Tackling climate change with machine learning [part 2] – Transportation

By loading the video, you agree to YouTube's privacy policy.
Learn more

Load video

Always unblock YouTube

On 10th of June, 2019, twenty-two AI researchers, including Andrew Ng and Yoshua Bengio, published a paper on how climate change can be tackled with machine learning. I really enjoyed reading it and I am convinced that the paper as well as the climatechange.ai initative, which emerged from it, deserve more attention. For that reason i created a series of blog posts and videos which provide a dense summary, listing many of the proposed solutions and linking research work as well as ongoing projects. In the big picture, all solutions aim to reduce greenhouse gas emissions.

As my contribution to the global #ClimateStrike week from September 20th to 27th, i will post one chapter (video and blog post) on every working day. You can subscribe to our YouTube channel or follow me on Twitter to be notified when new content is published.

This is part two of a six-part series:

1. Electricity Systems
2. Transportation
3. Buildings & Cities
4. Farms & Forests
5. Industry & Carbon Dioxide Removal
6. Datasets & further resources

Transportation

Transportation systems cause about ¼ of global energy-related CO2 emissions. Passenger and freight transportation are each responsible for about half of transport greenhouse gas emissions. Transportation includes all modes (road, rail, water, air). Two thirds of emissions are caused by road travel but emissions of air travel are on the rise.

Machine Learning can help reduce the carbon footprint of transportation by

• ↓ understanding transportation patterns
• ↓ modeling transportation demand
• ↓ estimating the impact of shared mobility concepts
• ↓ optimizing freight routing and bundling
• ↓ boosting alternatives to transport
• ↓ designing for vehicle efficiency
• ↓ enabling autonomous vehicles
• ↓ improving electric vehicles
• ↓ improving low-carbon transportation

City planners and other decision-makers often plan transportation infrastructure without sufficient information about transportation patterns.

ML can help

• count pedestrians, cyclists and cars with computer vision [Paper]
• classify roads with similar traffic patterns (support vector machines, neural networks) [Paper – SVM] [Paper – Neural Network]
• estimate average traffic by detecting vehicles on satellite and aerial imagery [Paper – Fast/er R-CNNs], [Paper – Convolutional Neural Network], [Paper – Convolutional Neural Network]

[high leverage]

Understanding transportation demand can be a good decision-making basis for policies and businesses that encourage low-carbon modes of transport.

Map of underground and rail stations in London visualised by the proportion of regular trips ending at each location [Manley et al., 2016 ]

ML can help

• by learning about the behavior of public transit users from smart card data or online booking data [Paper – DBSCAN] [Paper – Clustering] [Paper – Support Vector Regression, Neural Networks]
• detect transportation modes using GPS [Paper – Convolutional Neural Network]

[high risk]

Shared mobility concepts are on the rise and lead to reconsiderations regarding vehicle ownership. However, we don’t yet know if shared mobility will lead to lower greenhouse gas emissions in the long run. For example, if somebody starts to use car sharing instead of using public transport, the energy impact of car sharing is negative in this scenario.

ML can help

• understand the energy impact of a shared mobility concept, including if a ride sharing service is taking away customers from low-carbon transit modes [Paper – Ensemble Models]

Potential pitfall: Autonomous shared cars don’t necessarily lead to lower GHG emissions when based on combustion engines or batteries charged from a carbon emitting energy source.

[high leverage]

By bundling shipments together, the number of trips can be reduced significantly. Furthermore, optimizing routes so that e.g. freight vehicles don’t return empty reduces greenhouse gas emissions as well.

ML can help

• predict arrival times and freight vehicle demand [Report]
• identify and planning around transportation disruptions [Report]
• cluster suppliers by their geographical location and common shipping destinations
• optimize freight auctions [Paper ]

[high risk]

3D printing has the potential to reduce freight transport by producing lighter goods and moving the production closer to customers.

ML can help by

• advancing design of materials which are used in 3D printing [Paper – Convolutional Neural Network]
• detecting faults in 3D printed objects and thereby ensure quality

The energy consumption of all kinds of vehicles can be reduced through technology.

ML can help by

• improving design of advanced combustion engines [Paper – Stochastic Gradient Descent based online learning algorithm]
• improving the power management methods of hybrid electric vehicles [Paper ]
• detecting aerodynamically inefficient loading on freight trains [Paper ]
• lowering fuel consumptions of airplanes by predicting taxi time and thereby improve runway scheduling [Paper – Linear Regression, Support Vector Machines, k-Nearest Neighbors, Random Forest]

[high risk]

It is unclear if autonomous vehicles will decrease or increase traffic. There are a few scenarios where autonomy has the potential to reduce emissions.

ML can help by

• enabling trucks to drive very close together (platooning) to reduce air resistance. Their autonomy would allow them to brake and accelerate simultaneously. [Paper ]
• smoothing out traffic – although non-autonomous cars are involved – and reducing emissions caused by congestion; [Paper – Reinforcement learning] [Paper – Reinforcement learning]
• reducing emissions of last-mile delivery with small & light autonomous vehicles, such as delivery robots and drones could [Paper ]

[high leverage]

Electric vehicles are considered to be the main solution for decarbonizing transport. Batteries, hydrogen fuel cells and electrified roads & railways are all considered electric vehicle technology. These technologies can have very low greenhouse gas emissions, of course only if the electricity they consume is produced from mostly renewables.

ML can help

• improve charge scheduling, congestion management and vehicle-to-grid algorithms [Paper ]
• improve battery energy management, e.g. charge estimation [Paper – Support Vector Machines]
• detect faults during wireless charging [Paper – Neural Network]
• improve model charging behavior and informing grid operators about predicted electric load;
• improve the placement of charging stations based on in-vehicle sensors and communication data [Paper – Genetic algorithm]
• improve vehicle-to-grid technology (when car batteries act as energy storage for the grid) [Paper – Reinforcement learning]
• predict battery state, degradation and remaining lifetime [Paper ] [Paper ] [Paper ] [Paper ] [Paper ] [Paper ] [Paper ]

[high leverage]

Low-carbon options such as mobility sharing services face imbalance problems. In the case of bike sharing, bicycles accumulate in certain areas and lack completely in others, making such services less attractive. Furthermore, missing enforcement of regulations makes high-carbon modes more competitive to low-carbon options.

Bike sharing demand distribution imbalance problem in Shanghai; left: morning, right: evening [Pan et al., 2018 ]

ML can help

• lower rail maintenance costs by predicting track degradation and modelling maintenance; [Paper – Neural Networks]
• model bikeshare station usage [Paper – Clustering]
• solve the rebalancing problem by improving forecasts of bike demand and inventory [Paper – Gradient Boosting Machines] and proposing a pricing mechanism to set incentives to balance the distribution of bikes [Paper – Online Learning], also in dockless systems [Paper – Reinforcement Learning]
• detect overloaded trucks, which exist in places where regulation is not enforced and makes trucks more competitive compared to lower carbon modals, such as freight trains [Paper – Convolutional Neural Network]

Credits

Many thanks to all researchers of the paper:

David Rolnick, Andrew Y. Ng, Priya L. Donti, Lynn H. Kaack, Kelly Kochanski, Alexandre Lacoste,  Kris Sankaran, Andrew Slavin Ross, Nikola Milojevic-Dupont, Natasha Jaques, Anna Waldman-Brown, Yoshua Bengio, Alexandra Luccioni,  Tegan Maharaj, Evan D. Sherwin, S. Karthik Mukkavilli, Konrad P. Kording, Carla Gomes, Demis Hassabis, John C. Platt, Felix Creutzig and Jennifer Chayes.

→ Continue to Part 3: Buildings & Cities