Queensland OpenAPI
Create a project using one or more of Queensland's Open-API’s
Go to Challenge | 39 teams have entered this challenge.
Rufus! (from accounting ...)
"Are you really going to drive tomorrow?" combines a geospatial analysis of commute options with a statistical model to support an AI recommender for social nudge and incentive based behavioural change.
The return on investment has been maximised by targeting the highest value times for road users to select alternatives ahead of time. This return on investment can be measured on economic terms but also by using the NSW Human Services Outcomes Framework to realise health, social and community benefits.
We set out to ask: Are you really going to drive today? AI can predict the days when traffic is going to be really congested. Governments have good cause to invest in incentives to reduce this congestion. Small changes on the busiest days can deliver huge benefits, both in economic terms, and using the New South Wales Human services outcomes framework. A number of social nudge and incentive based systems are shown influencing transport choices.
So, how did we choose our data?
-travel patterns analysed and then cross referenced that with the tendency for some routes to get very slow at peak times.
- AI model could predict congestion and we have demonstrated a statistical relationship between rainy days and higher traffic congestion.
-ABS data allowed us to select suburbs where a number of commuters originate.
We built a custom statistical model which indicates routes where travel times are more likely to spike during congestion.
We selected 3 source suburbs of interest based on these visualisations and then used Google Maps to find the actual routes.
This allowed us to estimate the travel time for cars, trains and cyclists on each route – the car time varies a lot based on how busy it is. This supported our travel times analysis.
We can already see that when travel times spike, due to congestion, this is the time when the most benefit can be realised from switching some car drivers to alternative means.
Convincing drivers to switch modes of transport on these days is more effective than on regular days. This correlates with the analysis of spikes in travel times observed in historical data from the Queensland government.
We constructed a radar chart using the New South Wales Human Services outcome framework to look at broader benefits, on the worked example focussing just on the switch from car to train, both on quiet and busy days. A score was allocated for each scenario on a 5% switch basis.
There are two parts to our system. The first part estimates the cost of driving (or, more accurately, the benefit of removing a car) on a given day at a given level of traffic congestion
Traffic congestion is highly non-linear with the number of cars on the road. At times of high demand, removing a small number of cars from the road can make a large difference to congestion. This estimate is the benefit of removing a single car from the road, and includes things like:
• Wear and tear on the vehicle
• Time spent commuting
• Opportunity cost of missed exercise
• Likelihood of an accident (higher during peak times)
This cost then feeds into the next component of the model
The second part of the system is a model of the user. How receptive they are likely to be to different transport modalities, their social networks, etc.
Not everyone is prepared to catch public transport, ride-share, or cycle. This is a complex model of user behaviour, personality and motivations that aims to assess what opportunities there are to nudge a given user into causing less traffic congestion. Remember that we only need to reduce traffic by 5-10% to make a huge difference to vehicle flow. We start with the most effective opportunities and go from there.
The sort of nudges that the system can give:
-Attempt to spread out commuter traffic, and hence reduce congestion, on a stormy day
-Notify people of accidents and recommend alternative routes
-Encourage potentially receptive people to use public transport
-The system will use social media connectivity to facilitate carpooling and other network benefits. This will require a reputation to ensure a good experience.
-allow social media and community integration, allowing the system to facilitate ride-sharing and reputation management – this is an important feature for building trust in a rideshare system
- gives rewards, which helps build awareness and prestige
-also nudge towards active transport for receptive users.
-Sharing achievements to social media helps build awareness and prestige around other options
-The model identifies people who might consider active transport, and for whom this would be a good transport option
The system can suggest public transport alternatives to receptive people when needed. Again, it can recommend and facilitate carpooling.
The system identifies which people have similar work times and journey termini, as well as using social networks to identify people in, or close to, a person’s social network – these people will be Sam’s friends, or friends-of-friends.
Also, people get a reputation based on user-feedback – people with better reputation will be preferred.
There are also opportunities for rewards to connect with local businesses
If we tell people when congestion will be high, some will choose an alternative option.
The benefits to the community go beyond pure economic measures and align with areas of responsibility for all 3 levels of government.
Our targeted approach allows measures to be taken on the days where the return on investment will be highest.
After all you now know, Are you really going to drive today?
Description of Use Used to validate our model
Description of Use Used to estimate educational and other health benefits from activity
Description of Use Used to inform models and upper limits of costings and net benefits
Description of Use Used to determine mix and gross numbers of commuters
Description of Use This dataset contains information about the route codes used in the "Travel Times" table. By joining these data, we were able to access start/stop coordinates for each route segment, allowing us to create a map of transit times showing the skewness of transit times (how much longer it takes to drive the route segment on a busy day versus an average day)
Description of Use Used to inform the associated health benefits from "Are you really going to drive tomorrow?" combining a geospatial analysis of commute options with a statistical model to support an AI recommender for social nudge and incentive based behavioural change.
Description of Use Used the figures to determine the window and criteria for success in our rules based system.
Description of Use Used to determine the savings window of various choices on commuters, extrapolated for bike and train use.
Description of Use Used to determine community impact and disadvantage with relations to people who have no access or unreliable access to transport to feed our simulations.
Description of Use Used to write rules for congestions metrics and bases
Description of Use Used to design and optimise the user experience in our User Interface
Description of Use The 7 elements of the human services framework were assigned new values so as to be optimised for the empowerment of the the community.
Description of Use Provides an argument for our solution -- namely, that a combination of public transport and information services provides much better value than building more car infrastructure
Description of Use Table 3 supports our analysis, as quoted in the video
Description of Use This report was used to validate our modeling and conclusions. In particular, p100 supports our analysis of the costs of congestion in Brisbane
Description of Use Joined with the traffic congestion data to build a model relating increased transit times with precipitation
Description of Use This dataset was used to examine: 1) Which traffic routes in Brisbane were particularly susceptible to greatly-increased travel time on busy days (i.e. skewness of the transit time distribution) 2) create a model linking increased in transit time with meteorological effects
Description of Use This dataset was used to examine: 1) Which traffic routes in Brisbane were particularly susceptible to greatly-increased travel time on busy days (i.e. skewness of the transit time distribution) 2) create a model linking increased in transit time with meteorological effects
Go to Challenge | 39 teams have entered this challenge.
Go to Challenge | 21 teams have entered this challenge.
Go to Challenge | 45 teams have entered this challenge.
Go to Challenge | 5 teams have entered this challenge.
Go to Challenge | 12 teams have entered this challenge.
Go to Challenge | 15 teams have entered this challenge.
Go to Challenge | 15 teams have entered this challenge.
Go to Challenge | 39 teams have entered this challenge.