Vehicle Trajectory Generation Without HD Maps for Autonomous Maneuvers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches for autonomous vehicles to perform maneuvers like lane changes, lane splits, and turns rely heavily on high-definition (HD) maps, which are computationally expensive and require large amounts of processing power, energy, and bandwidth. Additionally, these systems are limited to areas with available and accurate HD maps, restricting their universal implementation.
Innovation Solution
The use of machine learning models that compute vehicle control data based on sensor data from cameras, RADAR, and LIDAR sensors, along with low-resolution map data and vehicle status information, allows autonomous vehicles to perform maneuvers without relying on HD maps. This approach enables navigation in any location, regardless of HD map availability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If high-definition (HD) maps are used for autonomous vehicle maneuvers, then navigation accuracy and reliability are improved, but computational cost and energy consumption increase significantly
Solution Approach 1:
The patent extracts and removes the dependency on HD maps from the autonomous vehicle system. Instead of relying on pre-recorded HD map data for localization and maneuver planning, the system uses sensor data from cameras, RADAR, and LIDAR to directly perceive the environment and generate control commands, eliminating the computationally expensive HD map processing while maintaining navigation accuracy
Solution Approach 2:
The patent substitutes the mechanical HD map-based localization system with a sensor-based perception system. The machine learning model replaces the traditional HD map matching and localization algorithms, using real-time sensor inputs to understand the vehicle's position and surroundings, thereby reducing computational burden while preserving navigation reliability
2Measurement precision
If high-definition (HD) maps are used for autonomous vehicle maneuvers, then localization precision is improved, but system complexity and data requirements increase
Solution Approach 1:
The patent removes the HD map data structure and associated processing infrastructure from the system. By extracting this dependency, the system achieves localization precision through direct sensor perception and machine learning-based environment understanding, significantly reducing system complexity and data management requirements
Solution Approach 2:
Instead of using pre-recorded HD map copies of the environment, the system creates real-time digital representations of the surroundings through sensor data processing. The machine learning model generates localized environmental models on-demand, eliminating the need to store and process extensive HD map datasets while maintaining localization accuracy
3Reliability
If HD maps are required for all locations, then safety and reliability are improved, but adaptability to unmapped locations deteriorates
Solution Approach 1:
The patent implements a universal sensor-based perception system that functions in all locations regardless of HD map availability. The machine learning model processes sensor data from cameras, RADAR, and LIDAR to provide safe and reliable autonomous operation in both mapped and unmapped environments, making the system adaptable to any location without requiring pre-existing HD maps
Solution Approach 2:
The system enables autonomous vehicles to self-localize and navigate in unmapped locations by using their own sensors to perceive and understand the environment. The machine learning model allows the vehicle to independently generate navigation decisions based on real-time sensor inputs, eliminating dependency on external HD map infrastructure and providing safety and adaptability in diverse locations
Data Source
AI summary
In various examples, a trigger signal may be received that is indicative of a vehicle maneuver to be performed by a vehicle. A recommended vehicle trajectory for the vehicle maneuver may be determined in response to the trigger signal being received. To determine the recommended vehicle trajectory, sensor data may be received that represents a field of view of at least one sensor of the vehicle. A value of a control input and the sensor data may then be applied to a machine learning model(s) and the machine learning model(s) may compute output data that includes vehicle control data that represents the recommended vehicle trajectory for the vehicle through at least a portion of the vehicle maneuver. The vehicle control data may then be sent to a control component of the vehicle to cause the vehicle to be controlled according to the vehicle control data.