Stanford Intelligent Systems Laboratory

New academic year, new SISL lab photo!

New academic year, new SISL lab photo!

We’re proposing a workshop at ICLR 2025 in Singapore on human-AI co-evolution—exploring how humans shape AI and how AI is shaping us across diverse fields like robotics, healthcare, education, legal systems, social media, and public service. We’re looking for participants from academia and industry to join this important conversation. Whether you're researching new AI trends, navigating legislation, or developing cutting-edge applications, we want to hear from you! Interested in submitting your work, speaking (keynote/panel), or attending? Fill out our form to help us gauge interest —no formal commitment needed at this stage.   Form: https://lnkd.in/gFwrsFCa

SISL is excited to be at #IROS2024! "Semantic Belief Behavior Graph: Enabling Autonomous Robot Inspection in Unknown Environments" Authors: Muhammad Fadhil Ginting, David D. Fan, Sung Kyun Kim, Mykel Kochenderfer, Ali Agha Paper: https://lnkd.in/g4Gu_PvE Session: Thursday, 17 October, 09:00-10:00, ThPI4T08.01 See you there!

Consider an agent that is tasked with exploring a large unknown environment. The agent must collect measurements to build an accurate representation of the state of the environment. The agent has a finite set of resources and is therefore constrained by time, battery life, or fuel capacity. As a result, it must plan a path to maximize the amount of information acquired from its noisy observations while remaining within the allowed resource budget. This is known as the informative path planning (IPP) problem and involves choosing a subset of sensing locations from the environment. A variety of real-world problems can be formulated as IPP problems, such as planetary rover exploration, search and rescue, and environmental monitoring. IPP merges the fields of robotics, artificial intelligence, and spatial data analysis. In the recent work by Joshua Ott, Mykel Kochenderfer and Stephen P. Boyd, we consider the problem of finding an informative path through a graph, given initial and terminal nodes and a given maximum path length. We assume that a linear noise corrupted measurement is taken at each node of an underlying unknown vector that we wish to estimate. The informativeness is measured by the reduction in uncertainty in our estimate, evaluated using several Gaussian process-based metrics. We present a convex relaxation for an informative path planning problem, which we can readily solve to obtain a bound on the possible performance. We develop an approximate sequential method where the path is constructed segment by segment through dynamic programming. This involves solving an orienteering problem, with the node reward acting as a surrogate for informativeness, taking the first step, and then repeating the process. The method scales to very large problem instances and achieves performance close to the bound produced by the convex relaxation. We also demonstrate our method’s ability to handle adaptive objectives, multimodal sensing, and multi-agent variations of the informative path planning problem. Learn more: https://lnkd.in/gPE5PWfj

Automation in aviation has recently gained technological advancements based on data-driven models applied to vision, decision-making, planning, and human interaction. In the recent work presented at Digital Avionics Systems Conference (DASC) 2024, Romeo Valentin, Sydney Katz, Joonghyun Lee, Don Walker, Matt Sorgenfrei, and Mykel Kochenderfer address the challenge of probabilistic parameter estimation given measurement uncertainty in real-time. We provide a general formulation and apply this to pose estimation for an autonomous visual landing system. We present three probabilistic parameter estimators: a least- squares sampling approach, a linear approximation method, and a probabilistic programming estimator. To evaluate these estimators, we introduce novel closed-form expressions for mea- suring calibration and sharpness specifically for multivariate normal distributions. Our experimental study compares the three estimators under various noise conditions. We demonstrate that the linear approximation estimator can produce sharp and well- calibrated pose predictions significantly faster than the other methods but may yield overconfident predictions in certain scenarios. Additionally, we demonstrate that these estimators can be integrated with a Kalman filter for continuous pose estimation during a runway approach where we observe a 50% improvement in sharpness while maintaining marginal calibration. This work contributes to the integration of data-driven computer vision models into complex safety-critical aircraft systems and provides a foundation for developing rigorous certification guidelines for such systems. Learn more: https://lnkd.in/g4kRe4yY Video: https://lnkd.in/g75kFQ7c

Consider an agent that is tasked with exploring a large unknown environment. The agent must collect measurements to build an accurate representation of the state of the environment. The agent has a finite set of resources and is therefore constrained by time, battery life, or fuel capacity. As a result, it must plan a path to maximize the amount of information acquired from its noisy observations while remaining within the allowed resource budget. This is known as the informative path planning (IPP) problem and involves choosing a subset of sensing locations from the environment. A variety of real-world problems can be formulated as IPP problems, such as planetary rover exploration, search and rescue, and environmental monitoring. IPP merges the fields of robotics, artificial intelligence, and spatial data analysis. In the recent work by Joshua Ott, Mykel Kochenderfer and Stephen P. Boyd, we consider the problem of finding an informative path through a graph, given initial and terminal nodes and a given maximum path length. We assume that a linear noise corrupted measurement is taken at each node of an underlying unknown vector that we wish to estimate. The informativeness is measured by the reduction in uncertainty in our estimate, evaluated using several Gaussian process-based metrics. We present a convex relaxation for an informative path planning problem, which we can readily solve to obtain a bound on the possible performance. We develop an approximate sequential method where the path is constructed segment by segment through dynamic programming. This involves solving an orienteering problem, with the node reward acting as a surrogate for informativeness, taking the first step, and then repeating the process. The method scales to very large problem instances and achieves performance close to the bound produced by the convex relaxation. We also demonstrate our method’s ability to handle adaptive objectives, multimodal sensing, and multi-agent variations of the informative path planning problem. Learn more: https://lnkd.in/gPE5PWfj

Validation is a critical component of the development process for decision-making systems in a variety of domains including autonomous vehicles, robotics, and healthcare. At the recent Stanford Center for AI Safety annual meeting, Sydney Katz presented the ongoing work on the upcoming textbook Algorithms for Validation, which offers a broad introduction to algorithms for validating safety-critical systems. Check out the highlight from Sydney’s talk: https://lnkd.in/eAUn9k-p A draft of the textbook is available at: https://lnkd.in/eBRJaTs9 Authors: Mykel Kochenderfer, Sydney Katz, Anthony Corso, and Robert Moss

Join Our AI Benchmarking Study! We are studying how AI benchmarks are used across different fields. We are seeking researchers, industry professionals, and policymakers who have used AI benchmarks to participate in a 45-60 minute interview. If you are interested in contributing to our study, please visit the link below for more details. Project Team: Amelia Hardy, Anka Reuel, Dylan Asmar, Lisa Soder, Kiana Jafari, Allie Griffith, Sanmi Koyejo, Mykel Kochenderfer Link: https://lnkd.in/g8sPK94R

Algorithms for Decision Making by Mykel Kochenderfer, Tim Wheeler, and Kyle Wray is now available in Korean! For more details: https://lnkd.in/gmUM2Gjh English version: https://lnkd.in/gJUn4_d