Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics
From MaRDI portal
Abstract: Robotic algorithms typically depend on various parameters, the choice of which significantly affects the robot's performance. While an initial guess for the parameters may be obtained from dynamic models of the robot, parameters are usually tuned manually on the real system to achieve the best performance. Optimization algorithms, such as Bayesian optimization, have been used to automate this process. However, these methods may evaluate unsafe parameters during the optimization process that lead to safety-critical system failures. Recently, a safe Bayesian optimization algorithm, called SafeOpt, has been developed, which guarantees that the performance of the system never falls below a critical value; that is, safety is defined based on the performance function. However, coupling performance and safety is often not desirable in robotics. For example, high-gain controllers might achieve low average tracking error (performance), but can overshoot and violate input constraints. In this paper, we present a generalized algorithm that allows for multiple safety constraints separate from the objective. Given an initial set of safe parameters, the algorithm maximizes performance but only evaluates parameters that satisfy safety for all constraints with high probability. To this end, it carefully explores the parameter space by exploiting regularity assumptions in terms of a Gaussian process prior. Moreover, we show how context variables can be used to safely transfer knowledge to new situations and tasks. We provide a theoretical analysis and demonstrate that the proposed algorithm enables fast, automatic, and safe optimization of tuning parameters in experiments on a quadrotor vehicle.
Recommendations
- Bayesian optimisation for constrained problems
- SMGO-\(\Delta\): balancing caution and reward in global optimization with black-box constraints
- No-regret Bayesian optimization with unknown hyperparameters
- Robust reinforcement learning with Bayesian optimisation and quadrature
- Data-driven algorithm selection and tuning in optimization and signal processing
Cites work
- A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems
- A taxonomy of global optimization methods based on response surfaces
- Convergence rates of efficient global optimization algorithms
- Gaussian processes for machine learning.
- scientific article; zbMATH DE number 43985 (Why is no real title available?)
- scientific article; zbMATH DE number 1094954 (Why is no real title available?)
- scientific article; zbMATH DE number 2107836 (Why is no real title available?)
- Information-Theoretic Regret Bounds for Gaussian Process Optimization in the Bandit Setting
- Kernels for vector-valued functions: a review
- No-regret Bayesian optimization with unknown hyperparameters
- Posterior consistency of Gaussian process prior for nonparametric binary regression
- Provably safe and robust learning-based model predictive control
- Reinforcement learning. An introduction
- Safe Exploration of State and Action Spaces in Reinforcement Learning
- Support Vector Machines
This page was built for publication: Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6053801)
