EVs pose no greater risk to pedestrians than conventional vehicles
image:
Graph showing pedestrian casualty rate per billion vehicle miles travelled, by vehicle propulsion type and year.
Credit: Professor Zia Wadud, University of Leeds.
view moreCredit: Credit: Professor Zia Wadud, University of Leeds.
Electric vehicles (EVs) are no more dangerous to pedestrians than traditional petrol or diesel cars, according to a new study.
Research by the University of Leeds into UK collisions involving pedestrians and cars found no significant difference in the pedestrian casualty rates between EVs and conventional vehicles.
It also found that in these crashes, injuries sustained by pedestrians were no more severe when caused by an EV than a non-electric car - despite the heavier weight of EVs.
The paper ‘Comparing pedestrian safety between electric and internal combustion engine vehicles’ by Zia Wadud, Professor of Mobility and Energy Futures at Leeds, is published today (December 9) in Nature Communications.
It found that with around 250 billion miles driven by cars in the UK each year, the average pedestrian casualty rates were 57.8 for EVs and 58.9 for non-electric vehicles, per billion miles of driving, between the years 2019 and 2023.
Professor Wadud, based in the Institute for Transport Studies and School of Chemical and Process Engineering at Leeds, said he hoped the findings would dispel any misconceptions around electric vehicles’ safety.
He added: “There were two worries about EVs and road safety. First, whether EVs would increase the number of collisions with pedestrians because they were quieter than traditional vehicles.
“Second, where there is a collision, whether the injuries to the pedestrians would be more severe when involving an EV because the vehicles are heavier. Our results show that this is not the case.”
Better safety technologies
One possible explanation for the findings, Professor Wadud suggests, is that because most of the EV fleet is much newer and more expensive, the vehicles generally have better safety technologies than most internal combustion engine vehicles on the road today, which help them to evade crashes or limit impact.
EVs typically weigh about 0.3 metric tonnes more than conventional cars due to heavy battery packs - an additional weight of around five washing machines. This prompted concerns they could cause more severe injuries to pedestrians. However, the study found no statistical evidence that EV-related injuries were more severe.
Early EVs were initially known for being very quiet, which raised fears about more low-speed accidents involving pedestrians. However, since July 2019 all new types of electric and hybrid vehicles must be fitted with Acoustic Vehicle Alerting System (AVAS), meaning they emit a sound when moving, reducing the risk.
Hybrid differences
The study distinguishes fully electric vehicles from hybrids, which combine some battery power with combustion engines. Previous research often grouped hybrids with EVs, which Professor Wadud believes skewed results.
In this study, hybrids showed higher pedestrian casualty rates than EVs and conventional vehicles - 120.14 per billion miles. Professor Wadud contends this could be due to their substantial use as private hire vehicles in the UK. This means they clock up far greater mileage than the average car, and are predominantly driven in and around city centres, increasing the chance of crashes involving pedestrians.
However, while hybrids are involved in more collisions, injuries tend to be less severe than those caused by conventional cars.
Larger vehicles and injury severity
The risks to vulnerable road users posed by sports utility vehicles (SUVs) have been highlighted in some news reports. While this study did not look into the casualty rates of SUVs, it found that large SUVs did increase the likelihood of a more severe injury to pedestrians in a collision.
Professor Wadud said: “We should worry less about the potential dangers of electrified vehicles and more about the growing prevalence of SUVs on the nation’s roads. Whether electric or conventionally powered, these larger, heavier vehicles not only pose greater safety risks, they also take up more road space and emit more carbon over their lifecycle.”
Greater understanding
Electrifying vehicles is seen as a major pathway to reducing greenhouse gas emissions from transport. EV use is now actively encouraged by Government policies in many countries, including the UK. As such, EV numbers have been growing rapidly, so it is more important than ever to understand their wider impacts.
Professor Wadud said: “One of the ways we can fight climate change is by decarbonising transport and drivers switching to EVs is an important aspect of that.
“These findings suggest we can reassure the public and policy makers that not only are EVs better for the planet, but they also pose no greater risk to pedestrians than current petrol or diesel vehicles on the road.”
The study analysed collision data from Great Britain's STATS19 road safety database - the official Department for Transport dataset used to record and analyse road traffic collisions reported to the police, across the country. It used the most recent figures available, from 2019 to 2023.
According to these, 71,979 pedestrians were hit by cars, taxis or private hire vehicles in that timeframe. Of these, hybrid vehicles were responsible for 5,303 pedestrian casualties (7.36%), while electric vehicles were responsible for 996 pedestrian casualties (1.38%). The remaining 65,680 incidents (91.25%) involved conventional vehicles.
Although casualty figures for EVs and conventional vehicles differ significantly, when miles driven and vehicle volume on the road are considered, their casualty rates are very similar.
The figures combine slight and serious injuries, plus fatalities. The study also developed a separate statistical model to compare injury severities across the vehicle groups.
Professor Wadud said that although current EVs are found to be just as safe as internal combustion engine vehicles being driven on the nation’s roads, future research should investigate whether that would still be the case if both had similar levels of safety technology.
Journal
Nature Communications
Method of Research
Data/statistical analysis
Subject of Research
Not applicable
Article Title
Comparing pedestrian safety between electric and internal combustion engine vehicles
Article Publication Date
9-Dec-2025
New multi-modal AI framework brings human-like reasoning to self-driving vehicles
image:
Comparison of different autonomous driving systems. (a) is rule-based with manually defined rules, (b) is data-driven but lacks diversity in training data, and (c) integrates large language model (LLM) capabilities with aligned decision states for closed-loop planning.
view moreCredit: Visual Intelligence
Autonomous driving has advanced rapidly, transitioning from rule-based systems to deep neural networks. Yet end-to-end models still face major deficits: they often lack world knowledge, struggle in rare or ambiguous scenarios, and provide minimal insight into their decision-making process. Large language models (LLMs), by contrast, excel at reasoning, contextual understanding, and interpreting complex instructions. However, LLM outputs are linguistic rather than executable, making integration with real vehicle control difficult. These gaps highlight the need for frameworks that combine multi-modal perception with structured, actionable decision outputs grounded in established driving logic. Addressing these challenges requires deeper research into aligning multi-modal reasoning with autonomous driving planners.
A research team from Shanghai Jiao Tong University, Shanghai AI Laboratory, Tsinghua University, and collaborating institutions has developed DriveMLM, a multi-modal large language model framework for closed-loop autonomous driving. The findings were published (DOI: 10.1007/s44267-025-00095-w) on 26 November 2025 in Visual Intelligence. DriveMLM integrates multi-view camera images, LiDAR point clouds, system messages, and user instructions to produce aligned behavioral planning states. These states plug directly into existing motion-planning modules, enabling real-time driving control while generating natural-language explanations of each decision.
DriveMLM tackles a core challenge in LLM-based driving: converting linguistic reasoning into reliable control behavior. The framework aligns LLM outputs with the behavioral planning states used in modular systems such as Apollo, covering both speed decisions (KEEP, ACCELERATE, DECELERATE, STOP) and path decisions (FOLLOW, LEFT_CHANGE, RIGHT_CHANGE, and others).
A specialized multi-modal tokenizer processes multi-view temporal images, LiDAR data, traffic rules, and user instructions into unified token embeddings. A multi-modal LLM then predicts the appropriate decision state and produces an accompanying explanation, ensuring interpretability.
To support training, the team created a large-scale data engine that generated 280 hours of driving data across eight CARLA maps and 30 challenging scenarios, including rare safety-critical events. The pipeline automatically labels speed and path decisions and uses human refinements and GPT-based augmentation to produce rich explanatory annotations.
In closed-loop evaluation on the CARLA Town05 Long benchmark, DriveMLM achieved a Driving Score of 76.1, outperforming the Apollo baseline by 4.7 points, and recorded the highest miles per intervention (0.96) among all compared systems. DriveMLM also demonstrated strong open-loop decision accuracy, improved explanation quality, and robust performance under natural-language guidance—such as yielding to emergency vehicles or interpreting user commands like “overtake” under varying traffic conditions.
“Our study shows that LLMs, once aligned with structured decision states, can serve as powerful behavioral planners for autonomous vehicles,” the research team noted. “DriveMLM goes beyond rule-following. It understands complex scenes, reasons about motion, and explains its decisions in natural language—capabilities essential for safety and public trust. By combining perception, planning, and human instruction within a unified framework, DriveMLM offers a promising direction for next-generation autonomous driving systems.”
DriveMLM demonstrates how multi-modal LLMs can enhance transparency, flexibility, and safety in autonomous driving. Its plug-and-play design allows seamless integration into established systems such as Apollo or Autopilot, enabling improved decision-making without major architectural changes. The ability to interpret natural-language instructions expands possibilities for interactive driving assistance and personalized in-vehicle AI copilots. More broadly, DriveMLM highlights a path toward reasoning-driven autonomous systems capable of understanding complex environments, anticipating risks, and justifying their actions—key capabilities for deploying trustworthy AI in real transportation networks.
Funding information
The work is supported by the National Key R&D Program of China (No. 2022ZD0161300) and the National Natural Science Foundation of China (Nos. U24A20325, 62321005 and 62376134).
About Visual Intelligence
Visual Intelligence is an international, peer-reviewed, open-access journal devoted to the theory and practice of visual intelligence. This journal is the official publication of the China Society of Image and Graphics (CSIG), with Article Processing Charges fully covered by the Society. It focuses on the foundations of visual computing, the methodologies employed in the field, and the applications of visual intelligence, while particularly encouraging submissions that address rapidly advancing areas of visual intelligence research.
About the Authors
Dr. Jifeng Dai is an Associate Professor at Department of Electronic Engineering of Tsinghua University. His current research focus is on learning intelligent models from visual data for understanding the complex world. Prior to that, he was an Executive Research Director at SenseTime Research, headed by Professor Xiaogang Wang, between 2019 and 2022. He was a Principal Research Manager in Visual Computing Group at Microsoft Research Asia (MSRA) between 2014 and 2019, headed by Dr. Jian Sun and Dr. Baining Guo.
Dr. Wenhai Wang is a Postdoctoral Researcher at The Chinese University of Hong Kong. His research interests include computer vision, machine learning and large language models (LLMs) toward artificial general intelligence (AGI).
Journal
Visual Intelligence
Article Title
DriveMLM: aligning multi-modal large language models with behavioral planning states for autonomous driving
COI Statement
NA
