Neuromorphic LiDAR Inference for Lower ADAS Data Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in efficiently processing and analyzing lidar sensor data, leading to increased computational workload and data traffic, which can impact the safety and performance of autonomous vehicles.
Innovation Solution
Integration of a neuromorphic memory device with a Spiking Neural Network (SNN) or Deep Neural Network (DNN) in lidar sensors to perform local analysis of lidar images, generating inference results that are then communicated to the ADAS, reducing the need for raw data transfer and enhancing processing capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If raw lidar data is transmitted to the ADAS for processing, then the computational system can analyze the data, but the data traffic and computational workload increase
Solution Approach 1:
The patent segments the data processing function by dividing it between the lidar sensor (edge device) and the ADAS (central system). The lidar sensor performs local preprocessing and generates inference results, while the ADAS receives only the processed results rather than raw data, thereby segmenting the computational workload and reducing data traffic.
Solution Approach 2:
The patent adds a new dimension of processing by integrating neuromorphic memory devices and SNN/DNN engines directly into the lidar sensor architecture. This transforms the system from a traditional centralized processing model to a distributed edge-processing model, enabling local inference and reducing the dimension of data that needs to be transmitted.
2Productivity
If neural network processing is performed in the lidar sensor, then data traffic is reduced, but the device complexity increases
Solution Approach 1:
The patent merges multiple functions into the lidar sensor by integrating the neuromorphic memory device, SNN/DNN engine, and inference engine directly into the sensor architecture. This consolidation enables the sensor to perform both data acquisition and local neural network processing, improving productivity while managing complexity through functional integration.
Solution Approach 2:
The lidar sensor is designed with multi-functionality by incorporating neuromorphic memory devices that can store and process data locally, SNN/DNN engines for various types of neural network operations, and an inference engine for generating results. This universal design allows the sensor to handle multiple processing tasks within a single device, improving efficiency without requiring separate dedicated components for each function.
Data Source
AI summary
Systems, methods and apparatuses of lidar sensors of autonomous vehicles. A lidar sensor can include: a memory configured to store a lidar image and an Artificial Neural Network (ANN); an inference engine configured to use the (ANN) to analyze the lidar image and generate inference results; and a communication interface coupled to a computer system of a vehicle to implement an advanced driver assistance system to operate the controls according to the inference results and a sensor data stream generated by sensors configured on the vehicle.