The emergence of robotics foundation models, including large language models (LLMs), vision-language-action (VLA) models, and multimodal generative AI, has the potential to significantly transform robotic system design. Originally leveraged for natural language interactions, these models have started to show remarkable capabilities when applied to autonomous tasks and adaptive behavior planning, across various robotic applications. This opens exciting opportunities for new design paradigms that blend human creativity, AI-driven reasoning, and semi- or fully-automated experimentation.

This Research Topic seeks to highlight novel research at the intersection of foundation models and robotic system design, encouraging contributions highlighting both theoretical advancements and practical implementations. We encourage submissions from diverse disciplines, including (but not limited to) mechanical design, software design, control design, human-centered design, and cognitive science.

Topics of Interest:

We welcome original research contributions and comprehensive review articles on topics including, but not limited to:

1. Human-AI Co-Design and Iterative Development

• Co-design using robotics foundation models for iterative, AI-assisted robotic system development
o AI-driven iterative design loops with minimal human intervention
o (Semi-)Automated experimentation guided by AI-generated hypotheses
o Robotics foundation models enabling rapid prototyping and experimental iteration
o Augmentation of real-world experimental design, control, and testing data sets with physics/AI-derived data sets
• User-centered design involving approaches that integrate human factors, usability testing, and user experience into AI-driven design loops

2. Collaboration, Cognition, and Decision-Making

• Instruction-tuned LLMs and multimodal models for collaborative robotics and co-manipulation tasks
• LLMs for high-level task planning and goal decomposition for lower-level robot controllers
• Cognitive cloning for replicating expert knowledge and reasoning for enhanced robot design
• Explainability, transparency, and trust in robotic system designs driven by robotics foundation models
• Safe, responsible, and regulatory-compliant use of foundation models in robotic systems, including considerations of bias, data privacy, and fail-safe mechanisms.

3. Robotic System Architecture

• Robotics foundation models enhancing manipulation capabilities, with implications for grasper and manipulator design
• Adaptive and morphologically dynamic robot designs driven by foundation models
• Robotics foundation models for design and optimization of soft, deformable, wearable, and bio-inspired robotic systems
• Modular and origami-based robots benefiting from foundation model-guided reconfigurability
• Integration of robotics foundation models with smart materials and novel actuators to achieve unprecedented functionality
• Use of VLA models in digital twin environments for iterative design, testing, and operation of robotic systems

4. Robotic Applications of Foundation Models

• Low-level perception and planning behaviors in coordination with high-level planning
• Rewards and motion constraints using low- and high-level perception
• Generating maps and scene graphs from multimodal using low- and high-level perception for task planning or user’s situational awareness
• End-to-end learning to correlate low-level perception to high-level commands using data augmentation
• Benchmarking strategies, simulation-to-reality transfer assessments, and performance metrics (e.g., sample efficiency, safety, reliability) for foundation-model-based robotics applications.

Submissions should offer novel insights into the integration, optimization, and practical application of robotics foundation models in the robotic system design process.

Are you interested in contributing a manuscript to this Research Topic?

Foundation Models for Robot Design