the key components: the science behind our tactile strip design, how our VR game works, and how we measure success.

Tactile Strips

VR Game

Measuring Impact

Our tactile strips use Edge Rate theory, where thick chevron shapes placed in the rider’s peripheral vision trick the brain into perceiving higher speed, prompting instinctual deceleration. The colour system transitions from yellow (slow down) to red (stop/dismount).

The design is also inclusive. Progressive increases in arrow size across the three zones mean users can distinguish between Maintain, Slow Down, and Stop/Dismount by shape alone, without relying on colour. This ensures accessibility for the roughly 8% of male users with colour vision deficiency.

Green strips were removed from the final design as green is commonly associated with “go” and could have inadvertently encouraged AMD users to speed up. Strips were tested on a similar surface over 10 days before deployment and were confirmed to adhere firmly and remove cleanly with no residue.

Players choose to experience the shared path as either a pedestrian or a cyclist, each with tailored key messages and in-game mechanics. Pedestrians carry a virtual phone that generates distraction pop-ups, while cyclists manage a speed indicator, bike bell, and tactile strip speed zones.

The game runs in three phases. In Simulation Round 1, players navigate freely with no guidance to establish a baseline. In the Lesson Rounds, three checkpoints each teach one key message through guided tutorials with real-time feedback. In Simulation Round 2, players apply everything they have learned to complete the level without any guidance.

The environment was built from two LiDAR scans of 38 Commonwealth Avenue and rendered using Gaussian Splatting on Unity, giving players a realistic and immersive experience grounded in actual on-the-ground conditions.

Pedestrians are scored on Safety (lives lost from NPC collisions), Lane Compliance (points per second on the correct path, with penalties for crossing into the cyclist lane), and Focus (penalties for tapping virtual phone pop-ups). Cyclists are scored on Safety, Lane Compliance, and Speed Control across three tactile strip zones: Green (at or below 25 km/h), Yellow (at or below 10 km/h), and Red (at or below 6 km/h). Penalties apply for every 0.25 seconds of delay in reaching the required speed upon entering each zone.

Impact is measured through a pre and post-game survey with 17 scored questions per player, split into 9 shared questions and 8 role-specific questions. Every question maps directly to one of the 6 key messages across both roles, so any knowledge gain is attributable to specific in-game content rather than general awareness.

We track four metrics: an average score lift of at least 5 points per player, a per-message pass rate of at least 70% correct post-game, role-specific gains for both pedestrians and cyclists, and an increase in awareness of LTA’s Stay on Track campaign.

Outdoors, success is measured by observing at least 40% of AMD users slowing down at the relevant tactile strip areas. Together, these metrics give us a clear and quantifiable picture of whether our solution achieved real behavioural impact.

Building the VR Booth

The VR booth was designed with four priorities in mind: scalability, portability, ease of set-up, and safety of usage. The primary structure uses PVC pipes, chosen for being lightweight, rust-resistant, and easy to modularise. Connections between pipes are made using custom 3D-printed fittings, fabricated in-house using school resources, allowing the booth to be assembled and disassembled efficiently.

The booth measures 2.0m x 2.0m x 1.5m and houses a monitor, seating for one player, and a facilitator station. Graphics and poster boards are hung from the top pipes using string, keeping the setup tool-free and adaptable.

A Snake and Ladder game with magnetic pieces is also mounted on the booth, allowing bystanders waiting to play the VR game to engage with path safety content while they wait.

Otterly Lost: VR Game

Game Results: Simulation Score Improvement

113 players took part at Queenstown Library: 52 as pedestrians, 61 as cyclists.

Pedestrians

Average score shifted from −20.9 to +37.4, a jump of 58.3 points. Time spent on the correct path nearly doubled (+96.6%), time on the wrong path dropped by almost two thirds (−64.7%), and phone distractions fell from an average of 2.3 instances per run to zero. That last figure is telling — awareness alone was enough to fully correct the habit within a single session.

Cyclists

Average score moved from −272.7 to −37.9, an improvement of 234.8 points. Time on the correct path almost doubled (+99.1%), time on the wrong path fell by 78.4%, and speed penalties dropped by 62.2%. Scores remaining negative even in Simulation Round 2 reflects the intentionally punishing weight of the speed penalty mechanic — no amount of good lane behaviour can fully offset a slow response to a speed zone. That cyclists still struggled with speed compliance after the lesson rounds, despite clear improvement, points to a genuine real-world finding: responding to visual speed cues is a harder habit to form than staying in a lane.

Across both roles, every tracked behaviour moved in the right direction. The before-and-after structure means that improvement is directly attributable to what the lesson rounds taught, not simply familiarity with the controls.

In partnership with :

Supported by :

Acknowledgements

Our team would like to express our sincere gratitude to our collaborating partners at the Municipal Services Office (MSO) — Siti, Haziah, and Esther — and at the Land Transport Authority (LTA) — Vania, Wilken, Kelvin, and Wulan — for their guidance and support throughout this project. This work builds upon LTA’s existing Stay on Track campaign and would not have been possible without their trust in our team.

We would like to thank the National Library Board (NLB) and Queenstown Library, in particular Luojia and Aqilah, for hosting our VR booth and providing us with the space and facilities to bring our indoor intervention to life.

Our gratitude goes to the SUTD Capstone Office, especially Hasmidah, and to the SUTD Marketing team, Yvonne and Zerlina, for their support throughout the project.

We would also like to thank our suppliers: Will from Unitech, Jaslyn from Pisti Prints, and William from Simple Digital, for their support in bringing our physical components to life.

Finally, we would like to thank our Capstone instructors, Professor Edwin Koh and Dr. Bernard Tan and for their invaluable feedback and guidance across every stage of this project, from research and ideation through to our final deployment.