This paper investigates the influence of tire-road friction on the performance of a long combination vehicle, specifically the A-double, during low-speed manoeuvres such as navigating roundabouts and intersections. Using real-world naturalistic driving data and applying the Performance-Based Standards scheme, the study assesses the low-speed swept path (LSSP) and the driver’s response to varying friction conditions. Results indicate that tire-road friction has minimal effect on the LSSP, which conforms with earlier simulation-based studies. However, it influences driver behaviour, particularly in roundabouts where speed reductions are observed under low-friction conditions. 

Recently, the Swedish government permitted five long combination vehicles (LCVs), with a length longer than 25.25 meters but limited to 34.5 meters, along the selected routes in Sweden. LCVs offer potential benefits such as reduced operational costs, improved fuel efficiency, and lower emissions. While simulations have provided insights into the performance of LCVs on the road, they are conducted in idealised conditions that exclude the complexities of real-world traffic. Therefore, the first major focus of this thesis is to investigate the performance of LCVs in real traffic. Two combinations have been studied: the A-double, a tractor hauling two semitrailers with a dolly in between, and the DuoCAT, a truck hauling two center-axle trailers, which is also called C-double.

The analysis is based on Performance-Based Standards (PBS) measures, including rearward amplification (RWA), low-speed swept path (LSSP), high-speed transient offtracking, and high-speed steady-state offtracking. The steering reversal rate is further applied to assess the driver’s cognitive workload during low-speed manoeuvres. The study covers four traffic scenarios: lane changes, roundabout manoeuvring, turning at intersections, and negotiating tight curves. The results show that both vehicle combinations perform within PBS limits in most cases, with a few outliers across the studied scenarios. The A-double exhibits slightly better stability in lane changes, while the DuoCAT demonstrates slightly better manoeuvrability in roundabouts and intersections. Furthermore, the tire-road friction is examined for the A-double in low-speed manoeuvres, showing that it influences the driver’s behaviour but has negligible effects on the performance of the vehicle.

In the thesis, an alternative approach is investigated that combines low-cost GPS and IMU instead of high-precision sensors to analyse the performance of an LCV. The performance measures, RWA and LSSP, obtained using sensor fusion of GPS/IMU are compared against those obtained from high-precision sensors. The results show a rather good agreement between the two sensor setups in most cases; however, further investigation is needed to improve the LSSP estimates. In addition, a method is presented that employs a computer vision-based approach using data from a forward-facing camera to estimate the relative positions of surrounding vehicles. This method is employed to study lane changes of an A-double, providing insights into its interactions with the surrounding traffic.

The second major focus of the thesis is the simulation-based assessment of collision risk for articulated heavy vehicles. CARLA is used as a simulation environment for this purpose. Since models of articulated vehicles, including LCVs, do not exist in CARLA vehicle libraries, a tractor–semitrailer model is developed as a first step, building upon existing studies. The modelled tractor-semitrailer is used to evaluate the efficacy of an improved two-dimensional time-to-collision (TTC) measure for articulated vehicles, proposed in this thesis. Unlike conventional TTC, which primarily addresses rear-end collisions, the results show that the pro-posed formulation captures both rear-end and sideswipe collision risks, making it more suitable for multi-unit vehicles.

Overall, the thesis enhances the understanding of the behaviour of articulated heavy vehicles, specifically LCVs in real traffic, and proposes frameworks to support future performance and safety studies related to these vehicles.

Nyligen tillät svenska regeringen 5 typer av fordonståg vars längd överstiger 25.25 meter men inte 34.5 meter (LCV) att trafikera utvalda rutter i Sverige. Långa fordonståg erbjuder fördelar såsom lägre driftkostnader, bättre bränsleeffektivitet och minskade utsläpp. Även om simuleringar ger värdefulla insikter om prestanda så bygger de på idealiserade förhållanden och tar inte hänsyn till egenskaper hos verklig trafik. Därför är det första huvudsakliga fokuset i denna avhandling att undersöka prestandan hos LCV i verklig trafik. Två kombinationer har studerats: A-dubbel, en dragbil som drar två påhängsvagnar med en dolly emellan, samt Duo-kärra, en lastbil som drar två släpvagnar med centrerade axlar (kärror), även kallad C-dubbel.

Analysen baseras på Performance-Based Standards (PBS)-mått, inklusive bakåtförstärkning samt svept bana i låg fart (LSSP), och spåravvikelse vid höga hastigheter. Styrreverseringshastighet används dessutom för att bedöma förarens arbetsbelastning vid lågfartsmanövrar. Studien omfattar fyra trafikscenarier: filbyten, manövrering i rondeller, svängar i korsningar samt körning i snäva kurvor. Resultaten visar att båda fordonskombinationerna generellt presterar inom PBS-gränserna men med vissa avvikelser. A-dubbeln uppvisar något bättre stabilitet vid filbyten medan C-dubbel uppvisar något bättre manövrerbarhet i rondeller och korsningar. Däck-väggrepp undersöks vidare för A-dubbeln vid lågfartsmanövrer och resultaten visar att det påverkar förarens beteende men har försumbar effekt på fordonets prestanda.

I avhandlingen undersöks också en analysmetod som kombinerar lågkostnads-GPS och IMU istället för dyra högprecisionssensorer, där de prestandamått som erhålls genom signalbehandling av dessa billigare sensorer jämförs med de som erhålls från högprecisionssensorer. Resultaten visar god överensstämmelse mellan de två sensorsystemen, men ytterligare undersökning krävs för att förbättra LSSP-uppskattningarna. Dessutom presenteras en kamerabaserad metod med en framåtriktad kamera för att uppskatta de relativa positionerna för omgivande fordon. Denna metod tillämpas för att studera filbyten hos ett A-dubbelfordon och analysera dess interaktion med den omgivande trafiken.

Avhandlingens andra huvudsakliga fokus är den simuleringsbaserade bedömningen av kollisionsrisk för tunga fordonskombinationer där simuleringsmiljön CARLA används. Eftersom modeller av tunga fordonskombinationer, inklusive LCV, inte finns i CARLA:s fordonsbibliotek, har en modell av en dragbil med påhängsvagn utvecklats som ett första steg. Den modellerade dragbil-påhängsvagn-kombinationen används för att utvärdera effektiviteten hos ett förbättrat tvådimensionellt tid-till-kollision (TTC)-mått för tunga fordonskombinationer. Till skillnad från konventionell TTC, som främst behandlar påkörningar bakifrån, fångar den föreslagna formuleringen både bakifrån- och sidokollisionsrisker, vilket gör den mer lämplig för långa fordonskombinationer.

Sammanfattningsvis ämnar avhandlingen öka förståelsen för beteendet hos tunga fordonskombinationer, särskilt LCV:er i verklig trafik, och föreslår ramverk för att stödja framtida studier av deras prestanda och säkerhet.

The deployment of High Capacity Transport (HCT) vehicles is in process in different countries. Although their performance has been assessed through simulations and test-track experiments, a question that remains unanswered is: how do these vehicles perform in real traffic? In this paper, the question is addressed for one of the transient manoeuvres, i.e., a lane change using Naturalistic Driving Data (NDD). First, an algorithm is proposed to extract lane changes from the NDD of HCT vehicles using GPS, road data and IMU signals. Following this, the performance of two A-double combinations is assessed in the extracted lane changes using measures commonly used in performance-based standards (PBS) schemes like offtracking and rearward amplification. The dependency of these measures on the factors such as the vehicle’s speed, load and lateral displacement is investigated. The assessment concludes that the vehicles satisfy the PBS requirements proposed for them and are driven safely in the extracted lane changes.

The focus of this paper is to use Naturalistic Driving Data to understand how the drivers manoeuvre an A-double combination in the roundabouts and evaluate performance in the roundabouts using measures like Low-Speed Swept Path (LSSP) and Tail Swing (TS). The analyses of the steering patterns and speed variations depict that the standard deviations of the responses of the drivers for a given travel direction in a roundabout are within 35o (17 % of the baseline) for the steering wheel angle and 8 km/h (40 % of the baseline) for the speed. It is also found that the cognitive workload of the drivers due to the steering pattern is higher in right turns compared to straight crossings through the roundabout. The performance analyses show a dependency of LSSP on the instantaneous radius obtained from the vehicle's path, and the vehicle's travel direction in the roundabout. LSSP ranges from 7.7 m for a left turn in a roundabout with an inner radius of 12 m to 3.1 m for a straight crossing in a roundabout with a 30 m inner radius. TS is observed in only one roundabout and its magnitude goes up to 0.4 m in a roundabout of 30 m inner radius.

The deployment of long combination vehicles (LCVs) is currently in progress in Sweden. LCV refers to heavy vehicles that are longer than 25.25 m, which is the conventional length limit in Swedish regulations. LCVs reduce operational costs, improve fuel efficiency and reduce CO2 emission per ton-km. Despite their numerous advantages, a question that still revolves around these vehicles is how they perform on the road. Although this question has been answered using simulations, an analysis using real traffic data is still missing. 

This thesis assesses the performance of LCVs using naturalistic driving data (NDD). The performance assessment is done using Performance-based standards (PBS) measures. PBS is a regulatory scheme for heavy vehicles, such as LCVs, that includes performance measures with a quantified required level of performance. The main PBS measures used in this thesis are rearward amplification, low-speed swept path, high-speed transient offtracking, and high-speed steady-state offtracking. Rearward amplification represents the amplification of motions in the rear end of a vehicle combination, which relates to its stability, and the remaining three are indicative of the space that the vehicles occupy in different scenarios. The steering reversal rate is also employed to compute the cognitive workload of the drivers in low-speed scenarios. 

Two LCV combinations are considered for analysis, namely an A-double composed of a tractor-semitrailer-dolly-semitrailer/tractor-semitrailer-full trailer and a DuoCAT composed of a truck hauling two centre-axle trailers. Four scenarios are of interest to this thesis: lane changes, manoeuvring through roundabouts, turning in intersections and negotiating tight curves. The thesis presents three contributions outlining the analysis methodologies, followed by a discussion of the results obtained from the analysis. 

The first contribution involves developing an algorithm to extract lane changes from the NDD of LCVs. The algorithm is used against the data obtained from A-doubles. The results indicate that A-doubles adhere to proposed safety limits during lane changes. 

The second contribution assesses the performance of A-double in round-abouts using NDD. Various roundabouts with different radii are considered. The vehicle occupies more space with lower radius roundabouts than higher radius roundabouts. The space occupied in all the cases is below the proposed safety limit. However, steerable axles might be needed for smaller roundabouts than those considered in this study. Additionally, the driver’s cognitive work-load is found to vary with the roundabout’s radius, with drivers navigating higher-radius roundabouts more easily. 

The third contribution deals with the performance evaluation of DuoCAT across four scenarios, followed by the comparison with A-double. The results indicate that the A-double and the DuoCAT are stable and have good tracking performance in most cases. The A-double is observed to be slightly more stable in lane changes, whereas the DuoCAT has slightly better manoeuvrability at low-speed scenarios such as roundabouts and intersections. 

Introduceringen av långa fordonskombinationer (LCV) pågår för närvarande i Sverige vilket ger en möjlighet att minska driftskostnader samtidigt som bränsleeffektiviteten förbättras och koldioxidutsläppen per ton-km reduceras. LCV avser tunga fordon som är längre än 25,25 meter, vilket är den konventionella längdgränsen enligt svenska regler. Trots fördelarna är en fråga hur dessa fordon presterar på väg. Den här avhandlingen studerar och analyserar den frågan med hjälp av naturalistisk kördata (NDD) från experiment med LCV.

Prestandabedömningen görs med hjälp av prestandabaserade standarder (PBS). PBS är ett regleringssystem för tunga fordon, såsom LCV, som kvantifierar och kravställer fordonets beteende. De huvudsakliga PBS-måtten som används i denna avhandling är bakåtförstärkning, spåravvikelse och svept area i låg hastighet. Bakåtförstärkning representerar förstärkningen av rörelser från den främre till den bakre delen av en fordonskombination, vilket relaterar till dess stabilitet, och de två återstående indikerar det utrymme som fordonen upp-tar i olika scenarier. Dessutom används styråtervändningshastigheten (SRR) för att beräkna förarnas kognitiva arbetsbelastning vid låga hastigheter, exempelvis vid körning i rondeller och korsningar.

Två LCV varianter studeras i denna avhandling, en A-dubbel bestående av dragbil-påhängsvagn-dolly-påhängsvagn/dragbil-påhängsvagn-släpvagn samt en Duo-kärra bestående av en lastbil som drar två släpvagnar med centrerade ax-lar. Fyra scenarier är av intresse för denna avhandling: filbyten, manövrering genom rondeller, svängning i korsningar och körning i snäva kurvor. Avhandlingen presenterar tre bidrag som beskriver analysmetoderna och efterföljande resultatdiskussioner.

I det första bidraget utvecklas en algoritm för att extrahera filbyten från naturalistisk kördata av LCV fordon där metoden används på data från A-dubbelfordon. Resultaten indikerar att A-dubbelfordon håller sig till föreslagna säkerhetsgränser under filbyten.

I det andra bidraget bedöms prestandan hos ett A-dubbelfordon i rondeller med hjälp av NDD. Olika rondeller med olika radier studeras. Fordonet upptar mer utrymme i rondeller med mindre radie jämfört med rondeller med större radie och det upptagna utrymmet ligger i alla undersökta fall under den föreslagna säkerhetsgränsen. För mindre rondeller än de som tas upp i denna studie kan styrbara axlar behövas. Dessutom visas att förarens kognitiva belastning varierar med rondellens radie där förare som navigerar större rondeller har en lägre kognitiv belastning.

Det tredje bidraget handlar om prestandautvärdering av Duo-kärrafordon över fyra scenarier, följt av jämförelsen med A-dubbelfordon. Resultaten indikerar att både A-dubbelfordon och Duo-kärrafordon är stabila och har god spårningsprestanda i de flesta fallen. A-dubbelfordon observeras vara något mer stabil i filbyten, medan Duo-kärrafordon har något bättre manövrerbarhet vid låghastighetsscenarier som rondeller och korsningar.  

Time-to-collision (TTC) is a widely used measure for predicting rear-end collisions, assuming constant speed and heading for both vehicles in the prediction horizon. However, this conventional formulation cannot detect sideswipe collisions. A two-dimensional extension, TTC_2D, has been proposed in the literature to address lateral interactions. However, this formulation assumes both vehicles have the same heading and that their headings remain unchanged during the manoeuvre, in addition to the constant speed and heading assumptions in the prediction horizon. Moreover, its use for articulated vehicles like a tractor-semitrailer remains unclear. This paper proposes three enhanced versions of TTC_2D to overcome these limitations. The first incorporates the vehicle heading to account for directional differences. The standard assumption of constant speed and heading in the prediction horizon holds. The second adapts the formulation for articulated vehicles, and the third allows for constant acceleration, relaxing the constant speed assumption in the prediction horizon. All versions are evaluated in simulated cut-in scenarios, covering both sideswipe and rear-end collisions, using the CARLA simulation environment with a tractor-semitrailer model. Results show that the proposed versions predict sideswipe collisions with better accuracy compared to existing TTC2_D. They also detect rear-end collisions similar to the existing methods.