TOA Guidance Learning for Lightweight UAV Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current navigation systems for unmanned aerial vehicles (UAVs) that rely on image processing to detect obstacles are computationally expensive and heavy, making them impractical for commercial and recreational use due to the high computational and storage resources required for real-time image processing.
Innovation Solution
A machine learning system, known as the Sense and Guide Machine Learning (SGML) system, is trained using time-of-arrival (TOA) information from sensor arrays to determine object locations, allowing for lightweight and efficient guidance of UAVs by generating guidance information that can be used to avoid obstacles or navigate to targets, reducing the computational burden and weight of onboard systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing systems are used to detect obstacles in real-time, then obstacle detection accuracy is improved, but computational expense and system weight increase significantly
Solution Approach 1:
The patent replaces complex mechanical image processing systems with an acoustic sensor array that uses sound wave propagation and time-difference-of-arrival measurements to detect and locate objects. This substitution eliminates the need for computationally intensive image processing while maintaining object detection capabilities through acoustic signal analysis.
Solution Approach 2:
The system changes the measurement parameter from visual image data to acoustic time-of-arrival data. By measuring the time difference of acoustic signal arrival at multiple sensor elements, the system directly computes object position without requiring complex image processing algorithms, thereby reducing computational expense while preserving detection accuracy.
2Productivity
If high-end processing systems are deployed for real-time image processing, then real-time obstacle detection is achieved, but system weight and power consumption increase
Solution Approach 1:
The patent substitutes heavy mechanical image processing hardware with a lightweight acoustic sensor array and simplified signal processing system. The acoustic sensors and their associated processing electronics weigh significantly less than camera systems combined with high-performance computing hardware, enabling real-time processing in UAVs with strict weight constraints.
3Measurement precision
If complex algorithms are used to determine object locations from sensor data, then location accuracy is improved, but computational burden increases beyond UAV capabilities
Solution Approach 1:
The patent replaces complex computational algorithms with direct geometric calculations based on time-difference-of-arrival measurements from the acoustic sensor array. This substitution uses simple mathematical relationships to compute object position from acoustic signal timing data, dramatically reducing computational energy consumption while maintaining accurate location determination suitable for UAV operations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The SGML system enables real-time guidance with significantly reduced computational requirements, making it feasible for deployment on UAVs where weight and power are concerns, while maintaining effective obstacle avoidance and navigation capabilities.
Implementation Method 1
The sensor array includes transmitters that transmit signals at intervals and receivers that collects return signals, which are transmitted signals reflected by objects.
Implementation Method 2
The time-of-arrival of each return pulse represent the time between the transmitting of a signal and receiving of a return pulse.
Data Source
AI summary
A system for generating a machine learning system to generate guidance information based on locations of objects is provided. The system accesses training data that includes training time-of-arrival (“TOA”) information of looks and guidance information for each look. The guidance information is based on a training collection of object locations. The TOA of a look represents, for each object location of a training collection of object locations, times between signals transmitted by transmitters and return signals received by receivers. The return signals represent signals reflected from an object at the object location. The system trains a machine learning system using the training data wherein the machine learning system inputs TOA information and outputs guidance information.