Dynamic Bounding Boxes for Vehicle Trajectory Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining vehicle trajectories and object avoidance in dynamic environments suffer from inaccuracies due to errors in sensor data and system response variability, leading to an increased probability of contact with obstacles.
Innovation Solution
A dynamic occupancy system using a polyhedral marching processor to determine dynamic bounding boxes based on vehicle trajectory curvature and acceleration, employing neural networks trained on real-world data to predict occupancy probabilities and adjust trajectories in real-time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If static bounding boxes are used for trajectory prediction, then the system is simple to implement, but the accuracy of predicting vehicle locations deteriorates due to sensor errors and system variability
Solution Approach 1:
The patent applies dynamics by transitioning from static bounding boxes to dynamic occupancy grids that adapt in real-time. The system continuously updates the occupancy probability of each grid cell based on the vehicle's current state, trajectory, and sensor data, allowing the prediction model to respond to changing conditions and improve location accuracy without excessive complexity
Solution Approach 2:
The system changes parameters by using multiple possible trajectories instead of a single fixed path. Each trajectory is evaluated with probability weights, and the occupancy grid is updated by combining predictions from multiple trajectories. This parameter change approach allows the system to account for sensor errors and system variability while maintaining manageable computational complexity
2Measurement precision
If dynamic occupancy grids with multiple trajectories are used, then the accuracy of predicting vehicle locations is improved, but the device complexity increases due to real-time processing requirements
Solution Approach 1:
The patent applies segmentation by dividing the prediction space into discrete grid cells and evaluating multiple trajectories separately. Each trajectory is processed independently to determine occupancy probabilities for grid cells, allowing parallel computation and reducing the complexity of real-time processing while maintaining high prediction accuracy
Solution Approach 2:
The system uses partial action by evaluating only the most probable trajectories rather than all possible paths. By focusing computational resources on the most likely trajectories and using probability thresholds to prune less likely paths, the system achieves high accuracy without the excessive complexity of exhaustive trajectory evaluation
3Reliability
If broad static bounding boxes are used to account for errors, then the confidence level of contact avoidance is improved, but the precision of vehicle location prediction deteriorates
Solution Approach 1:
The patent applies local quality by assigning different occupancy probabilities to different grid cells based on their specific characteristics and distance from the predicted trajectory. Instead of using a uniform broad bounding box, the system calculates localized probability distributions that reflect the actual uncertainty at each location, maintaining high confidence in contact avoidance while preserving precision in vehicle location prediction
Data Source
AI summary
A computer that includes a processor and a memory can determine a final trajectory for a vehicle by determining a candidate trajectory of a first object based on a detected second object. The candidate trajectory can be input to a reachable polyhedral marching processor to determine dynamic occupancy polyhedrals based on a shape of the candidate trajectory. A reachable tube can be determined based on combining the dynamic occupancy polyhedrals and the final trajectory can be determined based on the reachable tube avoiding the detected second object.