Chalmers Open Digital Repository

Välkommen till Chalmers öppna digitala arkiv!

Här hittar du:

Studentarbeten utgivna på lärosätet, såväl kandidatarbeten som examensarbeten på grund- och masternivå
Digitala specialsamlingar, som t ex Chalmers modellkammare
Utvalda projektrapporter

Forskningspublikationer, rapporter och avhandlingar hittar du i research.chalmers.se

Enheter i Chalmers ODR

Välj en enhet för att se alla samlingar.

Visar 1 - 2 av 2

Senast publicerade

Cost-Latency Benchmarking for Time-Series Forecasting - Hardware-Aware Evaluation of Deep Learning Architectures for Financial Planning
(2026) Asplund, Gustaf; Bjerhem Aronsson, Felix
Enterprise financial planning is increasingly moving from statistical forecasting toward deep learning, which captures more complex patterns at the product level but raises the cost of serving forecasts. The hardware for these workloads is often chosen through heuristics rather than measurement, and existing serving research has focused on computer vision and language rather than time-series forecasting. This thesis develops a workload-aware benchmarking framework that characterizes the cost-latency trade-offs of deep learning forecasting architectures across commodity cloud hardware. Six forecasting models spanning distinct computational classes were benchmarked on eleven Azure instances, examining how the operational intensity of each architecture sits against the roofline limits of the hardware, how far cost-latency behavior on real enterprise resource planning data diverges from simpler synthetic data, and how a Pareto analysis can guide the choice of a hardware and model pair under a given latency constraint. Operational intensity stayed within a narrow band across the tested models, yet the point at which a workload becomes compute- or memory-bound shifted with the hardware, so the same model could be bound differently from one machine to the next. The cost-latency outcome thus depends on the model and hardware as a pair rather than on the architecture alone, and this shows in the provisioning results. Neither the largest CPU nor the newest accelerator reliably improved serving performance and an older, lower-tier GPU instance offered the best overall balance for the workloads tested. The framework itself, rather than any single figure, is the more transferable result, and the comparison with synthetic data suggests it can stand in as a cheap first pass for narrowing the hardware search before real-data runs settle a final deployment.
Control Design for Differential Lock Synchronization in Heavy-Duty Trucks
(2026) Johansson, Hampus; Karlhager, Lukas
Heavy-duty trucks operating in low-traction environments rely on differential locks to maintain traction when a wheel spins out. These locks are commonly implemented with dog clutches, which require the connected shafts to be speed-matched before they can engage. Following a spin-out, achieving this match can force the driver to slow down or stop, wasting vehicle momentum and creating a safety risk on slopes. This thesis develops and compares active control strategies that synchronize the differential shafts after a wheel spin-out, enabling faster and safer dog clutch engagement. Individual wheel brakes and engine torque are used as actuators. A driveline model is derived for both the open and locked inter-axle differential configurations. A tire force estimator based on a Kalman filter provides feedforward disturbance cancellation, and a state transformation resolves an observability problem that arises when the inter-axle differential is locked. Three model-based controllers are designed and evaluated: a Linear-Quadratic Regulator (LQR), a Model Predictive Controller (MPC), and a Sliding Mode Controller (SMC). They are compared in simulation across split-friction and gravel road scenarios, using performance metrics for synchronization time, velocity loss, driver disturbance, and control effort, with tuning parameters swept to map the trade-offs between objectives. No significant trade-off is found between synchronization time and the remaining metrics: faster synchronization consistently coincides with lower velocity loss and does not worsen driver disturbance or control effort. A control strategy that follows the principles of the SMC is found to be best suited to the problem’s disturbance-heavy nature. Active engine torque control reduces velocity loss when traction allows, while on low-traction surfaces it must instead be limited to avoid excessive brake demand.
Modeling Interest Rate Risk in Swedish Mortgages Using Survival Analysis
(2026) Rödén, Simon
This thesis investigates interest-rate risk in Swedish adjustable-rate mortgages by combining survival analysis with mortgage valuation techniques. The objective is to estimate the behavioral maturity of mortgage loans and model interest rate sensitivity. Using mortgage data from Handelsbanken, mortgage prepayment behavior is modeled through parametric survival analysis, with particular emphasis on shifted gamma frailty models to capture unobserved heterogeneity among borrowers. A mortgage-rate forecasting model is developed to describe the relationship between mortgage rates and risk-free market rates for a banking industry average. The estimated survival model is integrated with a valuation framework to determine effective mortgage maturity, duration, and portfolio sensitivity under various interest-rate environments. Results show that Swedish three-month adjustable-rate mortgages exhibit longer behavioral maturities. This is driven by incomplete passthrough of market interest rates to mortgage rates and persistent borrower behavior. Stress-testing based on Basel-inspired interest-rate shock scenarios demonstrates that mortgage portfolios are particularly vulnerable to rising interest rates. The findings indicate that financing adjustable-rate mortgages exclusively with shortduration liabilities may underestimate the true level of interest-rate risk. This highlights the importance of incorporating behavioral maturity into asset–liability management and provides a practical way for evaluating mortgage-related interest-rate risk in the Swedish banking sector. Keywords:
Execution Monitoring and Local Coordination for Multi-Agent Fleet Management
(2026) Hotvedt, Kim; Manglani, Toshith
Robot fleet-management systems combine high-level scheduling with local motion control to coordinate multiple robots operating in shared environments. However, during execution robots can deviate from a pre-computed schedule, resulting in delays or conflicts. This thesis extends an existing scheduling–control framework by introducing an execution-monitoring local coordinator and implementing the framework on a physical multi-robot platform. The coordinator monitors robot states and predicted trajectories and handles identified conflicts through waiting, priority changes, and re-planning. The proposed method is evaluated in simulation under execution-time disturbances. The distributed Model Predictive Control (MPC) framework and coordinator were also implemented on physical robots (Duckiebots). Hardware experiments showed mean tracking error values between 0.039 m and 0.076 m, while schedule-adherence experiments revealed accumulated delays caused by differences between idealized scheduling assumptions and physical robot behavior. In simulation, the coordinator resolved interactions that otherwise caused large delays or incomplete execution. The results demonstrate that execution monitoring and online coordination can complement high-level scheduling and distributed MPC, while highlighting the trade-off between conservative offline scheduling and dynamic conflict resolution during execution.
Kan beteendeförändring leda till minskad vattenförbrukning
(2026) Sölsnaes Sjöberg, Ida; Ylenfors, Mimmi
Klimatförändringar och stigande vatten- och avloppstaxor ökar behovet av resurseffektiv vattenanvändning. I fastigheter där tekniken redan är optimerad med vattensnåla installationer, innebär det att det krävs något ytterligare för att bli ännu mer resurseffektiva. Denna studie undersöker om nudgingstrategier kan främja beteendeförändring i kommersiella lokaler, samt om dessa kan användas till att uppmuntra hyresgäster att minska sin vattenanvändning. Studien genomfördes som en fallstudie i samarbete med Alecta Fastigheter, i deras fastighet Kvarteret Regina i Göteborg, under mars månad 2026. Som en del av företagets hållbarhetsarbete har beteendeförändring hos hyresgäster identifierats som en central åtgärd för att minska vattenanvändningen (Alecta Fastigheter, 2026). Eftersom fastigheten redan var utrustad med moderna, resurseffektiva installationer riktades fokus mot vad som kan uppnås utöver dem genom att främja beteendeförändring. De nudgingstrategier som implementerades baserades på ett urval av verktyg från Lemoine et al. (2019) och innefattade visuella påminnelseskyltar, veckovis återkoppling om vattenförbrukningen samt involvering av hyresgästerna genom enkät och löpande kommunikation. Data samlades in genom två metoder: veckovisa vattenmätningar som låg till grund för återkopplingen till hyresgästerna, samt en digital enkät besvarad av 16 respondenter. Resultaten visar att vattenförbrukningen minskade kraftigt i början av implementeringen av nudgingstrategierna, för att sedan återgå mot utgångsnivån. Enkätsvaren bekräftar att en majoritet av respondenterna upplevde en ökad medvetenhet och förändrat beteende till följd av påminnelseskyltarna. Sammantaget indikerar studien att nudging kan initiera beteendeförändring även i en komplex kontorsmiljö, men att effekten kräver kontinuitet, variation och en bredare räckvidd för att bli varaktig. Studien ger indikationer på hur fastighetsbolag kan arbeta med beteendepåverkan som ett komplement till tekniska lösningar, ett tillvägagångssätt som kräver mindre omfattande resurser men som i praktiken ställer krav på kontinuitet och uthållighet. Även om resultaten inte är generaliserbara bedöms de ha god överförbarhet och kan utgöra ett underlag för liknande initiativ i jämförbara kontexter.