⚡ electrical engineering

Post-Collision Trajectory Restoration for a Single-track Ackermann Vehicle using Heuristic Steering and Tractive Force Functions

This paper proposes a structured heuristic control law for a single-track Ackermann vehicle that jointly manages steering and tractive force to restore the intended trajectory following a collision, specifically accounting for nonlinear coupling and time-varying longitudinal velocity.

Original authors: Samsaptak Ghosh, M. Felix Orlando, Sohom Chakrabarty

Published 2026-02-10

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Samsaptak Ghosh, M. Felix Orlando, Sohom Chakrabarty

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 driving a self-driving car on a straight highway. Suddenly, a stray dog or a piece of debris clips the side of your car. You aren't in a massive wreck, but the impact sends your car skidding sideways and spinning slightly off the road.

In a normal car, a human would instinctively steer back and hit the gas or the brakes to steady themselves. But for a computer, this is a mathematical nightmare. Most self-driving software is programmed to follow a perfect line; they aren't great at "recovering" once they’ve been knocked off that line.

This research paper proposes a new "emergency recovery plan" for autonomous vehicles to help them get back on track after a bump.

The Problem: The "Spinning Top" Effect

Think of a self-driving car like a spinning top. When it’s moving straight, it’s stable. But a collision is like flicking the top with your finger. It doesn't just move to the side; it starts to wobble, tilt, and slide in ways that are hard to predict.

Most current AI models assume the car is always moving at a steady speed. But after a crash, the car might slow down or speed up unexpectedly. If the computer doesn't account for that change in speed, its attempt to steer back might actually make the skid worse—like trying to catch a falling tray by moving your hands in the wrong direction.

The Solution: The "Rhythmic Dance" (Heuristic Control)

Instead of using a super-complex, heavy math formula that takes too long to calculate, the researchers created something they call a "Heuristic Function."

Think of this like a professional dancer’s muscle memory. When a dancer loses their balance, they don't stop to solve physics equations; they perform a specific, rhythmic movement—a quick sway of the hips and a shift in weight—to find their center again.

The researchers designed two specific "rhythmic movements" for the car:

The Steering Sway (The Steering Function): Instead of just turning the wheel hard one way (which might cause a flip), the car performs a controlled, wave-like steering motion. It’s like a gentle "S" curve that nudges the nose of the car back toward the center of the lane.
The Power Pulse (The Tractive Force Function): At the same time, the car doesn't just slam on the brakes. It gives the engine a specific "pulse" of power. Imagine you are running on ice and slip; you wouldn't just stop dead, you’d push off the ground with a specific rhythm to regain your footing. This "pulse" helps stabilize the car's forward momentum so the steering works better.

How do we know it works?

The researchers tested this using two different digital "car models"—one simple and one very complex (which includes things like wind resistance and messy tire friction).

They simulated two scenarios:

Scenario 1 (The Side Bump): A hit that pushes the car sideways but keeps it facing forward.
Scenario 2 (The Corner Hit): A hit that pushes the car sideways and makes it spin.

In both cases, the "rhythmic" steering and power pulses worked. While a car without this system would have drifted off into the ditch, the car with this new "muscle memory" successfully wobbled, pulsed, and slid its way back onto the straight path.

The Bottom Line

This paper is essentially teaching self-driving cars how to "recover their balance." It moves AI away from being a perfectionist that panics when things go wrong, and toward being a resilient driver that knows how to dance its way back to safety after a bump in the road.

Technical Summary: Post-Collision Trajectory Restoration for a Single-Track Ackermann Vehicle

1. Problem Statement

In the field of autonomous driving, current research focuses heavily on path-tracking and obstacle avoidance under the assumption of a collision-free environment. However, there is a critical research gap regarding post-collision recovery. When an autonomous vehicle undergoes an impact, the resulting lateral motion and yaw transients can rapidly deviate the vehicle from its intended path, potentially leading to secondary accidents.

Existing literature often simplifies vehicle dynamics by assuming constant longitudinal velocity ( $v_x$ ). This paper identifies that such simplifications are inadequate for the post-collision regime, where sudden speed variations and nonlinear steering-coupled interactions significantly influence the vehicle's ability to stabilize and return to its trajectory.

2. Methodology

The authors propose a structured, heuristic, open-loop control law that jointly commands steering and tractive force to restore the vehicle's path.

Vehicle Modeling: The study utilizes two models:
- Generalized Ackermann Vehicle Model: A complex model that incorporates nonlinear frictional drag forces and nonlinear interactions between steering angle, longitudinal velocity, and lateral/yaw dynamics.
- Standard 3DOF Single-Track Model: Used as a reference model for comparative validation.
Control Strategy: Instead of traditional feedback control, the paper introduces two heuristic functions:
- Steering Function ( $\delta_s$ ): A composite function consisting of two sinusoidal components ( $\delta_1$ and $\delta_2$ ) modulated by unit step functions. $\delta_1$ is designed to bring the vehicle back toward the trajectory, while $\delta_2$ maintains the vehicle's direction once the path is reached.
- Tractive Force Function ( $F_{xt}$ ): This combines the initial post-collision force with a corrective control force ( $F_c$ ), also modeled as a time-limited sinusoidal pulse. This allows the vehicle to recover forward motion without relying solely on braking.
Collision Emulation: Collisions are modeled by applying specific initial lateral ( $v_y$ ) and angular ( $w_z$ ) velocities to the vehicle's Center of Gravity (C.G.) at $t=0$ , simulating the immediate aftermath of an impact.

3. Key Contributions

Novel Heuristic Control Law: The development of a dual-input (steering + traction) heuristic function specifically designed for the transient, high-velocity post-impact regime.
Nonlinear Dynamics Integration: Unlike many existing models, this approach explicitly accounts for time-varying longitudinal velocity and the nonlinear coupling between steering and lateral/yaw dynamics.
Versatility Demonstration: The proposed method is shown to be effective across different levels of model complexity (from a standard 3DOF model to a highly nonlinear generalized model), proving its adaptability.

4. Results and Validation

The methodology was validated through MATLAB simulations across two distinct collision scenarios:

Case 1 (Lateral Collision at C.G.): The collision induces lateral displacement but no initial yaw. The controller successfully attenuated the lateral velocity and restored the vehicle to its original straight-line path.
Case 2 (Lateral Collision off-C.G.): The collision induces both lateral displacement and a significant yaw moment (rotation). Despite the added complexity of the vehicle's orientation being altered, the joint steering and tractive force control successfully re-aligned the heading and position.

Key Findings:

The simulations showed that the vehicle could restore its trajectory while maintaining high longitudinal speeds, avoiding the need for emergency braking.
The results confirmed that the "generalized" model requires different control parameters than the "reference" model due to its nonlinearities, highlighting the importance of model accuracy in control design.

5. Significance and Future Work

This research is significant because it addresses a "safety-critical" failure mode in autonomous systems: the ability to recover after an unavoidable impact. By providing a method to manage the transient post-impact state, this work lays the groundwork for more resilient autonomous driving algorithms.

Future Directions:

Incorporating actuator saturation limits (steering angle/rate and traction limits).
Accounting for combined-slip tire-force limits.
Conducting Monte-Carlo simulations to test robustness against uncertainties in mass, friction, and tire stiffness.
Moving from open-loop heuristic control to a closed-loop feedback architecture.