Mobile Robot Trajectory Planning With Learned Optimizer Initialization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Constrained optimization-based planners for mobile robots face challenges in converging to globally optimal trajectories in real-time due to slow convergence and susceptibility to local optima, especially with complex cost functions and models, impacting performance.
Innovation Solution
A two-stage optimization approach using a trained function approximator to provide high-quality initialization for a runtime optimizer, combining a simpler first stage with a more complex second stage to mitigate local optima and ensure real-time convergence to globally optimal trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a constrained non-linear optimization problem is formulated for mobile robot planning, then the robot can navigate while guaranteeing safety constraints, but the convergence to optimal control inputs becomes slow and uncertain
Solution Approach 1:
The patent applies preliminary action by using a trained function approximator to generate an initial guess for the optimization algorithm before the main optimization process begins. This pre-computed initialization is based on learned patterns from training data, allowing the optimizer to start closer to the optimal solution and converge faster while still guaranteeing safety constraints through the constrained optimization framework.
2Measurement precision
If the complexity of the cost function and mobile robot model increases, then the planning accuracy improves, but the optimizer takes longer to converge or may never successfully converge
Solution Approach 1:
The trained function approximator serves as an intermediary between the complex cost function/model and the optimizer. It learns the mapping from scenarios to optimal control inputs during training, and during execution, it provides an informed initial guess that guides the optimizer through the complex landscape, enabling the system to handle high complexity while maintaining fast convergence.
3Reliability
If non-linear constrained optimization solvers are used, then locally optimal solutions can be found, but the solver has a tendency to converge to local optima which may be far from the globally optimal solution
Solution Approach 1:
The trained function approximator performs preliminary action by providing a well-informed initial guess that is already close to the global optimum based on learned patterns from training data. This significantly reduces the risk of the non-linear solver getting trapped in local optima, as the optimization starts from a position much closer to the global solution rather than a random or naive initial point.
Data Source
AI summary
A computer system for planning mobile robot trajectories, the computer system comprising: an input configured to receive a set of scenario description parameters describing a scenario and a desired goal for the mobile robot therein; a runtime optimizer configured to compute a final mobile robot trajectory that substantially optimizes a cost function for the scenario, subject to a set of hard constraints that the final mobile robot trajectory is guaranteed to satisfy; and a trained function approximator configured to compute, from the set of scenario description parameters, initialization data defining an initial mobile robot trajectory. The computer system is configured to initialize the runtime optimizer with the initialization data, in order to guide the optimizer from the initial mobile robot trajectory to the final mobile robot trajectory that satisfies the hard constraints, the function approximator having been trained on example sets of scenario description parameters and ground truth initialization data for the example sets of scenario description parameters.