Formally Guaranteed Control Adaptation for ODD-Resilient Autonomous Systems

Gricel Vázquez; Calum Imrie; Sepeedeh Shahbeigi; Nawshin Mannan Proma; Tian Gan; Victoria J Hodge; John Molloy; Simos Gerasimou

cs.LOApr 8, 2026

Formally Guaranteed Control Adaptation for ODD-Resilient Autonomous Systems

Authors: Gricel Vázquez, Calum Imrie, Sepeedeh Shahbeigi, Nawshin Mannan Proma, Tian Gan, Victoria J Hodge, John Molloy, Simos Gerasimou

Organizations: Department of Computer Science, University of York, UK · Cyprus University of Technology, Limassol, Cyprus

Abstract

Ensuring reliable performance in situations outside the Operational Design Domain (ODD) remains a primary challenge in devising resilient autonomous systems. We explore this challenge by introducing an approach for adapting probabilistic system models to handle out-of-ODD scenarios while, in parallel, providing quantitative guarantees. Our approach dynamically extends the coverage of existing system situation capabilities, supporting the verification and adaptation of the system's behaviour under unanticipated situations. Preliminary results demonstrate that our approach effectively increases system reliability by adapting its behaviour and providing formal guarantees even under unforeseen out-of-ODD situations.

Figures & tables

Figure 1. SAVE approach overview: pre-deployment (A,B) and deployment (M,A,P,E) phases with shared knowledge.

ID	Property violated	Worst critic- ality score	SAVE success (no violations)	Critical situations avoided
1	[ $\Phi_{2}$ ]	0.03890	True	[s2, s3]
2	[ $\Phi_{2}$ , $\Phi_{1}$ ]	0.03482	True	[s2, s3, s5]
3	[ $\Phi_{2}$ , $\Phi_{1}$ ]	0.03556	True	[s3, s4]
4	[ $\Phi_{2}$ , $\Phi_{1}$ ]	0.04483	True	[s5, s1, s3, s2]
5	[ $\Phi_{2}$ , $\Phi_{1}$ ]	0.04905	False	-
…16	[ $\Phi_{2}$ ]	0.04994	True	[s1]

Table 1. SAVE adaptation results after a violation is detected. Without adaptation, the baseline fails in all cases.

Event	Baseline Outcome	SAVE Response
Vessel deceleration (out-of-ODD)	Collision expected within the next 50 timesteps ( $f2$ ), detected at time $t_{1}$ . Collision happens in situation $i$ at time $t_{2}$ .	Controller adapted at $t_{1}$ . Critical situation $i$ and collision avoid.
Requirement $\Phi_{1}$ (safe TTC)	Violated at time $t_{2}$ .	Satisfied post-adaptation at time $t_{3}$
Requirement $\Phi_{2}$ (near collision)	Violated at time $t_{4}$ .	Satisfied post-adaptation at time $t_{5}$ .

Table 2. SAVE adaptation example.

Figure 2. SAVE execution times to get criticality scores.

Explore similar work

Staggered Integral Online Conformal Prediction for Safe Dynamics Adaptation with Multi-Step Coverage Guarantees
Apr 7, 2026Daniel M. Cherenson, Dimitra PanagouConformal Prediction Adaptive Conformal Inference
Robotics Department, University of Michigan, Ann Arbor, MI, USA
Latent Dynamics-Aware OOD Monitoring for Trajectory Prediction with Provable Guarantees
Mar 15, 2026Tongfei Guo, Lili SuOOD Detection Change-Point Detection
Department of Electrical and Computer Engineering, Northeastern University, Boston, MA, USA.
Neural Barriers: An Online Certifiable Learning-enhanced Adaptive High Order Safety Critical Control
Oct 4, 2026Lishuo Pan, Mattia Catellani, An Cao +2Safety-Critical Control Control Barrier Functions
Department of Computer Science, Brown University, Providence, RI 02912 USA · Department of Sciences and Methods for Engineering, University of Modena and Reggio Emilia, 41121 Modena, Italy · Division of Physics, Mathematics and Astronomy, California Institute of Technology, Pasadena, CA 91125 USA

Apr 7, 2026J/K move · Enter open · S save

Staggered Integral Online Conformal Prediction for Safe Dynamics Adaptation with Multi-Step Coverage Guarantees

Daniel M. Cherenson, Dimitra Panagou

Robotics Department, University of Michigan, Ann Arbor, MI, USA

Conformal Prediction Adaptive Conformal Inference Safety-Critical Control

Open paper AskPDF arXiv

Safety-critical control of uncertain, adaptive systems often relies on conservative, worst-case uncertainty bounds that limit closed-loop performance. Online conformal prediction is a powerful data-driven method for quantifying uncertainty when truth values of predicted outputs are revealed online; however, for systems that adapt the dynamics without measurements of the state derivatives, standard online conformal prediction is insufficient to quantify the model uncertainty. We propose Staggered Integral Online Conformal Prediction (SI-OCP), an algorithm utilizing an integral score function to quantify the lumped effect of disturbance and learning error. This approach provides long-run coverage guarantees, resulting in long-run safety when synthesized with safety-critical controllers, including robust tube model predictive control. Finally, we validate the proposed approach through a numerical simulation of an all-layer deep neural network (DNN) adaptive quadcopter using robust tube (MPC), highlighting the applicability of our method to complex learning parameterizations and control strategies.

Abstract

Explore similar work

Staggered Integral Online Conformal Prediction for Safe Dynamics Adaptation with Multi-Step Coverage Guarantees

Latent Dynamics-Aware OOD Monitoring for Trajectory Prediction with Provable Guarantees

Neural Barriers: An Online Certifiable Learning-enhanced Adaptive High Order Safety Critical Control