Embedded Files
Frontiers
Frontiers
Numerical Methods and Differentiability
Numerical Methods and Differentiability
Robotics applications present unique challenges for optimization due to real-time requirements, high dimensionality, and non-smooth dynamics. This frontier explores recent advances in numerical methods and solver design aimed at meeting these demands, while also identifying key remaining obstacles and future research directions.
Robotics applications present unique challenges for optimization due to real-time requirements, high dimensionality, and non-smooth dynamics. This frontier explores recent advances in numerical methods and solver design aimed at meeting these demands, while also identifying key remaining obstacles and future research directions.
Learning and Optimization
Learning and Optimization
The integration of data-driven methods with optimization creates new opportunities, such as learning initialization heuristics, reward functions, and system dynamics. It also introduces challenges in certifiability and scalability. This frontier investigates the interplay between learning and optimization, examining how each field can effectively inform and improve the other within robotic systems.
The integration of data-driven methods with optimization creates new opportunities, such as learning initialization heuristics, reward functions, and system dynamics. It also introduces challenges in certifiability and scalability. This frontier investigates the interplay between learning and optimization, examining how each field can effectively inform and improve the other within robotic systems.
Global Optimization and Zero-Order Methods
Global Optimization and Zero-Order Methods
Many robotics problems, from task and motion planning to perception, are inherently non-convex and difficult to solve to global optimality. This frontier focuses on recent advances in algorithms with global performance ambitions, covering techniques ranging from convex relaxations to derivative-free and sampling-based methods.
Many robotics problems, from task and motion planning to perception, are inherently non-convex and difficult to solve to global optimality. This frontier focuses on recent advances in algorithms with global performance ambitions, covering techniques ranging from convex relaxations to derivative-free and sampling-based methods.
Target speakers and audience
Target speakers and audience
Our target speakers and panelists are will include:
Our target speakers and panelists are will include:
- Pioneers in optimization and control theory whose foundational work is enabling new possibilities in robotics, but whose voices are not always heard at mainstream robotics conferences.
- Robotics researchers and practitioners applying novel concepts from optimization theory to robotics; finding, translating and adapting these concepts for the efficient use in the field of robotics.
This workshop is for you if you are:
This workshop is for you if you are:
- A researcher or practitioner seeking to expand your toolkit of optimization methods to tackle more challenging robotics problems.
- A developer of optimization algorithms who wants to understand the unique demands of the robotics domain and find new avenues for impact.
- A student or early-career researcher eager to learn from the leading minds in the field and identify the most exciting open questions for your future work.
Page updated
Report abuse