Robotic MPC Trajectory Priors for Crowded Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Standard Information Theoretic Model Predictive Control (IT-MPC) techniques face challenges in navigating mobile robots in crowded environments, as they tend to exhibit poor convergence to optimal control sequences due to significant changes in optimal control sequences, such as new goal locations or unexpected dynamic obstacles, leading to inefficient trajectory planning.
Innovation Solution
An informed sampling process is introduced, combining Information Theoretic MPC with deep reinforcement learning, using a data-driven trajectory prediction model, such as a neural network, to generate control trajectory priors based on the current state of the environment and detected obstacles, thereby improving convergence by focusing on low-cost areas of the state space.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard IT-MPC generates control samples by local sampling around the previous control sequence, then the method performs well when optimal control sequence changes slightly, but poor convergence occurs when optimal control sequence significantly changes due to new goal locations or unexpected dynamic obstacles
Solution Approach 1:
The method performs preliminary exploration by generating control samples not only around the previous control sequence but also around randomly generated control sequences. This preliminary action prepares the system to quickly adapt to significant environmental changes by having diverse control samples ready, improving convergence reliability when optimal control sequences change dramatically due to new goals or dynamic obstacles
Solution Approach 2:
The sampling strategy dynamically adjusts based on the situation: it uses local sampling when optimal control sequences change slightly (maintaining efficiency) and incorporates random sampling when significant changes occur (improving adaptability). This dynamic adjustment of sampling behavior resolves the contradiction between reliability and adaptability across different operational conditions
2Measurement precision
If IT-MPC iteratively evaluates multiple control trajectory samples at each time step, then the trajectory planning becomes more accurate, but the computational complexity and processing time increase
Solution Approach 1:
The method evaluates multiple control trajectory samples (excessive action) to ensure high trajectory planning accuracy, but applies this extensively only when necessary (e.g., when significant environmental changes are detected). In stable environments, it relies more on the previous control sequence with fewer samples, thus balancing accuracy requirements with computational complexity constraints
Solution Approach 2:
The system changes the number of control trajectory samples evaluated based on environmental conditions and convergence status. When the optimal control sequence changes significantly or accuracy requirements are high, it increases the number of samples evaluated. When conditions are stable and convergence is good, it reduces the number of samples, thereby adjusting the trade-off between trajectory planning accuracy and computational complexity dynamically
Data Source
AI summary
A computer-implemented method for determining a control trajectory for a robotic device. The method includes: performing an information theoretic model predictive control applying a control trajectory sample prior in each time step to obtain a control trajectory for a given time horizon; determining the control trajectory sample prior depending on a data-driven trajectory prediction model which is trained to output a control trajectory sample as the control trajectory sample prior based on an actual state of the robotic device.