Linear Constrained Optimization is a powerful tool capable of solving a range of optimization tasks, with millions of variables and constraints. Discover more in Joshua Ott's lecture, covering KKT conditions, the simplex algorithm, and dual certificates. youtu.be/O0GN1AquD6o?si…

To plan safely in uncertain environments, agents must balance utility with safety constraints. Learn more about ConstrainedZero, accepted to IJCAI and authored by Robert Moss and other SISLers, which advances how we solve CC-POMDPs in belief space. arxiv.org/abs/2405.00644

To solve high-dimensional POMDPs, prior methods use online planning with problem-specific heuristics. Robert Moss's accepted paper for RLC 2024, BetaZero, proposes a belief-state planning algorithm that replaces heuristics with learned approximations. arxiv.org/abs/2306.00249

Excited to announce that Fadhil Ginting's paper on a new probabilistic planning method for object-goal navigation, using relational semantic knowledge and prior spatial configurations for real-world inspection, has been accepted to #RSS2024. Learn more: arxiv.org/abs/2405.09822

Deep RL is often seen as a black box with limited explainability and suboptimal performance. Jiachen Li’s publication integrates auxiliary tasks with spatio-temporal relational reasoning into standard DRL, enhancing both performance and explainability. arxiv.org/abs/2311.16091

Validating safety-critical autonomous systems in high-dimensional domains is challenging. Harrison Delecki's work explores using conditional denoising diffusion models for better sample efficiency and mode coverage. Learn more: arxiv.org/abs/2406.14761

Autonomous robotic inspection in complex, unknown environments is a critical challenge. In collaboration with Field AI, Fadhil Ginting proposes a belief-space task planning framework for semantic-based navigation, accepted to IROS 2024! arXiv: arxiv.org/abs/2401.17191

Our new paper "Open Problems in Technical AI Governance" led by Ben Bucknall & me is out! We outline 89 open technical issues in AI governance, plus resources and 100+ research questions that technical experts can tackle to help AI governance efforts🧵 t.ly/Y-mQ1

🥁And the #cav24 Award goes to...🥁 Clark Barrett Stanford University, David Dill Stanford University, Kyle Julian Wing, Guy Katz HUJI CSE and Mykel Kochenderfer Mykel Kochenderfer Stanford University for their #cav17 paper “Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks” Congratulations! 👏

👏👏👏

We introduce history-dependent dual variables in Lagrangian-guided MCTS to tackle myopic action selection in CPOMDPs. Discover recent work by Paula Stocco, Suhas Chundi, Arec Jamgochian, and Mykel Kochenderfer: youtube.com/watch?v=yCZCk7…

Congratulations to Joshua Ott on successfully defending his PhD on Autonomous Exploration of Unknown Environments! Efficient and safe exploration demands strategies that quantify uncertainty, optimize resources, and maximize information gain. Learn more: youtube.com/watch?v=UFA4zK…

Excited to share XPlaneConnectX, a cross-platform tool by our PhD candidate Marc Schlichting! It works seamlessly with X-Plane 12, using the built-in UDP feature—no plugin needed. Compatible with Python 3 & Julia, it's perfect for research. Check it out: t.ly/LUp2h

“algorithms for decision making”, Massachusetts Institute of Technology (MIT) publishing book is freely available.

Validation is key for safety-critical systems like autonomous vehicles and robotics. At the recent Stanford's AI Safety Annual Meeting, Sydney Katz presented the ongoing work on the book Algorithms for Validation: youtube.com/watch?v=V-meos…. Book draft: algorithmsbook.com/validation

Excited to share our new website on Open Problems in Technical AI Governance, a companion to our recent paper on the topic! Check out the website here → taig.stanford.edu ← and see the thread below for more details. Huge thanks to Anka Reuel for putting this together!

Technical AI Governance research is moving quickly (yay!), so Ben Bucknall and I are excited to launch a living repository of open problems and resources in the field, based on our recent paper where we identified 100+ research questions in TAIG: taig.stanford.edu 🧵

Many problems, such as rover exploration, can be framed as informative path planning. Joshua Ott tackles the problem of finding an informative path through a graph with specified start and terminal nodes, within a maximum path length. Learn more: authors.elsevier.com/c/1jtVf3HdG3p4U~

In the recent work presented at DASC 2024, Romeo Valentin addresses probabilistic parameter estimation under measurement uncertainty in real-time, applying the approach to pose estimation for an autonomous visual landing system. youtube.com/watch?v=BWiqpm…

Paper: arxiv.org/abs/2407.16223