Automated Repair of Cyber-Physical Systems
Pablo Valle · 2025
Read Automated Repair of Cyber-Physical Systems by Pablo Valle (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17678.txt.
Introduction
Pablo Valle sets out a doctoral research plan for automatically repairing the software of cyber-physical systems in 2025. These systems mix code with physical machinery, from robots to satellites, and checking their software by hand consumes much of the development budget.
Automated program repair is mature for ordinary software but barely tried here, where fault localisation is hard, tests run slowly and fitness scoring is weak. Valle proposes combining spectrum-based fault localisation with patch generation and modern AI methods. The plan will be tested on open and industrial code bases. The goal is repair tools that scale to the long, expensive test cycles of embedded systems.
Catalog record
| CREATOR | Pablo Valle |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 11 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-17678
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Automated Repair of Cyber-Physical Systems Pablo Valle Abstract Cyber-Physical Systems (CPS) integrate digital technologies with physical processes and are common in different domains and industries, such as robotic systems, autonomous vehicles or satellites. Debugging and verification of CPS software consumes much of the development budget as it is often purely manual. To speed up this process, Automated Program Repair (APR) has been targeted for a long time. Although there have been advances in software APR and CPS verification techniques, research specifically on APR for CPSs is limited. This Ph.D. research project aims to develop scalable APR techniques for CPSs, addressing problems of fault localization, long test execution times, and fitness function limitations. A new method combining spectrum-based fault localization (SBFL) with patch generation and advanced artificial intelligence techniques will be investigated. The approach will be validated by empirical studies on open and industrial code bases of CPSs. I Motivation Cyber-Physical Systems (CPS) integrate digital technologies with physical processes [ 1 ] , and can be found in different domains and industries. Debugging, testing and verification are estimated to account for 50% to 75% of the total software development cost budget [ 2 ] . Software debugging and verification is important as bugs are part of human nature. Because of this, Automated Program Repair (APR) has been deemed as a long-standing dream among software engineers. APR aims at providing an appropriate patch to a software program that contains a fault (i.e., a bug). Although this is a complex task, its applicability would save hundreds of hours to individual software engineers every year. For this reason, in recent years, the software engineering research community has been devising novel approaches targeting APR [ 3 ] . Despite this growth, as well as the growth of articles in the field of CPS verification and testing [ 4 , 5 , 6 ] , only three studies address APR of CPSs [ 7 , 8 , 9 ] . Of these three studies, two of them [ 7 , 8 ] , focus on misconfiguration repair. This is because CPSs are highly configurable and there may be cases where misconfiguration leads to failure of the running system [ 7 ] . The other paper [ 9 ] focuses on faults in attribute interactions in autonomous driving systems. However, none of these studies has considered faults introduced by CPS programmers at the code level, which unlike simple software, CPS code is tightly coupled with physical components and processes. This code must interact with hardware, manage real-time constraints, and ensure safety-critical operations, making it more complex than traditional software. Additionally, CPS code must handle uncertainties inherent in physical processes, which makes faults in CPS code have more severe consequences, requiring specialized techniques for repair compared to simple software. Traditional APR techniques can be classified into three types [ 10 ] : (1) search-based, (2) constraint-based and (3) learning-based, in which recently there has been a strong trend to include Large Language Models (LLM) to provide solutions to the detected faults. Constraint-based techniques face scalability problems [ 10 ] and, therefore, are not applicable in CPSs. Traditional learning techniques based on Neural Machine Translation (NMT) learn correctness patterns from different projects, limiting solutions to common bugs. This Ph.D. research work focuses on researching about the integration between the first and third type techniques.That is, a search-based technique will be used as the main part of the solution, as it can explore multiple potential parches for bugs at the same time and is better suited for CPSs due to its ability to directly optimize within a well-defined search space, providing more predictable and verifiable solutions. Moreover, we propose to integrate the archive-based search algorithm with an LLM that helps to provide possible fixes for detected bugs, thus overcoming some limitations of previously implemented fixes, e.g., related to fixes based on correction patterns implemented programmatically. II Problem Statement and Hypothesis In APR, GenProg [ 11 ] the most known tool, which uses genetic programming to evolve real patches by maintaining and modifying program variants across generations. Genetic operators like crossover and mutation are applied to evolve potential solutions. Moreover, most APR techniques [ 10 ] incorporate a pre-processing phase using fault localization methods such as Spectrum-Based Fault Localization (SBFL) to prioritize suspicious code fragments, improving repair efficiency. In this way, potentially faulty code elements are prioritized and modifications are targeted to the most suspicious code fragments. Inspired by GenProg, other search-based approaches, such as Arja [ 12 ] have also been proposed. Recently, LLM-based approaches [ 13 ] have gained attention for generating large numbers of patches to fix errors. However, all these approaches face scalability challenges when applying to CPSs due to the following issues: I1 - Ineffective fault localization: The number of test cases in CPSs is usually small, but test runs are long and system-wide, resulting in large code coverage with few test cases [ 14 ] . This makes current fault-finding techniques inaccurate, increasing the search space for potential patches and the likelihood of over-fitted patches. I2 - Long test execution time: CPS tests are expensive and time consuming [ 5 ] , lasting several minutes [ 7 , 9 ] compared to the milliseconds of unit tests of current APR approaches [ 10 , 12 , 13 ] , which makes them inapplicable in the CPS domain due to the significant time required to validate potential patches. For example, initializing GenProg would require 10 hours applied to an industrial case study to which this thesis will have access [ 7 ] . I3 - Use of the test suite as fitness function: Current approaches [ 10 , 12 , 13 ] use the number of failed and passed test cases as a fitness function, which is problematic in CPS due to the low number of test cases and the lack of internal information of test runs, deriving in almost random exploration of patches. This Ph.D. research aims at generating new knowledge in the field of APR by proposing a novel technique that addresses these three critical issues overlooked by current methods and that to the best of my knowledge, it has not been explored so far. The technique begins with a time-aware SBFL approach to localize faults, tackling the first issue (I1). The results feed into a search algorithm that integrates LLMs with an archive-based strategy to track partial repairs, addressing the second issue (I2) and differing from existing population-based algorithms. This archive evolves new patches using two strategies: one leveraging test case details like failure criticality and timing [ 9 ] , and another utilizing CPS failure diagnosis [ 15 ] to assist LLMs in patch generation. This offers higher granularity in evaluating partial patches, resolving the third issue (I3) for scalability in CPSs. Based on these conjectures, the initial hypothesis of this thesis can be formulated as follows: Starting Hypothesis: The inclusion of the notion of time in spectrum-based fault localization (SBFL) techniques combined with archive-based search algorithms that leverage LLMs can automate the repair process of CPSs’ software faults within an affordable time budget, saving significant debugging cost to CPS developers. III Research Objectives The main objective of this thesis is to