A Novel Occupancy Mapping Framework for Risk-Aware Path Planning in Unstructured Environments.

Area of Science:

• Robotics
• Artificial Intelligence
• Autonomous Systems

Background:

• Autonomous robot navigation requires damage prevention, often relying on probabilistic obstacle detection.
• Bayesian occupancy maps are common but limited for continuous path risk assessment due to their discrete nature.
• Unstructured environments pose challenges for traditional obstacle detection and mapping methods.

Purpose of the Study:

• To introduce a novel metric map, the Lambda Field, for enhanced risk assessment in robot navigation.
• To develop a framework for computing generic risk expectations over continuous paths.
• To demonstrate the Lambda Field's utility in physical-based path planning and navigating unstructured environments.

Main Methods:

• Proposed a method for computing the Lambda Field map.
• Developed a technique to calculate the expectation of a generic risk over a path.
• Utilized a use case defining risk as expected collision force for path planning.

Main Results:

• The Lambda Field enables physical-based path planning, integrating risk assessment directly into trajectory generation.
• Demonstrated effective navigation through unstructured environments, such as tall grass, by quantifying and managing risk.
• Showcased the framework's capability to handle scenarios where traditional methods fail.

Conclusions:

• The Lambda Field offers a superior approach to risk-aware navigation for autonomous robots compared to discrete maps.
• This novel mapping technique facilitates safer and more efficient navigation in complex and unstructured terrains.
• The Lambda Field provides a foundation for advanced physical-based risk assessment in robotics.