Self-adaptive robotic systems operate autonomously in dynamic and uncertain environments, requiring robust real-time monitoring and adaptive behaviour. Unlike traditional robotic software with predefined logic, self-adaptive robots exploit artificial intelligence (AI), machine learning, and model-driven engineering to adapt continuously to changing conditions, thereby ensuring reliability, safety, and optimal performance. This paper presents a research agenda for software engineering in self-adaptive robotics, structured along two dimensions. The first concerns the software engineering lifecycle, requirements, design, development, testing, and operations, tailored to the challenges of self-adaptive robotics. The second focuses on enabling technologies such as digital twins, AI-driven adaptation, and quantum computing, which support runtime monitoring, fault detection, and automated decision-making. We identify open challenges, including verifying adaptive behaviours under uncertainty, balancing trade-offs between adaptability, performance, and safety, and integrating self-adaptation frameworks like MAPE-K/MAPLE-K. By consolidating these challenges into a roadmap toward 2030, this work contributes to the foundations of trustworthy and efficient self-adaptive robotic systems capable of meeting the complexities of real-world deployment.

Recently, machine learning (ML) has become a popular approach to support self-adaptation. ML has been used to deal with several problems in self-adaptation, such as maintaining an up-to-date runtime model under uncertainty and scalable decision-making. Yet, exploiting ML comes with inherent challenges. In this paper, we focus on a particularly important challenge for learning-based self-adaptive systems: drift in adaptation spaces. With adaptation space we refer to the set of adaptation options a self-adaptive system can select from at a given time to adapt based on the estimated quality properties of the adaptation options. Drift of adaptation spaces originates from uncertainties, affecting the quality properties of the adaptation options. Such drift may imply that eventually no adaptation option can satisfy the initial set of the adaptation goals, deteriorating the quality of the system, or adaptation options may emerge that allow enhancing the adaptation goals. In ML, such shift corresponds to novel class appearance, a type of concept drift in target data that common ML techniques have problems dealing with. To tackle this problem, we present a novel approach to self-adaptation that enhances learning-based self-adaptive systems with a lifelong ML layer. We refer to this approach as lifelong self-adaptation. The lifelong ML layer tracks the system and its environment, associates this knowledge with the current tasks, identifies new tasks based on differences, and updates the learning models of the self-adaptive system accordingly. A human stakeholder may be involved to support the learning process and adjust the learning and goal models. We present a general architecture for lifelong self-adaptation and apply it to the case of drift of adaptation spaces that affects the decision-making in self-adaptation. We validate the approach for a series of scenarios using the DeltaIoT exemplar.

Self-adaptive systems offer several attack surfaces due to the communication via different channels and the different sensors required to observe the environment. Often, attacks cause safety to be compromised as well, making it necessary to consider these two aspects together. Furthermore, the approaches currently used for safety and security analysis do not sufficiently take into account the intermediate steps of an adaptation. Current work in this area ignores the fact that a self-adaptive system also reveals possible vulnerabilities (even if only temporarily) during the adaptation. To address this issue, we propose a modeling approach that takes into account the different relevant aspects of a system, its adaptation process, as well as safety hazards and security attacks. We present several models that describe different aspects of a self-adaptive system and we outline our idea of how these models can then be combined into an Attack-Fault Tree. This allows modeling aspects of the system on different levels of abstraction and co-evolve the models using transformations according to the adaptation of the system. Finally, analyses can then be performed as usual on the resulting Attack-Fault Tree.

Self-adaptive software can assess and modify its behavior when the assessment indicates that the program is not performing as intended or when improved functionality or performance is available. Since the mid-1960s, the subject of system adaptivity has been extensively researched, and during the last decade, many application areas and technologies involving self-adaptation have gained prominence. All of these efforts have in common the introduction of self-adaptability through software. Thus, it is essential to investigate systematic software engineering methods to create self-adaptive systems that may be used across different domains. The primary objective of this research is to summarize current advances in awareness requirements for adaptive strategies based on an examination of state-of-the-art methods described in the literature. This paper presents a review of self-adaptive systems in the context of requirement awareness and summarizes the most common methodologies applied. At first glance, it gives a review of the previous surveys and works about self-adaptive systems. Afterward, it classifies the current self-adaptive systems based on six criteria. Then, it presents and evaluates the most common self-adaptive approaches. Lastly, an evaluation among the self-adaptive models is conducted based on four concepts (requirements description, monitoring, relationship, dependency/impact, and tools).

Design time uncertainty poses an important challenge when developing a self-adaptive system. As an example, defining how the system should adapt when facing a new environment state, requires understanding the precise effect of an adaptation, which may not be known at design time. Online reinforcement learning, i.e., employing reinforcement learning (RL) at runtime, is an emerging approach to realizing self-adaptive systems in the presence of design time uncertainty. By using Online RL, the self-adaptive system can learn from actual operational data and leverage feedback only available at runtime. Recently, Deep RL is gaining interest. Deep RL represents learned knowledge as a neural network whereby it can generalize over unseen inputs, as well as handle continuous environment states and adaptation actions. A fundamental problem of Deep RL is that learned knowledge is not explicitly represented. For a human, it is practically impossible to relate the parametrization of the neural network to concrete RL decisions and thus Deep RL essentially appears as a black box. Yet, understanding the decisions made by Deep RL is key to (1) increasing trust, and (2) facilitating debugging. Such debugging is especially relevant for self-adaptive systems, because the reward function, which quantifies the feedback to the RL algorithm, must be defined by developers. The reward function must be explicitly defined by developers, thus introducing a potential for human error. To explain Deep RL for self-adaptive systems, we enhance and combine two existing explainable RL techniques from the machine learning literature. The combined technique, XRL-DINE, overcomes the respective limitations of the individual techniques. We present a proof-of-concept implementation of XRL-DINE, as well as qualitative and quantitative results of applying XRL-DINE to a self-adaptive system exemplar.

Modern systems are designed to operate in increasingly variable and uncertain environments. Not only are these environments complex, in the sense that they contain a tremendous number of variables, but they also change over time. Systems must be able to adjust their behaviour at run-time to manage these uncertainties. These self-adaptive systems have been studied extensively. This paper proposes a definition of a safety-critical self-adaptive system and then describes a taxonomy for classifying adaptations into different types based on their impact on the system's safety and the system's safety case. The taxonomy expresses criteria for classification and then describes specific criteria that the safety case for a self-adaptive system must satisfy, depending on the type of adaptations performed. Each type in the taxonomy is illustrated using the example of a safety-critical self-adaptive water heating system.

Many software systems today face uncertain operating conditions, such as sudden changes in the availability of resources or unexpected user behavior. Without proper mitigation these uncertainties can jeopardize the system goals. Self-adaptation is a common approach to tackle such uncertainties. When the system goals may be compromised, the self-adaptive system has to select the best adaptation option to reconfigure by analyzing the possible adaptation options, i.e., the adaptation space. Yet, analyzing large adaptation spaces using rigorous methods can be resource- and time-consuming, or even be infeasible. One approach to tackle this problem is by using online machine learning to reduce adaptation spaces. However, existing approaches require domain expertise to perform feature engineering to define the learner, and support online adaptation space reduction only for specific goals. To tackle these limitations, we present 'Deep Learning for Adaptation Space Reduction Plus' -- DLASeR+ in short. DLASeR+ offers an extendable learning framework for online adaptation space reduction that does not require feature engineering, while supporting three common types of adaptation goals: threshold, optimization, and set-point goals. We evaluate DLASeR+ on two instances of an Internet-of-Things application with increasing sizes of adaptation spaces for different combinations of adaptation goals. We compare DLASeR+ with a baseline that applies exhaustive analysis and two state-of-the-art approaches for adaptation space reduction that rely on learning. Results show that DLASeR+ is effective with a negligible effect on the realization of the adaptation goals compared to an exhaustive analysis approach, and supports three common types of adaptation goals beyond the state-of-the-art approaches.

In this paper, we focus on exploiting neural networks for the analysis and planning stage in self-adaptive architectures. The studied motivating cases in the paper involve existing (legacy) self-adaptive architectures and their adaptation logic, which has been specified by logical rules. We further assume that there is a need to endow these systems with the ability to learn based on examples of inputs and expected outputs. One simple option to address such a need is to replace the reasoning based on logical rules with a neural network. However, this step brings several problems that often create at least a temporary regress. The reason is the logical rules typically represent a large and tested body of domain knowledge, which may be lost if the logical rules are replaced by a neural network. Further, the black-box nature of generic neural networks obfuscates how the systems work inside and consequently introduces more uncertainty. In this paper, we present a method that makes it possible to endow an existing self-adaptive architectures with the ability to learn using neural networks, while preserving domain knowledge existing in the logical rules. We introduce a continuum between the existing rule-based system and a system based on a generic neural network. We show how to navigate in this continuum and create a neural network architecture that naturally embeds the original logical rules and how to gradually scale the learning potential of the network, thus controlling the uncertainty inherent to all soft computing models. We showcase and evaluate the approach on representative excerpts from two larger real-life use cases.

Autonomous robots may be able to adapt their behavior in response to changes in the environment. This is useful, for example, to efficiently handle limited resources or to respond appropriately to unexpected events such as faults. The architecture of a self-adaptive robot is complex because it should include automatic mechanisms to dynamically configure the elements that control robot behaviors. To facilitate the construction of this type of architectures, it is useful to have general solutions in the form of software tools that may be applicable to different robotic systems. This paper presents an original algorithm to dynamically configure the control architecture, which is applicable to the development of self-adaptive autonomous robots. This algorithm uses a constraint-based configuration approach to decide which basic robot behaviors should be activated in response to both reactive and deliberative events. The algorithm uses specific search heuristics and initialization procedures to achieve the performance required by robotic systems. The solution has been implemented as a software development tool called Behavior Coordinator CBC (Constraint-Based Configuration), which is based on ROS and open source, available to the general public. This tool has been successfully used for building multiple applications of autonomous aerial robots.

Security attacks present unique challenges to self-adaptive system design due to the adversarial nature of the environment. However, modeling the system as a single player, as done in prior works in security domain, is insufficient for the system under partial compromise and for the design of fine-grained defensive strategies where the rest of the system with autonomy can cooperate to mitigate the impact of attacks. To deal with such issues, we propose a new self-adaptive framework incorporating Bayesian game and model the defender (i.e., the system) at the granularity of components in system architecture. The system architecture model is translated into a Bayesian multi-player game, where each component is modeled as an independent player while security attacks are encoded as variant types for the components. The defensive strategy for the system is dynamically computed by solving the pure equilibrium to achieve the best possible system utility, improving the resiliency of the system against security attacks.