RSS 2025 Workshop on Reliable Robotics

RSS 2025 Workshop on Reliable Robotics:
Safety and Security in the Face of Generative AI

Email: rss2025-reliable-robotics AT googlegroups DOT com

Overview of Workshop

The groundbreaking capabilities of generative AI (GenAI) provide a trove of opportunities for robotics, enabling general-purpose skill acquisition and contextual reasoning for more intelligent and reliable operations beyond the traditional sense. However, the emergence of GenAI also introduces additional safety and security challenges for real-world robot autonomy (e.g., handling uncertainties and deliberate adversaries). These safety and security issues exacerbated by GenAI are ingrained in all aspects of robot autonomy, ranging from perception to understanding, reasoning and action.

This half-day workshop brings together researchers from academia and industry, drawing on expertise in cybersecurity, robot perception, learning, control, explainable AI, and evaluation for real-world robotics, to foster an interdisciplinary exchange on reliable robot autonomy with two connected themes: (i) enhanced reliable robot autonomy enabled by large-scale models and (ii) safety and security challenges introduced by GenAI in robotics. By uniting diverse perspectives, the workshop aims to drive discussion on pressing challenges at the intersection of GenAI, safety, and security in robotics while further promoting collaboration across disciplines.

Topics of Interest

Safe & secure deployment of language models for robots
- Can LLMs be trusted to enforce the safety of robots?
- How can we stop automated attacks by LLMs?
- What tools are needed to reliably test LLM-based robotic solutions?
Use of generative AI to address safety concerns
- How can language models be used to verify safe robot behavior or enable safe behavior specification?
- How can generative models be used effectively for safer sim-to-real transfers?
- How can we best leverage foundation models to build environment representations that would facilitate downstream reasoning and safe operation?
Leveraging generative AI to recover from off-nominal conditions
- Can the multimodal capabilities of GenAI improve the resilience of robots to cyber faults and real-world failures?
- How can foundation models compensate for uncertainty and disturbances?
- Can GenAI be leveraged to enhance system redundancy and resilience?
Combatting advanced adversarial attacks on robotics systems
- What is the ideal integration of safety and security measures and methods?
- What measures and benchmarks are required to evaluate attacker-robust algorithms?
- What security vulnerabilities to physical and cyber-attacks are introduced by incorporating generative AI in robots?

Call for Contributions

We invite researchers to share their findings with the community by submitting short, non-archival papers (4 pages + references). Submissions will be presented in-person during a short spotlight talk and a poster session on-site during the workshop.

Location: Olin Hall of Engineering, room 100D, or Zoom

Important Dates:

Submission Deadline: May 30, 2025, 23:59 (AOE)
Reviews and Decisions: June 6, 2025, 23:59 (AOE)
Workshop date: June 21, 2025, at USC, Los Angeles

Submission Instructions:

Recently published papers and novel research are both welcome
Single blind RSS format
Submission link: OpenReview

The contributed papers will be made available on this website. However, this does not constitute an archival publication, and no formal workshop proceedings will be made available, meaning contributors are free to publish their work in archival journals or conferences.

We will highlight the Best Paper Award and Most Popular Poster Award during the closing remarks at the workshop event.