Utvecklingen av automatiserade fordon, särskilt för kollektivtrafik och godstransporter, erbjuder möjligheter att effektivisera drift och minska kostnader. I Sverige är tillståndsprocessen för försök med dessa fordon komplex, tidskrävande och ofta beroende av tekniska detaljer, vilket skapar hinder för implementering. I en framtid med drift utan säkerhetsförare ombord och där fjärrassistans är alternativet kommer även nya krav att ställas på regelverken kring tillstånd. 

Projektets mål var att ta fram förslag på hur tillståndsprocessen kan förenklas för AI-baserade, automatiserade bussar och godstransporter, med fokus på förare utanför fordonet. Projektet harutforskat juridiska och tekniska aspekter av fjärrhantering. Projektet (org.nr. 2024-01679) finansierades av Vinnova inom ramen av utlysningen ”Reglering och banbrytande teknik”.  

Arbetet genomfördes i samverkan mellan VTI, RISE och Linköpings kommun och omfattade tre workshops som en central metod. Workshopen följde en trestegsmodell: (1) genomgång av gällande lagstiftning och erfarenheter, (2) analys av hinder och behov för regelverksutveckling, och (3) formulering av förbättringsförslag. Projektet Ride the Future med automatiserade små bussar i Linköping (https://ridethefuture.se/) användes som utgångspunkt för att identifiera barriärer och alternativa lösningar. 

Projektet identifierade behov av:

• etableringen av en nationell reglering som inte bara beaktar försök utan siktar mot permanenta lösningar
• en tydligare och mer förutsägbar tillståndsprocess
• harmonisering med internationella regelverk och ökad möjlighet att använda utländska typgodkännanden
• regelverk och rutiner som stödjer säker och pålitlig drift utan förare ombord
• klargörande av roller och ansvar vid fjärrhantering
• utbildningskrav och stödsystem för fjärroperatörer
• digitalisering av lokala trafikföreskrifter.

Slutsatserna pekar på att Sverige behöver agera mer aktivt angående lagstiftningen och tillståndsprocessen för att möjliggöra bred implementering av automatiserad kollektivtrafik och godstransporter. Målet bör vara att lämna försöksförordningar och etablera förordningar som stöttar permanent implementering och drift med automatiserade fordon. Möjliga nästa steg med hänsyn till projektet Ride the Future och utöver det har beskrivits baserat på en SWAP-analys (Bilaga 1).

The development of automated vehicles, especially for public transport and freight transport, offers opportunities to streamline operations and reduce costs. In Sweden, the permit process for trials with these vehicles is complex, time-consuming and often dependent on technical details, which creates obstacles to implementation. Future operations without a safety driver on board and where remote assistance is the alternative, new requirements will also be placed on the regulatory framework for permits.

The project's goal was to develop proposals for how the permit process can be simplified for AI-based, automated buses and freight transport, with a focus on drivers outside the vehicle. The project has explored legal and technical aspects of remote handling. The project (org.nr. 2024-01679) was funded by Vinnova within the framework of the call for proposals "Regulation and groundbreaking technology".

The work was carried out in collaboration between VTI, RISE and Linköping Municipality and included three workshops as a central method. The workshop followed a three-step model: (1) review of current legislation and experiences, (2) analysis of barriers and needs for regulatory development, and (3) formulation of improvement proposals. The Ride the Future project with automated shuttles in Linköping (https://ridethefuture.se/in-english/) was used as a starting point to identify barriers and alternative solutions.

The project identified the need for:

• establishment of a national regulation that not only considers trials but aims for permanent solutions
• a clearer and more predictable permit process
• harmonization with international regulations and increased possibility to use foreign type approvals
• a regulatory framework and routines supporting safe and reliable operation without a driver
• clarification of roles and responsibilities in remote operation
• training requirements and support systems for remote operators
• digitalization of local traffic regulations.

The conclusions indicate that Sweden needs to act more actively regarding the legislation and the permit process to enable a broad implementation of automated public transport and freight transport. The goal should be to leave experimental regulations and establish regulations that support permanent implementation and operation with automated vehicles. 

This paper presents a focused look at findings related to driver fatigue from Use Case A (UCA) of EU project PANACEA (grant agreement number 953426). UCA considers safety drivers of autonomous shuttles. For safety and regulation reasons all shuttles have a driver present ready to intervene if needed. In practice, this means that the safety driver is responsible for ensuring the safety of both passengers and surrounding road users. 8 shuttle safety drivers (100% of those available) participated. Subjective driver sleepiness was reported daily at the start and end of each shift using the Karolinska Sleepiness Scale. Potential risk for driver sleepiness was calculated using the Bordeaux Sleepiness Scale (BOSS). Overall, sleepiness was a rare experience (mean KSS start shift = 3, mean KSS end shift = 3.1). However, those identified as potentially at-risk using BOSS had some experiences of sleepiness (KSS = 7) on some shifts. The PANACEA system uses input from sensors in the vehicle, workplace depot and on the driver, creating a holistic monitoring and assessment system. This detects professional drivers who are not fit to drive and supports them and their employers to manage the situation and adopt preventive measures. 

This pre‑study examines the technological landscape, practical experience, and regulatory requirements related to in‑vehicle detection of alcohol‑ and drug‑impaired driving, focusing on alcohol and cannabis.  

Current detection approaches fall into five categories: physiology‑based systems (breath, sweat, biosignals), tissue spectroscopy, camera‑based driver monitoring, vehicle kinematics, and hybrid systems. No single modality today offers sufficient specificity to reliably detect alcohol or drug impairment during real driving. Physiological sensing best reflects substance presence, but integration hurdles remain; behavioural sensing (oculomotor patterns, posture, vehicle control) reveals impairment but lack specificity. Hybrid systems that combine presence‑based and behaviour‑based indicators has the greatest potential. 

Practical experience from prior intoxication studies carried out by the project partners highlighted key operational requirements. Reliable experiments depend on high staffing density, strict safety protocols, disciplined event logging, and carefully tuned alcohol dosing procedures. Recruitment is challenging and requires screening for health, medications, sleep patterns, and substance use history. Breath‑based ground truth measures require rigorous protocols, and evidential‑grade analysers are recommended in future studies. Closed‑track environments remain necessary for safety but may not fully reflect real‑world driving. Simulator extensions can complement realism. Across partners, there was strong alignment on experimental design for future alcohol studies, with growing interest in shifting focus toward cannabis impairment.

Conducting cannabis experiments in Sweden requires a full clinical‑trial application through the EU CTIS system, reviewed by both the Medical Products Agency and the Ethical Review Authority. A medically responsible investigator and a compliant drug manufacturer must be secured before submission. Additional requirements include medical screening, on‑site medical staff, blood and specimen handling procedures, and detailed documentation of the cannabis product. In addition, a separate government permit is needed because driving under drug influence is illegal even on closed tracks. Recruitment will be particularly challenging due to the illicit status of cannabis in Sweden, and the method of drug administration (vaporised, smoked, capsules, edibles) must be selected to balance safety, impairment dynamics, and sensor detectability.

Taken together, the project show that: 

• Robust detection of alcohol and drugs will likely require multi‑modal sensing and careful experimental design.
• SAFER partners already share substantial know‑how for alcohol studies, but cannabis research introduces additional legal, medical, and logistical complexities. 
• A future multi‑partner cannabis study is feasible but will require early coordination with medical investigators, compliant manufacturers, and regulators, alongside realistic timelines for approvals and recruitment. 

Binge drinking of alcohol leads to worsened driving performance the following morning, even when the blood alcohol concentration (BAC) has returned to or is close to zero. The objective of this study was to investigate the effects of social drinking (BAC = 0.05%) on next-day driver fatigue and driving performance. A homogenous sample of 32 experienced male drivers drove for 35 min on rural and urban roads in a driving simulator, both the day after drinking alcohol and the day after a sober evening. The main effects on next-day performance were ambiguous, where self-assessments showed lower next-day performance and higher subjective sleepiness after drinking alcohol in the evening, whereas variability in lateral position, heart rate variability and attention showed worse next-day performance in the control condition. Overall, the effect sizes were small. The results indicate that a BAC level of 0.05% does not influence next-day performance after a full night’s sleep to any greater extent. Further studies including a placebo condition are needed to verify this result, also considering more BAC levels, long-term effects of habitual drinking, and longer driving times.

Driver sleepiness contributes to a substantial proportion of road crashes. Drivers experiencing sleepiness are advised to take a break and have a caffeinated drink followed by a short nap (caffeine nap). However, previous research advocating this countermeasure has not considered participants with obstructive sleep apnoea (OSA), the most prevalent sleep disorder. Across three studies the effectiveness of caffeine, nap opportunity and caffeine nap countermeasures on subjective sleepiness (KSS), objective sleepiness (Alpha and Theta activity) and driving performance (standard deviation of lateral position and out-of-lane events) are considered. Twenty-one CPAP treated OSA participants (mean age = 59 years) engaged with a protocol of six laboratory visits: one after a normal night’s CPAP-treated sleep and five after sleep restriction (4 h CPAP-treated sleep), driving a monotonous simulated scenario before and after a countermeasure. Results showed that two cans of coffee (255 mg caffeine) mitigated driver sleepiness more than one can (127.5 mg) and little benefit to 30 min compared with 15 min nap opportunity. An optimised caffeine nap of two coffees followed by a 15 min nap opportunity provides some temporary benefit, but for OSA drivers a caffeine nap offers little practical improvement compared to two coffees alone. All countermeasures are temporary and cannot replace a good night of sleep before driving.

Hur trötthet mäts spelar roll för trafiksäkerheten. I det här projektet har vi undersökt om själva frågan ”Hur sömnig är du?” enligt Karolinska Sleepiness Scale (KSS) tillfälligt kan pigga upp föraren och därmed påverka hur vi tolkar både kroppens signaler och körbeteendet. Eftersom KSS ofta används som facit när trötthetsdetekteringssystem (Driver Drowsiness and Attention Warning, DDAW) utvecklas och typgodkänns inom EU:s förordning 2019/2144 om krav avseende fordonens allmänna säkerhet (General Safety Regulations, GSR), är det viktigt att skattningen inte förvränger tillståndet som mäts. Projektets övergripande mål har därför varit att kartlägga den uppiggande effekten av KSS och hur länge den sitter i.  

Data från fem körsimulatorstudier med totalt 84 deltagare och 2 701 KSS-skattningar ligger till grund för analyserna. Runt varje skattning följde vi ögonrörelser (blinkmönster), hjärtslag, hjärnaktivitet (EEG) och hur stadigt bilen hölls i körfältet (SDLP). Linjära regressionsmodeller med tre segment användes för att analysera (1) läget före frågan, (2) impulsen i frågeögonblicket och (3) återgången mot förenivån. En varaktighetsparameter τ ingick i modellen och fungerar som ett direkt mått på hur länge den uppiggande effekten sitter i.   

Resultaten visar att fysiologiska sömnighetsindikatorer reagerar snabbt vid KSS-frågan. Förändringar i blick, hjärtslag och EEG pekar på en uppiggande effekt som klingar av inom cirka 30–70 sekunder, medan körbeteendet (SDLP) påverkas längre tid, uppemot 2 minuter. Effekterna är tydligast vid hög sömnighet (KSS 7–9).   

Praktiskt innebär det att data direkt efter skattningstillfället, alltså där den uppiggande effekten är som störst, bör exkluderas när algoritmer tränas och testas mot KSS. Även proceduren för självskattning, inklusive hur försöksdeltagare tränas i att använda skalan, behöver standardiseras. Genom att förbättra hur KSS används vid datainsamling och produktvalidering kan olika detektionssystem bli mer tillförlitliga, ge färre falsklarm och i förlängningen bidra till färre trötthetsrelaterade olyckor. 

How fatigue is measured matters for traffic safety. In this project, we investigated whether sleepiness ratings on the Karolinska Sleepiness Scale (KSS) temporarily alert the driver and thereby influence how we interpret both physiological and behavioural sleepiness indicators. Since KSS is often used as the reference when developing and type-approving Driver Drowsiness and Attention Warning systems (DDAW, within the EU’s General Safety Regulation 2019/2144, GSR), it is important that the rating does not distort the state being measured. The overall aim of the project was therefore to map the alerting effect of KSS and how long it lasts. 

Data from five driving-simulator studies with a total of 84 participants and 2,701 KSS ratings formed the basis of the analyses. Around each rating, we tracked eye movements (blink patterns), heartbeats, brain activity (EEG), and how steadily the vehicle was kept within the lane (SDLP). Linear regression models with three segments were used to analyse (1) the state before the question, (2) the impulse at the rating, and (3) the return toward the pre-question level. A duration parameter τ was included in the model, serving as a direct measure of how long the alerting effect persists. 

The results show that physiological sleepiness indicators respond quickly to the KSS question. Changes in gaze behaviour, heart rate, and EEG indicate an alerting effect that fades within approximately 30–70 seconds, while driving behaviour (SDLP) is influenced for a longer period, up to about 2 minutes. The effects are most pronounced at higher sleepiness levels (KSS 7–9). 

Practically, this means that the period directly following each KSS rating, where the alerting effect occurs, should be excluded when algorithms are trained and tested against KSS. The self-rating procedure itself, including how participants are trained to use the scale, also needs to be standardized. By improving how KSS is used in data collection and product validation, different detection systems can become more reliable, produce fewer false alarms, and ultimately help reduce fatigue-related crashes. 

This paper presents a new conceptual framework, and stepwise approach to populate it, for informing countermeasure development to support fitness-to-drive for professional drivers. Professional drivers are vital to the transport network; however, the job is demanding and drivers are vulnerable to impairments which may impact safe driving. Countermeasures are any action or activity that mitigates the impact or frequency of occurrence of driver impairment. The framework proposes countermeasures to be delivered across three time points: Operational (during shift), Tactical (immediately after shift) and Strategic (outside of on-shift) and at multiple system levels, e.g., driver, manager, enforcement etc. The framework was successfully pilot tested with three different professional driver use cases: autonomous shuttles, taxi, and garbage truck drivers. This structured approach to countermeasure design offers potential to improve driver health and enhance road safety. The work was conducted within , an EU project, grant agreement number .

Driver fatigue is involved in between 15 and 30 percent of all crashes. The reasons behind this increased crash risk among fatigued drivers are numerous and not only due to sleepiness caused by lack of sleep. A clear view of the magnitude of driver fatigue is needed to understand when, how and where mitigation strategies are needed. There is also a need to understand the factors that are main contributors to driver fatigue, as well as an understanding of the effectiveness of different countermeasures. Studying the connection between driver fatigue and traffic safety is challenging for several reasons. One being the diverse nature of fatigue, others are difficulties in manipulating and measuring fatigue and individual differences between drivers.  Here, we will discuss some of the general challenges related to individual differences in sensitivity to fatigue and specific challenges related to fatigue research in the driving context.  

Actions against driver fatigue can be strategic, tactical or operational. Strategic actions are the most effective since they ensure that drivers who are at risk of developing fatigue are not seated behind the wheel. On a tactical level it is important to prevent drivers from continuing to drive when severe fatigue occurs, this could be mainly through information and warning. At an operational level it is crucial to support drivers and hence stop them from driving if they, for example, are about to leave the lane by mistake. In most countries it is illegal to drive fatigued, despite this people drive under fatigue. From a legal perspective it is a great challenge to produce evidence that a crash to some degree was caused by driver fatigue. To prosecute such a case in court and sentence drivers to punishment, clear laws and strong evidence are required, something that does not exist today. There is currently no biomarker that can be used to measure fatigue in an individual with a sensitivity of 100 percent. Furthermore, fatigue is a condition that is changeable, unlike the effects of alcohol. One way to measure fatigue can be based on factors such as time of day, number of hours slept, and number of hours awake in the last 24 hours. But since fatigue can also occur for reasons other than sleepiness, such as when the driving task has been going on for too long, there is a risk of missing situations where fatigue is a contributing factor to the accident.  In the present work various challenges and opportunities in relation to enforcement and their potential to help reduce the risk of fatigued drivers behind the wheel will be presented.  

The perception of comfort and safety among passengers of Autonomous Vehicles (AVs) is crucial and significantly influences their adoption in current Public Transport systems. It is essential to align the objective perception with an analysis of vehicle performance data to identify vulnerabilities and factors affecting passenger comfort and safety. This paper presents the first comprehensive correlation between objective and subjective data from autonomous fleets in three well-established pilot locations (Graz, Madrid, Linköping), each using different technologies and experiencing varying environmental conditions. Our analysis (i) revealed significant differences between the three pilot sites in terms of perceived safety and comfort (both perceived and actual) and (ii) confirmed a strong correlation between safety and comfort levels and the vehicles’ behaviour in terms of speed and acceleration, particularly noting the impact of hard braking events as those were defined by the SHOW consortium.