💻 computer science

Trajectory Planning for an Articulated Commercial Vehicle using Model Predictive Contouring Control

This paper presents an extended Model Predictive Contouring Control (MPCC) framework tailored for articulated commercial vehicles that incorporates scenario-dependent anchor point prioritization and explicit multi-axle road-boundary constraints to ensure safe and stable trajectory planning in both forward and reverse maneuvers.

Original authors: A. J. Aertssen, R. G. M. Huisman, I. J. M. Besselink, J. Elfring, M. J. G. van de Molengraft

Published 2026-04-28

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: A. J. Aertssen, R. G. M. Huisman, I. J. M. Besselink, J. Elfring, M. J. G. van de Molengraft

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to park a massive, 16-meter-long truck with a trailer attached. Now, imagine doing this not just in a straight line, but by backing it into a tight loading dock, or navigating a sharp city corner. If you were driving a regular car, you'd just turn the wheel and follow the center of the lane. But with a truck and trailer, it's like trying to walk a tightrope while carrying a long pole on your shoulders: the back end doesn't follow the front end; it swings wide or cuts the corner in a dangerous way. This is called "off-tracking," and if you aren't careful, the trailer can fold up against the cab like an accordion (a "jackknife"), or the wheels can run off the road entirely.

This paper presents a new "brain" for autonomous trucks that solves these problems using a method called Model Predictive Contouring Control (MPCC). Here is how it works, broken down into simple concepts:

1. The Problem: One Size Does Not Fit All

For regular cars, existing computer programs can easily plan a path. But a truck with a trailer is a different beast.

The Size Issue: A truck is huge. When it turns, the front sticks out into the opposite lane, and the trailer's wheels might cut across the sidewalk.
The Reversing Issue: Backing up a truck is unstable. It's like trying to balance a broomstick on your hand while walking backward; one wrong move and it tips over.
The "Who's Boss?" Issue: Sometimes, the most important part of the vehicle is the trailer (like when backing into a dock). Other times, it's the front of the truck (like when pulling up to a charging station). The computer needs to know which part of the vehicle to prioritize at any given moment.

2. The Solution: A "Virtual String" and a "Magic Rope"

The researchers upgraded an existing algorithm (MPCC) to handle these specific truck problems. Think of it this way:

The Virtual String (The Reference Path): Imagine a string laid out on the ground showing the perfect path the truck should follow.
The Magic Rope (The Anchor Points): In the past, computers only watched one point on the vehicle (like the center of the rear axle). This new system attaches "magic ropes" to three specific spots: the front of the truck, the back of the truck, and the back of the trailer.
Scenario-Dependent Priorities: The system can tighten or loosen the focus on these ropes.
- Docking: It tightens the rope on the trailer's back, ensuring the trailer hits the dock perfectly, even if the truck has to swing wide.
- Highway Driving: It keeps all ropes taut, ensuring the whole vehicle stays centered in the lane.

3. The Safety Net: Keeping All Wheels on the Road

The biggest innovation here is that the computer doesn't just check if the center of the truck is safe. It checks the front wheels, the rear truck wheels, and the trailer wheels individually.

Imagine driving a long boat through a narrow canal. If you only look at the middle of the boat, you might think you're safe, but the bow (front) or stern (back) could scrape the walls. This system draws a "corridor" for the road and ensures that every single wheel stays inside that corridor, preventing the truck from clipping curbs or hitting obstacles.

4. The Results: Testing the Theory

The researchers tested this "brain" in two ways:

Computer Simulations: They created a virtual world where the truck had to drive into a customer's yard, turn a sharp corner, and back into a dock. The system successfully navigated both forward and backward without the trailer jackknifing or the wheels leaving the road.
Real-World Prototype: They put the software on a real, full-size prototype truck. The truck successfully followed the planned path in the real world, proving the math works on actual metal and rubber.

The Takeaway

This paper doesn't just say "we can drive a truck." It says, "We can teach a computer to drive a truck by watching every part of the vehicle, not just the center, and by knowing which part of the vehicle matters most at any specific moment."

By adjusting the "weights" (how much the computer cares about the trailer vs. the truck), the system can act like a cautious driver backing into a tight spot or a smooth driver cruising down the highway, all while keeping every wheel safely within the lines.

1. Problem Statement

The paper addresses the challenge of autonomous trajectory planning for articulated commercial vehicles (specifically tractor-semitrailers) in complex logistics scenarios. Unlike passenger cars, these vehicles face unique difficulties:

Size and Geometry: They are significantly larger (approx. 16.5m length) and exhibit offtracking, where the semitrailer follows a different path than the tractor, often causing the inner wheels to cut corners or the outer wheels to encroach on adjacent lanes.
Instability: Reversing maneuvers are prone to jackknifing (unstable folding of the articulation angle).
Scenario-Dependent Priorities: Different driving tasks require prioritizing different "anchor points" on the vehicle. For example, during docking, the semitrailer's rear position is critical; during charging, the tractor's port is critical; on highways, both must stay centered.
Limitations of Existing Methods: Standard Model Predictive Contouring Control (MPCC) used for cars often simplifies the vehicle to a single point or assumes a fixed reference path, failing to account for the complex kinematics and boundary constraints of articulated vehicles.

2. Methodology

The authors propose an extended Model Predictive Contouring Control (MPCC) framework tailored for tractor-semitrailers.

A. Vehicle Modeling

Kinematic Model: A kinematic bicycle model is used for the tractor, coupled with the semitrailer via an articulation point.
State Vector: The state includes the tractor's position $(x_1, y_1)$ , heading angles ( $\psi_1, \psi_2$ ), velocity ( $v_1$ ), and steering angle ( $\delta_0$ ).
Augmented States: Crucially, the model introduces three independent progress states ( $\theta_0, \theta_1, \theta_2$ $θ_{0}, θ_{1}, θ_{2}$ ) corresponding to three anchor points:
1. Tractor front axle ( $i=0$ )
2. Tractor rear axle ( $i=1$ )
3. Semitrailer axle ( $i=2$ )
  Each anchor point has its own virtual velocity ( $v_{\theta}$ ) to approximate the distance traveled along the reference path.

B. Optimization Formulation

The problem is formulated as a constrained nonlinear optimization problem minimizing a cost function $J$ subject to dynamics and constraints:

Objective Function ( $J_m$ ): Minimizes contour error (lateral deviation from the path) and lag error (longitudinal deviation) for each anchor point, while maximizing path speed.
- Key Innovation: Scenario-dependent weighting allows the planner to prioritize specific anchor points (e.g., increasing the weight $q_{c,2}$ for the semitrailer during docking).
Comfort Term ( $J_c$ ): Penalizes high acceleration and steering rates to ensure passenger/payload comfort.
Constraints:
- Kinematic/Dynamic Limits: Bounds on steering angle, steering rate, velocity, and acceleration.
- Road Boundaries: Explicit constraints are applied to all three anchor points to ensure that the front tractor axle, rear tractor axle, and semitrailer axle all remain within the drivable corridor (defined by left/right boundaries). This prevents the "swept path" from colliding with obstacles.
- Safety: Constraints on lateral acceleration to prevent rollover.

C. Implementation

The solver uses a receding horizon scheme implemented in Python with CasADi and the IPOPT solver.
The prediction horizon is set to 15 seconds, balancing computational load (approx. 100ms per iteration) with sufficient look-ahead distance for both forward and reverse maneuvers.

3. Key Contributions

Extended MPCC for Articulated Vehicles: The first application of MPCC to tractor-semitrailers, overcoming the limitations of previous methods that treated the vehicle as a single point.
Multi-Anchor Point Constraint Formulation: The explicit formulation of road boundary constraints for the tractor front, tractor rear, and semitrailer axles simultaneously. This ensures the entire vehicle swept path remains within drivable limits, preventing jackknifing and off-tracking collisions.
Scenario-Dependent Prioritization: The ability to dynamically adjust optimization weights to prioritize specific vehicle parts (e.g., semitrailer for docking vs. tractor for charging) without changing the underlying algorithm structure.
Validation: Successful simulation and real-world testing on a full-scale prototype.

4. Results

Simulation Scenarios: The algorithm was tested in a logistics scenario involving a 90-degree turn and entry into a customer premises (forward and reverse).
- Success: The vehicle successfully navigated the corridor in both directions without jackknifing or exiting the drivable area.
- Necessity of Constraints: When the semitrailer constraints were disabled, the simulation resulted in the semitrailer cutting corners (forward) or jackknifing (reverse), proving the necessity of multi-point constraints.
- Weight Sensitivity: Varying the weight $q_{c,2}$ (semi-trailer priority) demonstrated that higher weights reduced the semitrailer's lateral deviation but required the tractor to steer more aggressively and slow down to satisfy lateral acceleration limits.
Real-World Prototype: The method was implemented on a full-scale DAF tractor-semitrailer prototype. The vehicle successfully followed the planned trajectory in a real environment, confirming practical applicability.

5. Significance

This work bridges a critical gap between autonomous passenger car planning and the complex needs of heavy logistics. By extending MPCC to handle articulated kinematics and multi-point boundary constraints, the authors provide a robust solution for:

Safety: Preventing jackknifing and ensuring all wheels stay within lane boundaries.
Efficiency: Enabling autonomous docking and reversing in tight spaces, which is currently a major bottleneck in logistics automation.
Flexibility: Offering a unified control framework that adapts to different operational modes (docking, charging, highway driving) through simple parameter tuning rather than algorithmic rewrites.

The paper concludes that while constructing robust drivable corridors in dynamic environments remains a challenge, the proposed MPCC extension is a significant step toward fully autonomous articulated commercial vehicles.