Runtime Self-Optimization Through Knowledge-Driven Design Space Reduction
21st International Conference on Software Engineering for Adaptive and Self-Managing Systems (SEAMS ’26), Rio de Janeiro, Brazil, pp. 1–12
- Status
- Published — Version of record available from ACM via the DOI.
- Date
- License
- CC BY 4.0
- Download PDF (Authors' version (CC BY 4.0))
- ACM version of record
Abstract
Modern Cyber-Physical Systems must operate in volatile environments, with conditions that are often unpredictable at design time, requiring engineering solutions that are able to learn and optimize their configuration at runtime. However, the dimensions of modern systems, which are often complex and highly interconnected, make self-optimization at runtime a challenging task. In this paper, we propose a novel approach to runtime self-optimization that leverages knowledge-driven design space reduction to enable efficient and formally correct optimization of complex systems. Contrary to naïve re-optimization strategies, which are often prohibitive, or heuristic search methods, which sacrifice optimality, our approach guarantees optimal solutions with heuristics-level runtime performance. Specifically, by storing and reusing proven-optimal configurations and intelligently pruning the design space based on learned knowledge, we are able to significantly reduce the search space, hereby enabling fast re-optimization. Upon changes in the operational domain, the knowledge base is efficiently queried and updated, with regions expansion and merging to maintain compactness. The key idea behind the approach is to use Zonotopes as the underlying representation, which offer a compact and efficient way to represent and manipulate high-dimensional design spaces. We evaluate the approach on a smart-farming drone scenario, comparing against common optimization methods demonstrating that our approach not only preserves optimality, but also improves performance up to ∼56x compared to baseline methods with negligible memory and computation overhead.
Summary
Research question
Cyber-physical systems such as autonomous vehicles and agricultural drones are configured at design time for a specific operational design domain (ODD). Once deployed, weather, new regulations or ageing components shift the operating conditions, and a configuration that was optimal at deployment can become suboptimal. Self-optimizing systems — a class of self-adaptive systems — should then find a new optimal configuration at runtime. Existing options force a trade-off: re-solving the full optimization problem is too slow for large, interconnected design spaces, while heuristic or search-based methods are fast but give up optimality guarantees. The paper asks whether runtime re-optimization can be both exact and fast.
Approach
The paper proposes knowledge-driven design space reduction, embedded in a digital-twin architecture with a MAPE-K feedback loop. Whenever an optimal configuration is computed for some operating conditions, it is stored in a knowledge base together with the region of conditions for which it is proven optimal. When conditions change, the knowledge base is queried first: a cached configuration is reused directly if it still applies, and otherwise the optimizer searches only the regions of the design space that have not been explored before. Regions are expanded and merged as knowledge accumulates, keeping the knowledge base compact.
The key representation choice is the zonotope, a compact geometric description of high-dimensional sets that supports efficient retrieval, expansion and merging. The motivating scenario comes from a multi-year national project on sustainable autonomous farming, where drones adjust parameters such as altitude, speed, spray rate or sensor resolution to changing conditions, and one drone can reuse configurations already found by another.
Results
On a smart-farming drone scenario, the approach is compared with common optimization methods. The authors report that it preserves optimality while improving performance by up to about 56× over the baselines, with negligible memory and computational overhead for managing the knowledge base.
Why it matters
For researchers in self-adaptive systems and search-based software engineering, the paper shows a way to combine the guarantees of exact optimization with runtime performance comparable to heuristics, by treating previously solved problems as reusable, formally scoped knowledge rather than discarding them.
Keywords
runtime self-optimization; self-adaptive systems; cyber-physical systems; design space reduction; zonotopes; MAPE-K; digital twins; operational design domain
Cite
@inproceedings{brodo2026selfoptimization,
author = {Brodo, Luca and Scalora, Giuseppe and Henkler, Stefan},
title = {Runtime Self-Optimization Through Knowledge-Driven Design Space Reduction},
booktitle = {Proceedings of the 21st International Conference on Software Engineering for Adaptive and Self-Managing Systems},
series = {SEAMS '26},
pages = {1--12},
year = {2026},
address = {Rio de Janeiro, Brazil},
publisher = {Association for Computing Machinery},
doi = {10.1145/3788550.3794861},
url = {https://doi.org/10.1145/3788550.3794861}
}Related papers
-
Towards a Digital Twin Framework for Secure and Efficient Cyber-Physical Transportation Systems
MoVe4SPS 2025 Workshop
-
Asymmetric Timed Refinement for Correct Post-Deployment Updates in Real-Time Systems
MEMOCODE 2026AcceptedBest Paper Award