MIT and QCRI Deep Learning Predicts Traffic Accidents

And forecasting generalizes to multiple cities. 

Deep Learning Helps Predict Traffic Crashes Before They Happen

MIT News, Rachel Gordon, October 12, 2021

A deep learning model trained on historical traffic crash data, road maps, satellite imagery, and global positioning system trajectory patterns can generate high-resolution crash risk maps. Scientists at the Massachusetts Institute of Technology and the Qatar Computing Research Institute (QCRI) developed the model, which yields risk maps that can define the expected number of crashes over a future period, identifying high-risk areas and forecasting future collisions. The maps are composed of 5x5-meter grid cells, a resolution that shows highway roads, for example, have a greater risk for traffic accidents than nearby residential roads, while highway ramps have higher risk than other roads. QCRI's Amin Sadeghi said, "Our model can generalize from one city to another by combining multiple clues from seemingly unrelated data sources."

Autonomous Driving from a Different Contextual View

From the most recent CACM Mag.   An Arab oriented view of the problem.  Instructional.

Autonomous Driving in the Face of Unconventional Odds   By Hesham M. Eraqi, Ibrahim Sobh

Communications of the ACM, April 2021, Vol. 64 No. 4, Pages 64-66   10.1145/3447729

Traffic accidents are a major unsolved problem worldwide. Yearly, it causes around 1.35 million deaths and 10 million people sustain nonfatal injuries9 in addition to having substantial negative economic and social effects. With approximately 90% of accidents being due to human errors, autonomous driving (AD) will play a vital role in saving human lives and substantial property damage. Moreover, it promises far greater mobility, energy saving, and less air pollution.

Despite the recent advances to achieve such promising vision, enabling autonomous vehicles in complex environments is still decades away.6 The problem turned out to be more difficult than expected and it is even harder in the extremely complex and challenging driving environments in many regions around the world, including most of the Arab region. In the coming subsections, we discuss the challenges facing the successful implementation of AD in the Arab region and map them to the four pillars of the Autonomous Vehicles Readiness Index.7 Finding solutions to those challenges will help deliver the benefits of AD to world regions that are in desperate need for it, as more than 90% of traffic deaths occur in low- and middle-income countries with Africa having the highest death rates.9 In addition, it is very beneficial to the automotive industry as 24% of the global automotive market sales are outside Europe, China, U.S., India, and Japan.5

Infrastructure challenges. The AD algorithmic pipeline employs a 'sense-plan-act' design,3 which is the basis of many robotic systems. Advanced sensors allowed for a more accurate sensing of the environment and the surrounding objects. Nevertheless, today's road infrastructure provides many open challenges to be fully solved in 'planning' the vehicle's actions based on understating the driving scene and eventually 'acting' by commanding the vehicle's control system. Based on our literature review, we conclude that road infrastructure quality is determined by 10 features: signs, marking, barriers, lane, shoulder, median, right-of-way, horizontal alignment, vertical alignment, and lighting. The number of features contributing to road safety problems increases considerably in the case of developing countries and the Arab region4 compared to developed countries. Arab-region specific problems include a combination of lack of lane markings, traffic signs, light poles, and roadside barriers. Add to that the higher traffic densities and unofficial pedestrian waiting areas, captured in Figure 1, and it creates a very challenging driving scene for the 'plan and act' algorithms. From another perspective, once adopted, a self-driving car is expected to generate more than four terabytes of data daily, while the communication infrastructure in many parts of the Arab region is not ready to accommodate such traffic requirements.  ... "