Robotic Object Detection Using CNNs for Noisy Multi-Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robotic object detection systems face challenges in accurately identifying and distinguishing multiple types of objects using scanning range sensors that provide semi-sparse spatial and infrared data, often struggling with sensor noise and requiring complex classification algorithms and fusion of outputs.
Innovation Solution
A system employing a convolutional neural network that processes combined spatial and infrared data from both 2D and 3D sensors, transforming sensor data into a multidimensional array format for object classification, enabling the detection and tracking of various objects, including humans and dynamic obstacles, and determining safe navigation paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex classification algorithms and data fusion methods are used to improve object detection accuracy, then detection precision improves, but device complexity increases
Solution Approach 1:
The patent replaces complex mechanical classification algorithms with a neural network system that automatically learns and classifies object types from sensor data. The neural network processes spatial and infrared data to identify multiple object types (humans, animals, inanimate objects) without requiring manual algorithm design for each object category, thereby reducing algorithmic complexity while maintaining high detection accuracy
Solution Approach 2:
The patent employs a universal neural network classifier that can identify multiple different object types using the same detection system. Rather than requiring separate classification algorithms for each object type, the neural network provides a multi-functional classification capability that handles humans, animals, inanimate objects, and moving objects uniformly, simplifying the overall system architecture
2Reliability
If multiple sensor types (2D and 3D) with combined spatial and infrared data are used, then object detection reliability improves, but device complexity increases
Solution Approach 1:
The patent merges 2D spatial sensor data and 3D infrared sensor data into a unified neural network processing pipeline. The neural network simultaneously processes both data types to detect and classify objects, combining the complementary information from each sensor type to improve detection reliability while avoiding the need for separate processing systems for each sensor
Solution Approach 2:
The neural network acts as an intermediary that receives and integrates data from multiple sensor types (2D spatial sensors and 3D infrared sensors). Rather than requiring complex external data fusion hardware or software, the neural network internally processes and combines the sensor inputs, simplifying the overall system architecture while maintaining high detection reliability
3Device complexity
If traditional object detection systems are used with scanning range sensors, then device simplicity is maintained, but detection precision deteriorates due to sensor noise and semi-sparse data
Solution Approach 1:
The patent replaces traditional mechanical object detection algorithms with a neural network-based system that is specifically designed to handle semi-sparse sensor data and sensor noise. The neural network learns robust features from the noisy scanning range sensor data, significantly improving object identification accuracy while maintaining relative system simplicity through the use of a single integrated detection pipeline
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 system effectively detects and tracks objects, predicts their trajectories, and determines safe navigation paths, overcoming previous limitations in object type and location identification, and maintaining high detection accuracy despite sensor noise, while reducing the complexity of detector stages and fusion of data types.
Implementation Method 1
a light detection and ranging (LIDAR) sensor using a convolutional neural network
Implementation Method 2
the 2D sensor further configured to provide infrared data related to one or more of a shape, a size, a type, a reflectivity, a location, and dynamic data regarding the object
Data Source
AI summary
A system includes a mobile robot, the robot comprising a sensor; and a server operably connected to the robot over a network, the robot being configured to detect an object by processing sensor data using a convolutional neural network. A pipeline for robotic object detection using a convolutional neural network includes: a system comprising a mobile robot, the robot comprising a sensor, the system further comprising a server operably connected to the robot over a network, the robot being configured to detect an object by processing sensor data using a pipeline, the pipeline comprising a convolutional neural network, the pipeline configured to perform a data collection step, the pipeline further configured to perform a data transformation step, the pipeline further configured to perform a convolutional neural network step, the pipeline further configured to perform a network output transformation step, the pipeline further configured to perform a results output step.