Automated Vehicle Trajectory Planning With Quality-Based Braking

The patent applies local quality by differentiating the treatment of detected objects based on their detection quality. High-quality detections trigger standard avoidance behavior with appropriate braking, while low-quality detections are treated differently to avoid excessive braking. This is implemented through quality-based cost assignment where the same object type receives different trajectory costs depending on detection quality metrics.

The patent changes the parameter of trajectory cost assignment based on detection quality. Instead of uniform cost assignment for all detected objects, the system dynamically adjusts trajectory costs according to detection quality parameters. This allows the trajectory optimization to account for uncertainty in detection reliability, preventing excessive braking responses to low-quality detections while maintaining appropriate responses to high-quality detections.

The patent segments the trajectory evaluation process by dividing trajectory candidates into filtered and unfiltered categories based on detection quality. This segmentation allows the system to apply different evaluation strategies: low-quality detections are filtered out or given minimal weight, while high-quality detections receive full evaluation. This reduces the effective search space and computational burden compared to evaluating all trajectory candidates with all detected objects equally.

The patent applies partial action by selectively applying full trajectory evaluation only to high-quality detections, while using simplified or filtered evaluation for low-quality detections. This partial evaluation approach reduces computational complexity by avoiding exhaustive evaluation of all trajectory candidates against all detected objects, particularly those with low detection quality that are less likely to represent real hazards.

The patent performs preliminary filtering of detected objects based on detection quality before assigning trajectory costs. By pre-identifying and filtering out low-quality detections, the system prevents these unreliable detections from triggering unnecessary braking interventions. This preliminary action maintains safety by focusing on high-quality detections while eliminating the source of excessive braking that would compromise comfort.

The patent converts the potentially harmful effect of low-quality detections (which would cause excessive braking and discomfort) into a benefit by using detection quality as a filtering mechanism. The presence of low-quality detections is transformed from a source of problematic braking into an opportunity to demonstrate robust, comfortable driving behavior by appropriately downweighting or filtering these unreliable detections while maintaining response to genuine hazards.