Vision-Based Follower Steering Using Dynamic Pursuit Pose
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous follower vehicle systems face challenges in efficiently and safely navigating behind a leader vehicle, particularly in scenarios where lane markings are inadequate or absent, and in maintaining a stable following trajectory while adapting to changes in the leader's path.
Innovation Solution
The system employs non-contact perception sensors like lidar, cameras, and radars to derive a trajectory for the follower vehicle based on Leader-Follower Relative Pose (LFRP) and Lane Relative Pose (LRP), allowing for adaptive steering and path adjustments, including extrapolation and interpolation of the leader's motion, to maintain a stable following distance and lane center alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If autonomous follower vehicle uses traditional lane marking detection methods, then lane following accuracy is improved, but system reliability deteriorates when lane markings are inadequate or absent
Solution Approach 1:
The patent introduces an intermediary reference frame transformation mechanism that mediates between sensor data and lane following control. By transforming sensor measurements into a leader-follower relative pose (LFRP) reference frame, the system can reliably follow the leader vehicle even when traditional lane marking detection fails, thus maintaining both accuracy and reliability under varying road conditions
Solution Approach 2:
The system employs a universal sensor fusion approach that combines multiple sensing modalities (lidar, cameras, radars) to perform both leader vehicle tracking and lane detection functions. This multi-functional sensing system ensures reliable operation whether lane markings are present or absent, resolving the contradiction between accuracy and reliability
2Device complexity
If autonomous follower vehicle maintains fixed following distance, then control simplicity is improved, but adaptability deteriorates when leader's path changes
Solution Approach 1:
The patent implements dynamic pursuit point calculation that adapts the follower vehicle's target position based on the leader's current pose and trajectory. Instead of maintaining a fixed following distance, the system dynamically adjusts the pursuit point in the LFRP reference frame, enabling smooth adaptation to path changes while maintaining relatively simple control logic through parametric adjustments
3Device complexity
If autonomous follower vehicle uses only leader perception inputs, then system complexity is reduced, but measurement precision deteriorates in scenarios with inadequate lane markings
Solution Approach 1:
The patent merges leader perception inputs with follower's own sensor data in a unified LFRP reference frame. By combining relative pose information from the leader with the follower's direct sensor measurements of the environment, the system achieves accurate position estimation without requiring complex multi-vehicle communication infrastructure, thus improving precision while maintaining reasonable system complexity
Data Source
AI summary
Techniques for operating an autonomous follower vehicle that is following a leader vehicle. A desired path to be traversed by the follower may be determined from a Leader Follower Relative Pose (LFRP) derived from sensor data. A pursuit pose is derived along the desired path from the present leader pose, such as either interpolating backward or forward. As a result, the pursuit distance no longer needs to be the same as the distance derived solely from the LFRP. This permits steering controls (lateral position) to be freed from requirements to satisfy safety constraints that might otherwise be imposed by other considerations (such as longitudinal spacing between vehicles).