Back to Projects

Team Name:

Best Käse Scenario


Team Members:


Evidence of Work

HAiLO

Project Info

Best Käse Scenario thumbnail

Team Name


Best Käse Scenario


Team Members


Logan Wood , Lizzie , Alison , David

Project Description


HAiLO is our solution to help keep Victoria's roads safe. It's an AI web tool with an interactive map that utilizes datasets across many sectors to find the main contributing factors of road crashes and provide suggestions. These suggestions can be anything from who to target in a marketing campaign to which intersections need changing to reduce road risks. This increases efficiency & transparency of data by providing easy access to all relevant information in a user-friendly form.


#ai #roads #crash #safety #interactive #maps

Data Story


The Issue

Reducing the road toll is not just about getting smaller numbers in some list, it's about saving lives. Organisations such as the DTP, VicRoads, VicPol and the Traffic Accident Commision (TAC) are doing a great job at preventing road fatalities, and we want to further support them with the best resources to continue with their work.
But there is so much data, and no practical way to use it, so how can we do this?

Our Solution

We've come up with a Ai bot and map that can help government officials visualise and utilize important data. Ask it anything and it will do it's best to help, providing fact-based suggestions (with the statistics to back it up!) as well as providing a pdf in executive brief format.

HAiLO is currently assisted to make more accurate recommendations by pulling in the excellent crash data set released by DTP, and a related data set about traffic signals.

HAiLO is built to be easily expanded upon, with the ability to integrate with separate AI agents, solely focused on areas such as:
* Parks and Land Use data, and Animal distribution data: Are wombats likely to wander across the road anywhere in gippsland, or just in a particular belt of farmland?
* BAC, fatigue: Would better bus travel in specific areas help reduce accidents from tired or intoxicated patrons returning home?
* Public Holiday dates: If many accidents happen on public holidays, how many of them are preventable by improving public holiday transport and would the cost be justified?
* sunrise/sunset: One of our Bendigo case study intersections involved right hand turn accidents, but interestingly they were mostly around 3pm-6pm (was the sun in their eyes?)
* Proximity to hospitals, ambulance response times: Does this change the odds of a major injury becoming a fatality? Is it statistically significant?
* Street trees: are large trees likely to be blocking the view of a driver approaching the intersection
* Speed Limit zones: the speed zone at the time of the accident is already covered by the crash data, however it would be good to ask a GPT to systematically review the areas with changed speed limits to identify if there have been less or no further accidents, which would be excellent. (OpenStreetMap commit data may help identify the time frame of speed limit changes if the data isn't available elsewhere)
* Population or car registration data: identify crash stats by LGA, combined with population to work out per-capita riskiest and where interventions may have best payoff
* Preventable Mortality: were there any risk factors related to cardiac, mental health or other conditions which became road incidents? Prevention for these measures will be different (e.g. heart screening for at-risk groups, mental health support).

Challenges Entered

  • AI applications using Open Road Crash data - This is our main challenge. We are creating an AI agent. Due to technical difficulties with locally run models (we spent hours trying, I promise), we're currently using Google Gemini 1.5 flash as a stand-in. ChatGPT 4o was also been used to explore the data as it incorporates a number of sub-models and has done an excellent job crunching a reduced dataset for Bendigo, creating some example briefs. For the curious, the example generated briefs from the Bendigo data are included in the Github repository under 'AI-generated-briefs', but these weren't generated by HAiLO.
  • AI in Governance - this project aims to build transparency and trust by interpreting data, and making recommendations (in the form of PDF briefs) which give their statistical reasoning behind recommending various interventions, based on historical data.

  • Smart Mobility: Optimizing Urban Infrastructure for a Sustainable Future - One of the things we realised on reading this challenge is that identifying ways to support the adoption of newer, safer cars and more public transport (Bus home from the pub) is well within the scope of this AI.


Evidence of Work

Video

Project Image

Team DataSets

Whole Fleet Vehicle Registration Snapshot by Postcode

Description of Use Density of car types is an interesting data set to ask the AI about, as you can make an argument from absence - e.g. an area with more cars but less accidents is probably a safer area than one with less cars and a similar number of accidents

Data Set

Victoria Road Crash Data

Description of Use This is a highly detailed, very useful data set. It gives information which we have demonstrated can help identify key information: * what locations are accidents occurring in * main age brackets of pedestrians involved in fatalities (hint: it's not kids!) * what the most dangerous maneuvers were (cough - right hand turns - cough) * how many wombats (73) and roos ( were injured in road crashes in the time covered by the data *

Data Set

Black Spot Program - Victoria 2024/25

Description of Use This data shows currently planned fixes to intersections and is made available as training data for analytical GPTs to reduce or hopefully prevent recommending interventions which are already planned, or to assist with focusing on other locations.

Data Set

Traffic Lights

Description of Use When making recommendations around light cycle changes etc it is helpful to have the traffic signal data available

Data Set

Victorian Integrated Survey of Travel and Activity (VISTA)

Description of Use Making this info available to GPT for analysis

Data Set

GovHack 2024 Brief Template

Description of Use HAiLO is intended to analyse road and crash data. This brief format is used to output identified recommendations for improving identified risk areas (intersections and road segments) and preventing future accidents.

Data Set

Challenge Entries

AI applications using Open Road Crash data

How might we leverage road crash statistics and multi-agent AI-based web applications to enhance road safety and inform policy making?

#Innovative solutions for Analysis, Risk Control and Presentation of Road Crash Statistics

Eligibility: Open to everyone. See the expected outcomes and the use of Crash Stats data listed above

Go to Challenge | 13 teams have entered this challenge.

AI in Governance

How can governments use AI to boost efficiency and transparency in public sector operations while addressing concerns regarding ethics, data privacy, and public trust?

#AI in Governance

Eligibility: Open to everyone.

Go to Challenge | 35 teams have entered this challenge.

Smart Mobility: Optimizing Urban Infrastructure for a Sustainable Future

How might we use data insights to promote the development of sustainable urban infrastructure and reduce dependency on private vehicles?

#Designing sustainable urban mobility solutions

Eligibility: Use at least one dataset from data.vic.gov.au

Go to Challenge | 26 teams have entered this challenge.