Software Engineering for Self-Adaptive Robotics: A Research Agenda

Sartaj, Hassan; Ali, Shaukat; Cavalcanti, Ana; Esterle, Lukas; Gomes, Claudio; Larsen, Peter Gorm; Tefas, Anastasios; Woodcock, Jim; Zhang, Houxiang

Software Engineering for Self-Adaptive Robotics

A Research Agenda

This page presents an extended version of Table 1 and Table 2 from the research paper listed below. Table 1 provides a detailed analysis of cross-cutting quality aspects for all challenges identified across the software engineering lifecycle (Requirements Engineering, Software Design, Software Development, Software Testing, and Operations) and enabling technologies (Digital Twins and Artificial Intelligence) for self-adaptive robotics. Table 2 maps self-adaptive robotics characteristics to all challenges across both the software engineering lifecycle and the enabling technologies. For a full description of the analysis and results, please refer to the paper below.

Reference Research Paper

Hassan Sartaj, Shaukat Ali, Ana Cavalcanti, Lukas Esterle, Cláudio Gomes, Peter Gorm Larsen, Anastasios Tefas, Jim Woodcock, Houxiang Zhang, "Software Engineering for Self-Adaptive Robotics: A Research Agenda," arXiv preprint arXiv:2505.19629, 2025.

📄 View on arXiv 📥 Download PDF

Table 1: Mapping of challenges in self-adaptive robotics (SAR) to cross-cutting quality aspects. Rows with more “+” symbols indicate higher criticality of the corresponding challenge, while columns with more “+” symbols show higher importance of the associated quality aspect.

Challenge ID	Challenge Description	Safety	Performance	Reliability	Explainability	Privacy	Adaptability	Security	Criticality
Requirements Engineering
SRE-Ch-1	Handling runtime requirement updates while ensuring system consistency and safety.	+		+			+		3
SRE-Ch-2	Balancing conflicting requirements.			+			+		2
SRE-Ch-3	Ensuring traceability of requirements and scalability of verification and validation activities.	+		+			+		3
SRE-Ch-4	Eliciting requirements due to diverse stakeholders and the nature of robotics applications.	+			+		+		3
SRE-Ch-5	Integrating ethical requirements into SRE processes.				+	+			2
NR-Ch-1	Assuring that the robot meets all normative requirements upon initial deployment, and continuously satisfies norms while working in uncertain, evolving scenarios.	+		+			+		3
NR-Ch-2	Providing clear explanations for norm-sensitive decisions.				+				1
Software Design
SD-Ch-1	Scalable and modular architectures that balance flexibility and efficiency while supporting continuous adaptation across diverse platforms.		+				+		2
SD-Ch-2	Managing trade-offs between adaptability, safety, and performance so that reliability is preserved without degrading system behaviour.	+	+	+			+		4
SD-Ch-3	Safe runtime evolution and reconfiguration with consistency and fault tolerance during software updates in operation.	+		+			+		3
SD-Ch-4	Handling uncertainty introduced by AI-driven adaptation, including mitigating unpredictability and ensuring trustworthiness.			+			+		2
SD-Ch-5	Keeping runtime models consistent with design-time models to support lifecycle adaptation in MDE workflows.			+			+		2
SD-Ch-6	Human-in-the-loop adaptation with interfaces and processes that minimise cognitive load and enable effective collaboration.	+			+		+		3
SD-Ch-7	Verification and validation techniques suited to highly dynamic and safety-critical self-adaptive software.	+		+			+		3
Software Design: Model-Driven Engineering
MDE-Ch-1	Lack of a formal standard for self-adaptive architectures (e.g., MAPE-K), hindering traceability between models and code.			+	+		+		3
MDE-Ch-2	Limited support for automated code generation that reflects adaptation-specific semantics.		+	+			+		3
MDE-Ch-3	Need for hybrid reasoning that combines software with models of hardware and environment.	+	+	+			+		4
MDE-Ch-4	Insufficient integration of AI components into MDE frameworks.			+	+		+		3
MDE-Ch-5	Limited treatment of human and ethical factors in adaptation decisions.	+			+	+	+		4
Software Design: Simulation
Sim-Ch-1	Compact representation of scenario ranges for adaptive simulations.			+			+		2
Sim-Ch-2	Coupling heterogeneous subsystem simulations to form a composite simulation.			+			+		2
Sim-Ch-3	Orchestration algorithms that can dynamically adapt to system reconfiguration.		+	+			+		3
Sim-Ch-4	Co-simulation interfaces often lack the richness required to capture multi-rate and hybrid dynamics.		+	+					2
Sim-Ch-5	Protection of intellectual property when coupling externally provided subsystem models.					+		+	2
Software Development
Dev-Ch-1	Ensuring effective integration and compatibility between various hardware and software components from different vendors.		+	+			+		3
Dev-Ch-2	Adapting to continuously evolving hardware and software without rebuilding existing architectures.		+	+			+		3
Dev-Ch-3	Developing real-time sensor fusion to integrate diverse sensor data.	+	+	+			+		4
Dev-Ch-4	Developing robotic software in an energy-efficient manner and providing sustainable operations in real environments.		+	+			+		3
Dev-Ch-5	Ensuring privacy, data security, and robustness against cyberattacks at runtime.	+				+		+	3
Software Testing
ST-Ch-1	Bridging the reality gap in simulation-based testing to ensure realistic yet cost-effective testing.	+	+	+			+		4
ST-Ch-2	Testing the interactions between AI and non-AI components, while considering uncertainty and ethical concerns.	+		+	+				3
ST-Ch-3	Testing SAR software in the presence of uncertainty.	+		+			+		3
ST-Ch-4	Ensuring the correctness of individual MAPE-K components and the overall loop during testing.	+	+	+			+		4
ST-Ch-5	Real-time testing of digital twins, while accounting for synchronisation delays with physical systems.		+	+			+		3
ST-Ch-6	Developing cost-effective regression testing methods to support the continuous evolution of SAR.	+		+	+		+		4
ST-Ch-7	Testing robustness and reliability of robotic foundation models.	+		+	+				3
ST-Ch-8	Ensuring resilience against adversarial threats, such as prompt-based and perception-based attacks.	+		+	+			+	4
Operations
Op-Ch-1	Overlapping model state estimation.			+			+		2
Op-Ch-2	Uncertainty as first-class citizen in MAPLE-K Loops.	+		+	+		+		4
Op-Ch-3	Active uncertainty reduction.	+		+			+		3
Op-Ch-4	Dynamic software updating for SARs.	+		+			+		3
Op-Ch-5	Continuous Dev(Sec)Ops pipelines for heterogeneous SAR fleets.	+		+			+	+	4
Digital Twins
DT-Ch-1	Ensuring trustworthy predictions due to data quality and uncertainties from physical systems.	+		+	+				3
DT-Ch-2	Requires sufficient computing infrastructure to support timely predictions for decision-making.	+	+	+					3
DT-Ch-3	Determining the exact moment of state transitions in physical systems.			+	+		+		3
DT-Ch-4	Composing DTs for collaborating robots.					+		+	2
DT-Ch-5	Adapting DT models across all software engineering lifecycle phases with sufficient fidelity.		+	+			+		3
DT-Ch-6	Integrating complex subsystems in large-scale robotic DTs.		+	+			+		3
DT-Ch-7	Achieving a fully integrated, dynamic, and responsive DT representation of multidisciplinary robotic systems.		+	+			+		3
Artificial Intelligence
AI-Ch-1	Gathering sufficient data for self-adaptation in AI-powered robotics is often expensive or infeasible.		+	+			+		3
AI-Ch-2	Retaining previously acquired knowledge while adapting to new data and avoiding catastrophic forgetting.	+		+			+		3
AI-Ch-3	Addressing AI overfitting to improve generalisation and prevent operational anomalies in self-adaptive robots.	+		+			+		3
AI-Ch-4	Ensuring trustworthiness in self-adaptive robots, including handling hallucinations, real-time uncertainty, and ethical concerns.	+		+	+	+		+	5
Quality Aspect Focus (Column Totals)		28	18	44	15	6	41	6	-

Table 2: Mapping of all challenges in self-adaptive robotics (SAR) to their corresponding characteristics. Rows with more “+” symbols indicate relevance of the characteristic as part of the challenge.

Challenge ID	Challenge Description	Embodiment	Physical Interaction	Localised	Heterogeneity	Distributed Nature	Modularity
Requirements Engineering
SRE-Ch-1	Handling runtime requirement updates while ensuring system consistency and safety.		+		+	+	+
SRE-Ch-2	Balancing conflicting requirements.	+			+	+
SRE-Ch-3	Ensuring traceability of requirements and scalability of verification and validation activities.					+
SRE-Ch-4	Eliciting requirements due to diverse stakeholders and the nature of robotics applications.	+	+
SRE-Ch-5	Integrating ethical requirements into SRE processes.	+		+
NR-Ch-1	Assuring that the robot meets all normative requirements upon initial deployment, and continuously satisfies norms while working in uncertain, evolving scenarios.	+	+		+		+
NR-Ch-2	Providing clear explanations for norm-sensitive decisions.			+	+	+	+
Software Design
SD-Ch-1	Scalable and modular architectures that balance flexibility and efficiency while supporting continuous adaptation across diverse platforms.					+	+
SD-Ch-2	Managing trade-offs between adaptability, safety, and performance so that reliability is preserved without degrading system behaviour.	+	+		+
SD-Ch-3	Safe runtime evolution and reconfiguration with consistency and fault tolerance during software updates in operation.	+	+		+
SD-Ch-4	Handling uncertainty introduced by AI-driven adaptation, including mitigating unpredictability and ensuring trustworthiness.	+	+	+
SD-Ch-5	Keeping runtime models consistent with design-time models to support lifecycle adaptation in MDE workflows.				+		+
SD-Ch-6	Human-in-the-loop adaptation with interfaces and processes that minimise cognitive load and enable effective collaboration.	+	+	+
SD-Ch-7	Verification and validation techniques suited to highly dynamic and safety-critical self-adaptive software.		+	+	+		+
Software Design: Model-Driven Engineering
MDE-Ch-1	Lack of a formal standard for self-adaptive architectures (e.g., MAPE-K), hindering traceability between models and code.	+		+	+	+	+
MDE-Ch-2	Limited support for automated code generation that reflects adaptation-specific semantics.	+		+	+		+
MDE-Ch-3	Need for hybrid reasoning that combines software with models of hardware and environment.		+	+	+
MDE-Ch-4	Insufficient integration of AI components into MDE frameworks.	+			+		+
MDE-Ch-5	Limited treatment of human and ethical factors in adaptation decisions.	+	+	+
Software Design: Simulation
Sim-Ch-1	Compact representation of scenario ranges for adaptive simulations.	+	+	+
Sim-Ch-2	Coupling heterogeneous subsystem simulations to form a composite simulation.		+		+	+	+
Sim-Ch-3	Orchestration algorithms that can dynamically adapt to system reconfiguration.			+	+	+
Sim-Ch-4	Co-simulation interfaces often lack the richness required to capture multi-rate and hybrid dynamics.				+		+
Sim-Ch-5	Protection of intellectual property when coupling externally provided subsystem models.	+
Software Development
Dev-Ch-1	Ensuring effective integration and compatibility between various hardware and software components from different vendors.	+					+
Dev-Ch-2	Adapting to continuously evolving hardware and software without rebuilding existing architectures.	+					+
Dev-Ch-3	Developing real-time sensor fusion to integrate diverse sensor data.	+	+	+
Dev-Ch-4	Developing robotic software in an energy-efficient manner and providing sustainable operations in real environments.	+		+	+	+
Dev-Ch-5	Ensuring privacy, data security, and robustness against cyberattacks at runtime.	+	+	+
Software Testing
ST-Ch-1	Bridging the reality gap in simulation-based testing to ensure realistic yet cost-effective testing.	+	+
ST-Ch-2	Testing the interactions between AI and non-AI components, while considering uncertainty and ethical concerns.				+		+
ST-Ch-3	Testing SAR software in the presence of uncertainty.	+	+
ST-Ch-4	Ensuring the correctness of individual MAPE-K components and the overall loop during testing.	+		+	+		+
ST-Ch-5	Real-time testing of digital twins, while accounting for synchronisation delays with physical systems.		+		+	+
ST-Ch-6	Developing cost-effective regression testing methods to support the continuous evolution of SAR.	+			+
ST-Ch-7	Testing robustness and reliability of robotic foundation models.	+	+		+		+
ST-Ch-8	Ensuring resilience against adversarial threats, such as prompt-based and perception-based attacks.	+
Operations
Op-Ch-1	Overlapping model state estimation.	+			+		+
Op-Ch-2	Uncertainty as first-class citizen in MAPLE-K Loops.	+	+	+
Op-Ch-3	Active uncertainty reduction.	+	+	+	+	+	+
Op-Ch-4	Dynamic software updating for SARs.	+			+		+
Op-Ch-5	Continuous Dev(Sec)Ops pipelines for heterogeneous SAR fleets.	+			+	+
Digital Twins
DT-Ch-1	Ensuring trustworthy predictions due to data quality and uncertainties from physical systems.	+	+	+	+
DT-Ch-2	Requires sufficient computing infrastructure to support timely predictions for decision-making.		+	+		+
DT-Ch-3	Determining the exact moment of state transitions in physical systems.		+
DT-Ch-4	Composing DTs for collaborating robots.				+	+
DT-Ch-5	Adapting DT models across all software engineering lifecycle phases with sufficient fidelity.	+			+
DT-Ch-6	Integrating complex subsystems in large-scale robotic DTs.				+	+	+
DT-Ch-7	Achieving a fully integrated, dynamic, and responsive DT representation of multidisciplinary robotic systems.				+	+	+
Artificial Intelligence
AI-Ch-1	Gathering sufficient data for self-adaptation in AI-powered robotics is often expensive or infeasible.	+	+	+
AI-Ch-2	Retaining previously acquired knowledge while adapting to new data and avoiding catastrophic forgetting.	+	+
AI-Ch-3	Addressing AI overfitting to improve generalisation and prevent operational anomalies in self-adaptive robots.	+	+
AI-Ch-4	Ensuring trustworthiness in self-adaptive robots, including handling hallucinations, real-time uncertainty, and ethical concerns.	+	+