Autonomous Motion Control Under Uncertain Position Estimates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for autonomous movement of mobile objects, such as robots, struggle to generate action plans when the object's position is unknown, particularly after power-up, collisions, or false environmental recognition, leading to random movement or reliance on predetermined assumptions.
Innovation Solution
A control apparatus and method that generates multiple action plan candidates based on surroundings, evaluates them using probability estimates from feature points, and determines a valid action plan through selection or merging based on evaluation values, enabling autonomous action even in situations where the object's position is unclear.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If multiple action plan candidates are generated and evaluated based on position probability, then autonomous action capability in unknown position situations is improved, but device complexity increases
Solution Approach 1:
The control apparatus segments the action planning process into three distinct functional modules: an action plan candidate generating section that creates multiple potential action plans, an action plan candidate evaluating section that assesses each candidate based on position probability, and an action plan determining section that selects the optimal plan. This segmentation allows the system to handle unknown position situations systematically while maintaining manageable complexity through modular design.
Solution Approach 2:
The system performs preliminary actions by pre-generating multiple action plan candidates before determining the final action. By preparing multiple potential plans in advance and evaluating them based on position probability estimates, the system ensures that autonomous action can proceed even when the exact position is unknown, rather than waiting for complete position information.
2Speed
If action plans are generated using only recent information, then responsiveness is improved, but reliability deteriorates in situations with no or unclear past information
Solution Approach 1:
The system dynamically adapts its information usage based on availability. When recent information is available, it uses that for rapid response. When past information is missing or unclear, it transitions to using position probability estimates and surrounding situation analysis. This dynamic adaptation maintains both responsiveness and reliability across different operational conditions.
Solution Approach 2:
The system changes the parameters used for action plan generation based on information availability. Instead of rigidly using only recent information, it adjusts to incorporate position probability estimates and surrounding situation data when recent information is insufficient, thereby maintaining reliability without sacrificing response speed.
3Measurement precision
If feature point matching is used to estimate own position, then position estimation accuracy is improved, but measurement precision requirements increase
Solution Approach 1:
The system uses feature point matching by creating a correspondence between observed features in the current environment and stored reference features. This copying approach allows position estimation without requiring direct measurement of absolute position, instead relying on relative feature correspondence that can be detected through standard image processing techniques.
Solution Approach 2:
Rather than requiring complete and precise detection of all feature points, the system accepts partial matches. By evaluating position probability based on the degree of feature point matching rather than requiring perfect matches, the system achieves practical position estimation accuracy while reducing the difficulty of feature detection and measurement requirements.
Data Source
AI summary
An action plan is generated even when the own position is unknown in order to move autonomously. A route is planned for each position constituting an own-position candidate on the basis of status of surroundings. Multiple candidates of an action plan constituting multiple action candidates are generated on the basis of the planned routes. An evaluation value is set to each of the generated multiple action plan candidates. The action plan is determined using the action plan candidates in accordance with their evaluation values. This technology is applied advantageously to multi-legged robots, flying objects, and onboard systems each controlled by an onboard computer to move autonomously.