Experimental Validation of Safe MPC for Autonomous Driving in Uncertain Environments

Batkovic, Ivo; Gupta, Ankit; Zanon, Mario; Falcone, Paolo · 2023 · Crossref

DOI: 10.1109/tcst.2023.3291562

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the challenge of ensuring provably safe autonomous driving in uncertain environments where the future motions of other road users (e.g., pedestrians, cyclists) cannot be known a priori. While Model Predictive Control (MPC) is widely used for vehicle planning, existing methods often rely on assumptions that are impractical for general urban settings, such as the existence of a central coordinator or the ability to guarantee recursive feasibility in the presence of uncontrolled obstacles. The authors propose a framework based on Model Predictive Flexible Trajectory Tracking Control (MPFTC) that guarantees constraint satisfaction at all times while tracking a reference trajectory as closely as obstacles allow. The methodology involves formulating an MPC problem that incorporates auxiliary inputs to adjust the timing of the reference trajectory, allowing the vehicle to slow down or stop safely if constraints become tight. The framework distinguishes between a-priori known constraints (e.g., vehicle dynamics limits) and a-priori unknown constraints (e.g., collision avoidance with dynamic obstacles). To ensure safety, the authors define specific assumptions regarding system regularity, reference feasibility, stabilizing terminal conditions, and the dynamics of unknown constraints. Specifically, they require that predicted reachable sets for other road users are consistent over time, ensuring that if a position is deemed safe at one step, it remains safe in subsequent predictions. The vehicle dynamics are modeled using a single-track model expressed in the frame of the reference path, with quadratic costs penalizing deviations from the reference. The study validates the framework through both simulations and real-world experiments on a test vehicle. The authors demonstrate that the proposed safe MPC controller maintains recursive feasibility and avoids collisions even in complex traffic situations. Simulations illustrate how the controller adapts to unforeseen events by modifying the reference trajectory's timing rather than violating safety constraints. The real-world experiments confirm the framework's real-time capability and effectiveness in handling uncertain environments, showing that the vehicle can safely navigate interactions with other road users without requiring cooperative coordination. The significance of this work lies in providing a general, verifiable framework for safe autonomous driving that does not rely on idealized assumptions about environment predictability or coordination. By integrating flexible trajectory tracking with robust constraint handling, the approach ensures that the vehicle can always find a safe control action, such as stopping, if necessary. This contributes to the broader goal of deploying fully autonomous systems in general urban settings by addressing the critical open problem of guaranteeing safety in the presence of unpredictable human-driven agents.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Theoretical Contribution: computational model