💻 computer science

Adaptive Sliding Mode Control for Vehicle Platoons with State-Dependent Friction Uncertainty

This thesis proposes a novel two-stage adaptive sliding mode controller for vehicle platoons that effectively manages state-dependent friction uncertainties and external disturbances without requiring prior knowledge of friction parameters, thereby ensuring robust speed regulation and safe inter-robot distance maintenance.

Original authors: Rishabh Dev Yadav

Published 2026-03-17

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Rishabh Dev Yadav

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 a group of friends trying to walk in a perfect line down a busy, unpredictable street. They want to stay exactly one meter apart, move at the same speed, and follow a specific winding path.

Now, imagine that the ground beneath them changes constantly. Sometimes it's smooth pavement, sometimes it's sticky mud, and sometimes it's a steep hill. Furthermore, their shoes wear out differently depending on how fast they run. This is the challenge of Vehicle Platooning: getting a group of autonomous cars (or robots) to drive in a tight, safe formation without crashing or falling apart.

This thesis by Rishabh Dev Yadav tackles a specific, tricky problem: How do you control these robots when you don't know exactly how "sticky" or "slippery" the ground is, and that stickiness changes as they move?

Here is a simple breakdown of the paper's solution:

1. The Problem: The "Mystery Mud"

Most self-driving car systems are like students who memorized a textbook perfectly. They know exactly how much force is needed to move a car on dry asphalt. But in the real world, the road is messy.

The Unknown: The friction between a tire and the road isn't a fixed number. It changes based on speed, tire wear, rain, or a patch of gravel.
The Flaw in Old Systems: Traditional controllers try to guess the friction or assume it stays within a safe, known limit. But if the road suddenly becomes super slippery (like ice) or super sticky (like mud), these old systems get confused. They either brake too hard (causing a crash) or don't brake enough (causing a collision).

2. The Solution: The "Adaptive Sliding Mode" Coach

The author proposes a new type of controller called Adaptive Sliding Mode Control (ASMC). Think of this not as a rigid rulebook, but as a highly intuitive coach standing next to the robots.

The "Sliding" Part (The Rail): Imagine the robots are on a train track. The "sliding mode" is the track itself. The controller's job is to keep the robot on the track, even if the wind blows or the engine sputters. If the robot starts to drift off the track, the controller pushes it back immediately.
The "Adaptive" Part (The Learner): This is the magic sauce. The coach doesn't need to know the road conditions beforehand.
- If the robot starts to slip because of "mystery mud," the coach feels the slip.
- Instead of panicking, the coach instantly says, "Okay, the ground is slippery now. I need to push harder to keep us on the track."
- It learns the friction level in real-time and adjusts its strength on the fly. It doesn't need a manual; it figures it out as it goes.

3. The Two-Step Dance

The paper splits the job into two distinct roles, like a dance partner and a muscle builder:

The Choreographer (Kinematic Controller): This part looks at the map and says, "We need to turn left here and move at 5 mph." It plans the path but doesn't worry about how hard the motors need to work.
The Muscle (Dynamic Controller): This is the new adaptive part. It takes the Choreographer's plan and figures out the actual force needed. It asks, "The Choreographer says turn left, but the road is icy! I need to apply 20% more torque to the left wheel to make that turn happen."

4. The Proof: The Video Game Test

To prove this works, the author didn't just do math on paper. They built a virtual world (using a simulator called Gazebo, which is like a high-tech video game engine for robots).

The Setup: Three robots were sent to drive in a figure-eight pattern.
The Trap: The virtual arena had different zones. Some parts were smooth, but one specific corner had "super sticky" friction, and there were even speed bumps to jolt them.
The Result:
- The Old Controller (the student who memorized the textbook) got confused in the sticky zone. It drifted off course, and the gap between the robots got messy.
- The New Adaptive Controller (the intuitive coach) felt the change, adjusted its grip instantly, and kept the robots perfectly aligned. They stayed in formation, even when the road tried to throw them off.

The Big Takeaway

This research gives self-driving platoons a "superpower": Resilience.

Instead of needing a perfect map of every pothole and patch of ice before they start driving, these robots can now handle the unknown. They can drive through construction zones, rainstorms, or changing terrain and still stay in a tight, safe line. It's the difference between a robot that breaks when the road gets weird and a robot that says, "No problem, I've got this," and keeps moving forward.

1. Problem Statement

The research addresses the challenge of controlling autonomous vehicle platoons (groups of wheeled mobile robots moving in formation) in the presence of unknown, state-dependent friction forces.

The Core Issue: Maintaining a safe inter-robot distance and tracking a desired trajectory requires robust control. However, friction between wheels and the ground is not a fixed parameter; it varies dynamically based on road surface conditions, tire wear, vehicle speed, and payload.
Limitation of Existing Methods: State-of-the-art Adaptive Sliding Mode Controllers (ASMC) typically assume that system uncertainties (like friction) are a priori bounded (i.e., their maximum value is known or can be estimated beforehand).
The Gap: Because friction is state-dependent (specifically, viscous friction is proportional to velocity), it cannot be strictly bounded a priori. Traditional controllers fail to accurately model or identify these forces, leading to performance degradation or instability when road conditions change (e.g., moving from dry asphalt to a wet or rough surface).

2. Methodology

The proposed solution is a two-stage Adaptive Sliding Mode Control (ASMC) framework designed for nonholonomic wheeled mobile robots (WMRs). The approach separates the control problem into kinematic and dynamic layers.

A. System Modeling

Kinematics: Models the robot's position and orientation ( $x, y, \theta$ ) relative to a reference trajectory.
Dynamics: Uses a Lagrangian approach to model forces, including mass ( $m$ ), inertia ( $J$ ), external disturbances, and friction forces ( $f_v, f_w$ ).
Friction Model: The thesis utilizes a combined Coulomb + Viscous friction model. Crucially, it acknowledges that the viscous component is proportional to velocity ( $v$ ), making the total friction force state-dependent and unbounded by a constant prior to execution.

B. Control Architecture

The controller operates in two distinct stages:

Kinematic Controller (Backstepping):
- Calculates desired linear ( $v_c$ ) and angular ( $\omega_c$ ) velocities based on the tracking error between the robot's current pose and the desired trajectory.
- Uses a standard backstepping approach with positive design constants ( $k_1, k_2, k_3$ ) to ensure the posture error converges.
Dynamic Controller (Proposed ASMC):
- Objective: Generate the actual force ( $F$ ) and torque ( $\tau$ ) commands to make the robot's actual velocities track the desired velocities ( $v_c, \omega_c$ ) despite unknown dynamics.
- Sliding Variables: Defined for linear velocity ( $s_v$ ) and angular velocity ( $s_\omega$ ) errors, incorporating integral terms to eliminate steady-state error.
- Uncertainty Handling: The controller assumes the system parameters (mass, inertia, friction coefficients) and disturbance bounds are unknown.
- Adaptive Laws: The core innovation lies in the adaptive gain laws. Instead of using fixed bounds, the controller estimates the upper bounds of the uncertainties ( $\hat{K}$ $\hat{K}$ ) in real-time.
  - The uncertainty bounds are structured as linear functions of the state errors ( $\xi_v, \xi_\omega$ ).
  - Adaptive laws update the estimates ( $\dot{\hat{K}}$ ) based on the magnitude of the sliding surface and the error states, allowing the controller to "learn" the magnitude of the friction forces as they change.
- Control Law: The control input is a combination of a stabilizing term ( $-\Lambda s$ ) and a robust switching term ( $-\rho \text{sgn}(s)$ ), where the gain $\rho$ is dynamically adjusted by the adaptive estimates.

C. Stability Analysis

Lyapunov Theory: The stability of the closed-loop system is rigorously proven using a Lyapunov function candidate.
Result: The analysis demonstrates that the system trajectories are Uniformly Ultimately Bounded (UUB). This means that despite the unknown and state-dependent nature of the friction, the tracking errors and parameter estimation errors will converge to a small, bounded neighborhood of zero.

3. Key Contributions

State-Dependent Uncertainty Handling: The primary contribution is an ASMC framework capable of handling unbounded, state-dependent uncertainties (specifically viscous friction) without requiring prior knowledge of system parameters or friction bounds.
Decentralized Platoon Control: The method is applied to a multi-robot platoon where each follower robot only relies on the state of its immediate predecessor, ensuring scalability.
Theoretical Proof: Provides a formal Lyapunov-based proof of UUB stability for a system with coupled, state-dependent dynamic uncertainties.
Realistic Simulation Validation: Extensive testing in the Gazebo simulation environment using TurtleBot3 models with a custom plugin to simulate varying friction coefficients and physical obstacles (speed breakers).

4. Simulation Results and Analysis

The proposed controller was compared against a standard ASMC (which assumes a priori bounded uncertainties) in a custom arena with varying friction zones and obstacles.

Scenario: Three robots followed a figure-of-eight path. The arena included a high-friction quadrant ( $\mu=0.13$ ) and speed breakers to induce sudden disturbances.
Performance Metrics: Root Mean Squared (RMS) error for trajectory tracking and gap maintenance.
Key Findings:
- High Friction Zones: In the high-friction quadrant (where state-dependent friction changed significantly), the standard ASMC exhibited large spikes in tracking error and gap deviation. The proposed ASMC maintained significantly lower errors.
- Disturbance Rejection: When encountering speed breakers, the proposed controller showed fewer and smaller error spikes compared to the standard controller.
- Quantitative Improvement:
  - Trajectory Tracking: The proposed ASMC reduced the RMS position error by approximately 6–12% across all robots compared to the standard ASMC.
  - Gap Maintenance: The RMS gap error was reduced by nearly 50% (e.g., from 0.050m to 0.036m between Robot 2 and 3), demonstrating superior formation keeping.

5. Significance and Conclusion

This work bridges a critical gap in robotic control theory by addressing the limitation of "a priori bounded" assumptions in adaptive control.

Practical Impact: It enables autonomous vehicle platoons to operate safely and efficiently in unstructured, real-world environments where road conditions are unpredictable and variable.
Robustness: The controller does not require complex friction modeling or sensor-based friction estimation; it adapts purely through control laws.
Future Directions: The author suggests future work involving real-world hardware implementation, testing under varying slopes and payloads, and extending the methodology to other autonomous systems like aerial drones and underwater vehicles.

In summary, the thesis presents a robust, theoretically sound, and practically validated control strategy that significantly enhances the reliability of autonomous vehicle platoons in the face of complex, unknown friction dynamics.