arXiv subjects

Armin Abdolmohammadi

Publications and source records attributed to Armin Abdolmohammadi.

5 recordsLinked to original sources

arXiv2026

Wheel-loader V-Cycle Automation with Deep Koopman MPC

The repeated forward-reverse maneuvers performed by wheel loaders during earthmoving operations make them well suited for automation. However, the nonlinear dynamics of articulated vehicles and complex vehicle-terrain interactions limit the effectiveness of conventional model-based approaches. This paper presents a hierarchical framework that combines long-horizon geometric planning with data-driven predictive control for autonomous wheel-loader operation. A reduced-order articulated kinematic model is used to generate the maneuver geometry, where the forward and reverse trajectories are jointly optimized through a shared intermediate state. To capture the vehicle dynamics, two data-driven deep bilinear Koopman models are learned for the forward and reverse motions using data generated from high-fidelity simulations in Algoryx Dynamics. The learned Koopman representations are subsequently incorporated into a computationally efficient model predictive control (MPC) formulation for trajectory tracking. The resulting controller operates in real time within a 50-ms execution loop. High-fidelity simulation results demonstrate that the proposed end-to-end framework enables accurate and computationally efficient execution of wheel-loader V-cycle maneuvers, providing a promising approach toward autonomous operation of articulated heavy-duty machinery.

cs.RO↗

arXiv2025

Fleet Size and Mix Capacitated Vehicle Routing Problem with Time Windows for Mobile Fast Chargers

The electrification of off-road heavy equipment presents operational challenges for agencies serving remote sites with limited fixed charging infrastructure. Existing mobile fast charging vehicle (MFCV) planning approaches typically treat fleet design and routing as separate problems, fixing vehicle characteristics before dispatch. This paper formulates a fleet size and mix capacitated vehicle routing problem with time windows (FSMCVRPTW) for MFCV deployment, jointly optimizing fleet composition, charger specifications, routing, and scheduling within a unified mixed-integer linear program. The model incorporates heterogeneous MFCV types with varying power ratings, battery capacities, fuel range, and cost structures, minimizing total daily cost from labor, fuel, amortized capital expenditure, and energy purchase under temporal service windows, resource budgets, and energy-delivery constraints. The formulation is implemented in Python/Gurobi and applied to two case studies using California Department of Transportation wheel-loader data in Los Angeles (dense urban) and Truckee (sparse mountainous). Results show that simultaneous optimization yields compact, well-utilized fleets that meet all service windows while revealing strong sensitivity of unit cost to demand density and geography. The proposed FSMCVRPTW framework provides a generalizable decision-support methodology that co-designs fleet size, charger power, routing, and service schedules in a single optimization layer for context-aware, cost-efficient mobile fast charging.

eess.SY↗

arXiv2025

Data-Efficient Excavation Force Estimation for Wheel Loaders

Accurate prediction of excavation forces is critical for enabling autonomous operation and optimizing control strategies in earthmoving machinery. Conventional approaches often depend on extensive data collection or computationally expensive simulations across multiple soil types, which limits their scalability and adaptability. This study presents a data-efficient framework that calibrates soil parameters using force measurements from the preceding bucket-loading cycle. The proposed method is based on an analytical soil-tool interaction model formulated through the fundamental earthmoving equation, and employs a multi-stage optimization procedure during the loading phase to identify relevant soil parameters. These estimated parameters are then used to predict excavation forces in the subsequent cycle, allowing the system to adapt its control inputs without relying on large-scale datasets or machine learning model training. The framework is validated through high-fidelity simulations in the Algoryx Dynamics engine under different soil types and excavation trajectories, achieving root-mean-square prediction errors between 10% and 15%. This cycle-to-cycle adaptation demonstrates strong potential for scalable, online force estimation and efficient path planning in wheel loader operations.

eess.SY↗

arXiv2025

Optimal Path Planning for Wheel Loader Automation Enabled by Efficient Soil-Tool Interaction Modeling

Earthmoving operations with wheel loaders require substantial power and incur high operational costs. This work presents an efficient automation framework based on the Fundamental Earthmoving Equation (FEE) for soil-tool interaction modeling. A reduced-order multi-step parameter estimation method guided by Sobol's global sensitivity analysis is deployed for accurate, online excavation force prediction. An optimal control problem is then formulated to compute energy-efficient bucket trajectories using soil parameters identified in the previous digging cycle. High-fidelity simulations in Algoryx Dynamics confirm accurate force prediction and demonstrate 15-40% energy savings compared to standard paths. The total computation time is comparable to a single digging cycle, highlighting the framework's potential for real-time, energy-optimized wheel loader automation.

eess.SY↗

arXiv2025

Multi-Step Deep Koopman Network (MDK-Net) for Vehicle Control in Frenet Frame

The highly nonlinear dynamics of vehicles present a major challenge for the practical implementation of optimal and Model Predictive Control (MPC) approaches in path planning and following. Koopman operator theory offers a global linear representation of nonlinear dynamical systems, making it a promising framework for optimization-based vehicle control. This paper introduces a novel deep learning-based Koopman modeling approach that employs deep neural networks to capture the full vehicle dynamics-from pedal and steering inputs to chassis states-within a curvilinear Frenet frame. The superior accuracy of the Koopman model compared to identified linear models is shown for a double lane change maneuver. Furthermore, it is shown that an MPC controller deploying the Koopman model provides significantly improved performance while maintaining computational efficiency comparable to a linear MPC.

eess.SY↗