Autonomous Robot Navigation With Motion Primitives in Cluttered Spaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for autonomous navigation, such as robot mapping and path planning, face challenges with inaccuracies due to environmental and sensor noise, leading to inefficiencies in processing power and memory usage, and are unsuitable for complex environments with non-convex robot geometries and cluttered spaces.
Innovation Solution
The approach involves a two-phase system for determining motor commands that minimizes a cost function, using a library of precomputed motion primitives and a forward model to simulate trajectories, allowing for precise navigation with minimal clearance and collision avoidance in complex environments, decoupling model and motion primitive generation from planner development, and employing deterministic algorithms for scalable testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional mapping and path planning methods are used, then navigation can be implemented, but processing time and memory usage increase significantly
Solution Approach 1:
The system precomputes and stores motion primitives (pre-planned motion sequences) offline before actual navigation. During online operation, the robot selects from pre-computed primitives rather than planning from scratch, dramatically reducing real-time processing time while maintaining navigation capability.
Solution Approach 2:
The navigation problem is segmented into two phases: offline precomputation of motion primitives and online selection/execution. This segmentation allows computationally intensive tasks to be performed beforehand, freeing up processing resources during actual navigation.
2Ease of operation
If simple geometric primitives are used to approximate robot shape, then path planning becomes easier, but navigation precision deteriorates
Solution Approach 1:
Motion primitives are pre-computed incorporating the exact robot geometry and kinematics offline. This allows complex precise geometry to be handled in advance, while online operation uses simplified primitive selection that maintains precision without requiring complex real-time calculations.
Solution Approach 2:
The system uses occupancy grids as discrete representations of continuous space and robot footprints as discrete masks. These discrete copies enable efficient computational operations while preserving the essential geometric information needed for precise navigation.
3Productivity
If robots navigate close to obstacles for efficiency, then productivity increases, but collision risk increases
Solution Approach 1:
Motion primitives are pre-computed to inherently satisfy safety constraints and collision avoidance requirements. By embedding safety checks in the offline precomputation phase, the system ensures that selected primitives are guaranteed safe, enabling close obstacle navigation without compromising reliability.
Solution Approach 2:
The system evaluates motion primitives based on cost functions that incorporate safety metrics and collision probabilities. Primitives that violate safety constraints are penalized or rejected, providing feedback that ensures reliable navigation even when operating close to obstacles.
Data Source
AI summary
The safe operation and navigation of robots is an active research topic for many real-world applications, such as the automation of large industrial equipment. This technological field often requires heavy machines with arbitrary shapes to navigate very close to obstacles, a challenging and largely unsolved problem. To address this issue, a new planning architecture is developed that allows wheeled vehicles to navigate safely and without human supervision in cluttered environments. The inventive methods and systems disclosed herein belong to the Model Predictive Control (MPC) family of local planning algorithms. The technological features disclosed herein works in the space of two-dimensional (2D) occupancy grids and plans in motor command space using a black box forward model for state inference. Compared to the conventional methods and systems, the inventive methods and systems disclosed herein include several properties that make it scalable and applicable to a production environment. The inventive concepts disclosed herein are at least deterministic, computationally efficient, run in constant time and can be deployed in many common non-holonomic systems.