Dynamic Window Approach Velocity Selection for Robot Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot navigation methods are inefficient in simultaneously avoiding both moving and stationary obstacles in dynamic warehouse environments, leading to suboptimal progress towards target locations within the allotted cycle time.
Innovation Solution
A method for robot navigation that involves receiving a goal pose, determining a goal path, generating candidate velocities, evaluating these velocities using objective functions that consider obstacle avoidance and progress along the path, and selecting a preferred velocity to ensure safe and efficient movement, incorporating both dynamic window approach (DWA) and optimal reciprocal collision avoidance (ORCA) techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional obstacle avoidance methods are used, then collision avoidance is achieved, but progress towards target location is suboptimal within cycle time
Solution Approach 1:
The patent implements dynamic window approach (DWA) to create a time-varying velocity space that adapts to changing environmental conditions and robot states. The velocity window is dynamically adjusted based on robot velocity, acceleration capabilities, and time horizon, allowing the robot to optimize its trajectory in real-time while maintaining collision avoidance. This dynamic adaptation enables faster convergence to optimal velocities that maximize progress toward the target within each control cycle.
Solution Approach 2:
The patent transforms the navigation problem by changing the parameter space from direct position control to velocity space optimization. By formulating the objective function in terms of velocity parameters (linear and angular velocities) and optimizing within a dynamically defined velocity window, the system can efficiently evaluate multiple candidate velocities and select the one that maximizes progress while ensuring safety. This parameter transformation enables computationally efficient optimization within the fixed cycle time constraint.
2Reliability
If comprehensive obstacle avoidance is implemented, then collision safety is improved, but computational complexity increases
Solution Approach 1:
The patent segments the complex navigation problem into distinct computational components: (1) defining a velocity window based on robot dynamics and time horizon, (2) generating candidate velocities within the window, (3) evaluating each candidate using an objective function that incorporates collision avoidance constraints, and (4) selecting the optimal velocity. This segmentation allows each sub-problem to be solved efficiently with specialized algorithms, reducing overall computational complexity while maintaining comprehensive safety checks.
Solution Approach 2:
The patent evaluates a finite set of discrete candidate velocities sampled within the dynamic window rather than exhaustively searching the entire velocity space. By selecting a sufficient but not excessive number of candidate points, the system achieves practical collision avoidance guarantees while keeping computational requirements tractable for real-time operation. The dynamic window itself is designed to be partial (limited by acceleration constraints and time horizon) yet sufficient for safety-critical decisions.
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
A method and system for navigation of a robot along a goal path and avoiding obstacles. The method includes receiving goal pose for one or more robots and determining a goal path for a first robot while avoiding moving and fixed obstacles of a received obstacle map. A first objective function is evaluated to select a preferred velocity from a generated set of candidate velocities, the selecting based on one or more weighted cost functions. A set of velocity obstacles created based on the poses of the one or more robots and the preferred velocity is used in evaluating a second objective function to determine the motion of the robot in the next time cycle. Creating the set of velocity objects includes converting the preferred velocity from a non-holonomic to a holonomic velocity.