AI Robot Navigation with Socially Compliant Crowd Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous navigation systems struggle with safe and socially compliant navigation in dynamic human-centric environments, particularly in crowded spaces, due to challenges in differentiating between humans and obstacles, leading to unpredictable behaviors and inefficient path planning.
Innovation Solution
An AI-based system utilizing sensors, AI models, and ML models for object tracking, probabilistic estimation, and socially compliant behavior to generate convex hulls, cost maps, and navigation paths that adapt to dynamic environments, ensuring safe and efficient robotic navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional collision avoidance technologies are used to avoid inanimate obstacles, then obstacle avoidance capability is improved, but the ability to differentiate between humans and other obstacles deteriorates, resulting in suboptimal and intrusive robot behaviors
Solution Approach 1:
The system applies different quality standards to different types of obstacles by implementing specialized detection models. Human detection uses specific computer vision algorithms that recognize human characteristics, while general obstacle detection uses different criteria. This allows the robot to differentiate between humans and other obstacles, applying appropriate navigation behaviors for each type.
Solution Approach 2:
The patent introduces an intermediary layer between basic obstacle detection and navigation decision-making. This intermediary consists of specialized detection models (human detection, group detection, corridor detection) that process sensor data and provide classified obstacle information to the path planning system, enabling nuanced differentiation and appropriate response.
2Device complexity
If static path planning models are used for navigation, then path planning simplicity is improved, but adaptability to dynamic changes in environment deteriorates, leading to inefficient routes in areas with rapidly changing crowd density
Solution Approach 1:
The system transitions from static to dynamic path planning by continuously updating the cost map based on real-time sensor data and detected environmental features. The path planning model dynamically adjusts navigation costs according to detected corridors, groups of people, and other obstacles, allowing the robot to adapt its route as environmental conditions change while maintaining a structured planning framework.
Solution Approach 2:
The patent implements feedback mechanisms where the robot continuously senses the environment, detects features (corridors, groups, obstacles), updates the cost map accordingly, and replans the path. This closed-loop feedback system enables continuous adaptation to dynamic changes in crowd density and environmental conditions while maintaining efficient navigation.
3Measurement precision
If high-precision sensors are used for accurate positioning and navigation, then positioning accuracy is improved, but system cost and complexity deteriorates
Solution Approach 1:
The system achieves accurate positioning and navigation using multi-functional sensors already present for other purposes. The same sensors used for obstacle detection and environmental perception are also utilized for positioning by detecting corridors, landmarks, and spatial features. This eliminates the need for separate high-precision positioning sensors while maintaining navigation accuracy.
Solution Approach 2:
The robot uses its own sensing capabilities and environmental features (corridors, walls, landmarks) to achieve accurate positioning without requiring external infrastructure or specialized sensors. The system processes sensor data to detect environmental structures and uses these for self-localization and navigation, making the positioning function self-sufficient.
4Reliability
If the robot halts suddenly or exhibits unpredictable behaviors when encountering obstacles, then collision safety is improved, but user experience and acceptance deteriorate
Solution Approach 1:
The system performs preliminary detection and classification of obstacles before collision risk becomes critical. By detecting corridors, groups of people, and individual obstacles in advance, the robot can plan appropriate avoidance maneuvers ahead of time, maintaining smooth and predictable motion while ensuring safety. This prevents sudden halts by preparing avoidance actions in advance.
Solution Approach 2:
The patent implements dynamic navigation behavior that adapts to the type of obstacle encountered. Instead of uniform sudden halts, the system dynamically selects appropriate responses: navigating along detected corridors, avoiding groups with appropriate clearance, or stopping for individual obstacles. This dynamic behavior maintains safety while providing smooth, predictable, and socially acceptable navigation.
Data Source
AI summary
An artificial intelligence (AI)-based system and method for autonomous navigation of robotic devices in dynamic human-centric environments are disclosed. The AI-based system comprises an object tracking subsystem, a probabilistic estimation subsystem, a socially compliant behavior subsystem, a constrained space navigation subsystem, a commands processing subsystem, a virtual cost-map layer subsystem, and a path-planning subsystem. The AI-based system obtains sensor data using sensors. The AI-based system employs artificial intelligence (AI) models and machine learning (ML) models for computing probabilistic position data and generating convex hulls. The AI-based system generates high-cost zones for identifying boundaries associated with groups to plan navigation paths. The AI-based system generates waypoints for the robotic devices based on detecting constrained spaces in the dynamic human-centric environments by analyzing the sensor data. The AI-based system extracts navigational insights based on processing natural language commands and adaptively selects the navigation paths for autonomous navigation of the robotic devices.