Hierarchical Trajectory Proposal for Faster Diverse Autonomous Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous navigation systems face inefficiencies in predicting diverse trajectories due to increased computation time and efforts, especially when dealing with multiple trajectory proposals and overcoming GPS errors.
Innovation Solution
A hierarchical 2-stage Neural Network architecture is employed, comprising a Trajectory Proposal Network (TPNet) and a Trajectory Sampling Network (TSNet), which generates diverse traversable regions and predicts fine-grained trajectories using Convolutional Neural Networks (CNN) and Long Short Term Memory (LSTM) networks, respectively, to efficiently propose multiple trajectory options.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional trajectory planning techniques based on hypothesizing multiple goal locations are used for diverse trajectory prediction, then multiple trajectory options can be generated, but computation time and computational efforts increase significantly
Solution Approach 1:
The trajectory planning problem is segmented into two independent stages: (1) traversable region proposal stage that identifies candidate regions using sensor data and occupancy maps, and (2) trajectory sampling stage that generates specific trajectories within those regions. This segmentation allows each stage to be optimized independently and reduces the overall computational burden compared to generating all trajectories from scratch.
Solution Approach 2:
The system performs preliminary action by first identifying traversable regions before generating trajectories. The Trajectory Proposal Network (TPNet) pre-processes the environment by marking safe traversable regions, which then constrains the trajectory sampling process. This preliminary region identification reduces the search space for trajectory generation, significantly cutting computation time while maintaining diverse trajectory options.
2Ease of operation
If neural networks predict perceptual input to continuous space control actions, then navigation control can be achieved, but efficiency for predicting multiple trajectory proposals is poor
Solution Approach 1:
The neural network architecture is segmented into two specialized networks: TPNet for traversable region prediction and TSNet for trajectory sampling. TPNet processes perceptual input to identify safe regions, while TSNet efficiently samples trajectories within those regions. This segmentation allows each network to be optimized for its specific function, improving overall trajectory prediction efficiency compared to a single network attempting to predict continuous control actions for multiple trajectories.
Solution Approach 2:
The traversable region proposals act as an intermediary between perceptual input and final trajectory predictions. Instead of directly mapping sensor data to continuous control actions for multiple trajectories, the system first generates discrete traversable region proposals that guide subsequent trajectory sampling. This intermediary representation simplifies the learning task for the neural networks and improves computational efficiency.
Data Source
AI summary
The disclosure herein generally relates to the field of autonomous navigation, and, more particularly, to a diverse trajectory proposal for autonomous navigation. The embodiment discloses a hierarchical network based diverse trajectory proposal for autonomous navigation. The hierarchical 2-stage neural network architecture maps the perceived surroundings to diverse trajectories in the form of trajectory waypoints, that an autonomous navigation system can choose to navigate/traverse. The first stage of the disclosed hierarchical 2-stage Neural Network architecture is a Trajectory Proposal Network which generates a set of diverse traversable regions in an environment which can be occupied by the autonomous navigation system in the future. The second stage is a Trajectory Sampling network which predicts a fine-grained trajectory/trajectory waypoint over the diverse traversable regions proposed by Trajectory Proposal Network.