Motion Trajectory Control Using Neural-Guided Mixed-Integer Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion planning systems for multiple moving devices in a workspace with obstacles require significant computational and memory resources, often leading to collisions due to the complexity of mixed-integer programming (MIP) problems, especially when using machine learning (ML) approaches that produce continuous trajectories for discontinuous obstacle avoidance.
Innovation Solution
The system employs a machine learning module, such as a neural network, to predict integer variables for a mixed-integer optimization problem, allowing for the reduction of the problem to a real-valued optimization, which is then solved using simpler methods, and includes focused sampling and barrier functions to enhance feasibility and reduce computational burden.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mixed-integer programming (MIP) is used to solve motion planning problems, then optimal motion trajectories can be obtained, but computational resources and time required increase significantly
Solution Approach 1:
The patent segments the MIP problem into two parts: (1) using ML to predict integer variables that represent discrete motion decisions, and (2) solving only the remaining real-valued optimization problem. This segmentation reduces the computational complexity from NP-Hard to polynomial time while maintaining collision avoidance guarantees through the barrier function formulation.
Solution Approach 2:
The patent applies preliminary action by using machine learning models to predict integer variables before solving the optimization problem. These predictions serve as initial guesses or constraints that guide the subsequent real-valued optimization, avoiding the need to search the entire discrete solution space and significantly reducing computation time.
2Productivity
If machine learning approaches are used to approximate MIP solutions, then computational requirements are reduced, but collision avoidance reliability decreases due to continuous trajectory approximation
Solution Approach 1:
The patent applies local quality by using different computational approaches for different parts of the problem: machine learning provides local predictions for integer variables (discrete decisions), while a barrier function-based optimization ensures global collision avoidance constraints are satisfied. This hybrid approach maintains reliability in critical areas while improving efficiency overall.
Solution Approach 2:
The patent introduces barrier functions as intermediaries that bridge the ML predictions and the final optimized trajectory. The barrier functions act as constraints that guide the optimization process to satisfy collision avoidance requirements, ensuring that the final trajectory is both efficient and reliable.
3Reliability
If full MIP solving is implemented in embedded processing units, then optimal solutions are obtained, but device complexity and memory requirements increase
Solution Approach 1:
The patent extracts the computationally intensive integer variable solving part from the embedded system and replaces it with a lighter ML-based prediction approach. The embedded system only needs to handle the real-valued optimization with barrier functions, significantly reducing memory requirements and computational complexity while maintaining solution quality.
Data Source
AI summary
A controller for controlling a motion of at least one device subject to constraints on the motion, is disclosed. The controller comprises a processor and a memory, where the controller inputs parameters of the task including the state of the at least one device to a neural network trained to output an estimated motion trajectory for performing the task. Further, the controller extracts at least some of the integer values of a solution to a mixed-integer optimization problem for planning an execution of the task that results in the estimated motion trajectory. Further, the controller solves the mixed-integer optimization problem for the parameters of the task with corresponding integer values fixed to the extracted integer values to produce an optimized motion trajectory subject to the constraint and changes the state of the at least one device to track the optimized motion trajectory.