Session Overview
Social robots are becoming integral to everyday life — in homes, care settings, schools, public spaces — yet the static models that power them often fail to adapt beyond controlled environments. To thrive in long-term, open-ended deployment, robots must be continual learners and generative thinkers.
This special session investigates how Continual Learning and Generative AI can jointly provide the foundation for social robots that learn, generalise, and create over time. Robots should be able to continuously assimilate new experiences without forgetting, while also generating contextually appropriate behaviours, dialogue, and adaptations. Together, these paradigms chart a path toward robots that grow with their users, refine their skills post-deployment, and engage in meaningful, human-like interaction.
We invite contributions spanning theory, architecture, and demonstration — including reinforcement and meta-learning, memory systems, multimodal post-training (e.g. LLMs and VLMs), and embodied generative frameworks — all aimed at socially intelligent, self-improving, and ethically aware robotic agents.
Key Themes
- Continual and Lifelong Learning in Social Robotics
- Generative AI for Behavioural, Cognitive, and Affective Modelling
- Post-Training and Online Adaptation for Robotic Intelligence
- Memory, Retention, and Knowledge Consolidation in Continual Systems
- Multi-Modal Generative Perception (Vision, Language, and Action)
- Human-in-the-Loop and Co-Adaptive Learning
- Evaluation, Safety, and Ethics of Continually Evolving Robots
List of Topics
Vision
We envision robots that continue to grow and adapt alongside humans — fine-tuning their behaviour, perception, and interaction style over weeks, months, and years. The combination of continual learning and generative AI enables this vision: robots that reason, create, and evolve.
These systems hold promise in domains such as assistive care, education, public engagement, and collaborative work, where long-term adaptability and trust are essential.
Related Research by the Organisers & References
Architecture integrating memory to guide adaptation in interactive robot systems.
Adaptive pipeline leveraging behaviour context to condition robot responses.
Adaptive attention mechanisms in social robots for gaze control and focus behaviour.
A recent survey and synthesis of continual learning approaches in robotic systems, highlighting challenges, benchmarks, and architectures.
Invitation
We welcome submissions from across robotics, AI, HRI, cognitive systems, and applied domains — especially those exploring how robots can retain knowledge, adapt over time, and generate creative responses.
Whether via algorithms, systems, or user studies, your contribution can push forward the frontier of socially intelligent, evolving robots.
Important Dates
Submission Deadline: TBA
Notification: TBA
Camera-ready: TBA
Dates will be announced soon. Check back for updates.
Conference: July 1-4, 2026
Location: University of London, London, UK
Organisers
Khashayar Ghamati
PhD Candidate in AI & Robotics
Robotics Research Group
University of Hertfordshire, UK