Autonomous Vehicle Path Sampling With Spatial Envelope Filtering

The patent segments the trajectory evaluation process into two distinct phases: (1) generating a reduced set of candidate trajectories using spatial envelope and vehicle configuration filtering, and (2) evaluating only these filtered candidates. This segmentation allows the system to maintain high reliability by thoroughly evaluating viable options while improving productivity by eliminating the need to evaluate infeasible trajectories.

The patent applies preliminary action by performing filtering operations before the main trajectory evaluation. The spatial envelope generation and vehicle configuration data filtering are executed in advance to identify and eliminate infeasible trajectories before they enter the costly evaluation phase. This preliminary filtering ensures that only promising candidates are subjected to detailed assessment, thereby reducing computational waste.

The patent applies local quality by making the trajectory generation process adaptive to local vehicle characteristics and spatial constraints. The spatial envelope and vehicle configuration data are used to locally filter trajectories based on specific vehicle capabilities (e.g., turning radius, acceleration limits) and environmental constraints at each segment of the path, rather than applying uniform filtering across all trajectories.

The patent changes parameters by dynamically adjusting the set of evaluated trajectories based on vehicle configuration parameters and spatial envelope characteristics. The filtering process modifies which trajectories are considered viable by changing the effective parameter space from all possible trajectories to only those satisfying vehicle-specific and environment-specific constraints, thereby reducing computational cost while maintaining adaptability.

The patent introduces intermediary structures (spatial envelope and vehicle configuration data representations) that mediate between the raw trajectory generation and the evaluation process. These intermediaries serve as filtering layers that translate complex vehicle constraints and spatial relationships into simplified criteria for trajectory selection, reducing the burden on the evaluation process while maintaining accuracy.

The patent uses copying by creating simplified representations of vehicle configuration data and spatial envelope characteristics that can be efficiently processed during filtering. Instead of working with full-precision, complex models during the filtering stage, the system uses copied or approximated data structures that capture essential constraints, enabling faster filtering while preserving the accuracy needed for reliable trajectory selection.