Trajectory Validation Using Occupancy Maps for Collision Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle collision avoidance systems face challenges in accurately detecting and validating trajectories to prevent collisions, often leading to errors and increased computational demands.
Innovation Solution
A collision avoidance system utilizing a secondary perception component with machine-learning models to validate or reject trajectories based on sensor data, incorporating ensemble voting techniques and threshold distances to ensure safe vehicle operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors and systems are used for detecting and tracking events surrounding the autonomous vehicle, then collision detection accuracy is improved, but computational bandwidth and processing complexity increase
Solution Approach 1:
The system segments the detection task by dividing the environment into discrete grid cells and processing sensor data from multiple sensors independently for each cell. This allows parallel processing of LIDAR, radar, and camera data without requiring centralized computation of all sensor inputs simultaneously, reducing computational bandwidth while maintaining detection accuracy through comprehensive multi-sensor coverage.
2Measurement precision
If traditional sensor systems are used for autonomous navigation, then basic collision detection is achieved, but accuracy in detecting occupied portions of the environment is insufficient
Solution Approach 1:
The system introduces an intermediary validation component that acts as a mediator between the primary perception system's trajectory suggestions and the actual vehicle control. This validation component independently assesses occupancy probability using the same sensor data and compares it against the suggested trajectory, providing an additional layer of verification that improves both detection accuracy and trajectory validation reliability without requiring entirely new sensor systems.
3Measurement precision
If a secondary perception component with validation is added, then trajectory validation accuracy is improved, but system complexity increases
Solution Approach 1:
The validation component is designed to perform multiple functions using the same sensor data infrastructure already present in the autonomous vehicle system. It simultaneously validates trajectories, detects occupancy, and provides backup detection capabilities, thereby improving trajectory validation accuracy without proportionally increasing system complexity through dedicated specialized components for each function.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A collision avoidance system may validate, reject, or replace (416, 418, 420) a trajectory generated to control a vehicle. The collision avoidance system may comprise a secondary perception component that may comprise one or more machine learned models, each of which may be trained to output one or more occupancy maps (406(1) -406(N)) based at least in part on sensor data of different types (402(1)-402(N)). The occupancy maps may include a prediction of whether at least a portion of an environment is occupied at a future time by any one of multiple object types. Occupancy maps associated with a same time may be aggregated (408) into a data structure that may be used to validate, reject, or replace the trajectory (416, 418, 420).