LIDAR Traversability Mapping for Sidewalk Robot Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicle navigation on sidewalks is challenging due to the absence of lane markers and the presence of dynamic and static obstacles, requiring the ability to identify safe traversable areas in unstructured environments without relying on pre-built maps or GPS.
Innovation Solution
A sidewalk robot equipped with LIDAR sensors and machine learning algorithms that project laser light to create a three-dimensional layout of the environment, identify safe traversable areas, and generate masks to avoid obstacles, using real-time data to navigate autonomously and interact with dynamic objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional navigation methods (lane markers, GPS, pre-built maps) are used, then navigation is simple and reliable, but they cannot be applied in unstructured environments like sidewalks
Solution Approach 1:
The patent replaces traditional mechanical navigation systems (GPS, pre-built maps, lane markers) with a vision-based deep learning system. The sidewalk robot uses LIDAR sensors to capture environmental data, which is then processed by convolutional neural networks to generate traversability masks, enabling navigation in unstructured environments without relying on infrastructure-based guidance systems.
2Adaptability or versatility
If deep convolutional neural networks are used to identify safe traversable areas, then the robot can navigate unstructured environments, but the computational complexity and processing time increase
Solution Approach 1:
The patent employs pre-trained deep convolutional neural network models that have been trained offline on large datasets. The training process is performed beforehand, and the trained models are deployed for real-time inference. This allows the robot to perform complex traversability assessment during operation without the computational burden of training, as the models have already learned the necessary patterns from extensive pre-training.
3Reliability
If LIDAR data is processed in real-time to generate traversability masks, then the robot can avoid obstacles dynamically, but the processing speed and energy consumption increase
Solution Approach 1:
The patent extracts and processes only the essential features from LIDAR data that are critical for traversability assessment. The deep convolutional neural network is designed to focus on key geometric and structural features of the environment rather than processing all raw LIDAR points uniformly. This selective feature extraction reduces computational load and energy consumption while maintaining the reliability of obstacle avoidance capabilities.
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
Enables the sidewalk robot to safely traverse sidewalks by predicting and avoiding static and dynamic obstacles, maintaining a safe path without relying on pre-built maps or GPS, improving navigation in complex, unstructured environments.
Implementation Method 1
projecting a laser light from a LIDAR sensor, wherein the projection is directed toward an environment occupied by the sidewalk robot; receiving one or more reflections of the laser light as LIDAR data
Data Source
AI summary
Techniques described in this application are directed to determining safe path navigation of an unmanned vehicle, including a sidewalk robot, using LIDAR sensors and/or other data.