Predicting initialization effectiveness for trajectory optimization

Trajectory optimization is a method for solving motion planning problems by formulating them as non-convex constrained optimization problems. The optimization process, however, can get stuck in local optima that are in collision. As a consequence, these methods typically require multiple initializations. This poses the problem of deciding which initializations to use when given a limited computational budget. In this paper we propose a machine learning approach to predict whether a collision-free solution will be found from a given initialization. We present a set of trajectory features that encode the obstacle distribution locally around a robot. These features are designed for generalization across different tasks. Our experiments on various planning benchmarks demonstrate the performance of our approach.

Authors

• Jia Pan 0001
• Zhuo Chen
• Pieter Abbeel

Keywords

• Trajectory
• Optimization
• Vectors
• Robots
• Planning
• Benchmark testing
• Collision avoidance
• Trajectory Optimization
• Optimization Problem
• Local Optimum
• Path Planning
• Non-convex Problem
• Constrained Optimization
• Constrained Optimization Problem
• Non-convex Optimization
• Trajectory Features
• Degrees Of Freedom
• Training Set
• Eigenvalues
• Learning Algorithms
• Objective Function
• Combination Of Features
• Final Outcome
• Inequality Constraints
• Good And Bad
• Hessian Matrix
• Spherical Harmonics
• Good Initial Guess
• Good Initialization
• Distance Vector
• General Optimization Problem
• Spherical Functions
• Different Combinations Of Features
• Task Planning
• Non-convex Objective
• Convex Objective
• Optimization Algorithm