Efficiently Combining Data from Different Sensors

This patent describes a system for combining information from many different types of sensors more efficiently by carefully selecting and processing the data, reducing the computer power needed.

What it covers

The system outlined in this patent collects data from multiple sensors, then carefully organizes and filters it, a process called curation. It links related pieces of data and combines them, which is known as fusion. A key part of this process is calculating 'conditional entropy' to decide which data is most important, which helps reduce the amount of computing power and time needed. The system then uses this combined data to make educated guesses or decisions (inference) and checks if the combined data is reliable enough by comparing it against a set mathematical standard (validation). For example, a robot could use this to combine camera images, lidar distances, and microphone sounds to understand its surroundings without slowing down.

What it doesn't cover

• —Does not cover sensor data fusion systems that do not curate or filter the raw data before combining it.
• —Does not cover methods of data fusion that do not calculate conditional entropy to reduce computational load.
• —Does not cover systems that combine data from only a single type of sensor.
• —Does not cover systems that fuse data without validating the output against a mathematical threshold.
• —Does not cover data fusion techniques that increase, rather than reduce, computational demand and processing time.

The clever bit

The clever part is reducing the computational effort required for data fusion. It achieves this by curating the data and calculating conditional entropy, which helps the system focus only on the most relevant information from multiple sensors, making the process much more efficient.

Why it matters

Combining data from various sensors is critical for technologies like self-driving cars, smart homes, and advanced robotics. This patent focuses on making that process faster and less demanding on computer resources. By reducing computational demands, it allows for more complex real-time decision-making in devices with limited power or processing ability, which can lead to more responsive and reliable systems.

Real-world examples

• 1.Autonomous vehicle navigation systems
• 2.Robotics for industrial automation
• 3.Smart home environmental monitoring
• 4.Wearable health tracking devices
• 5.Industrial IoT sensor networks