Modular Autonomous Vehicle Architecture for Safe Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated vehicle systems rely on human intervention for various operations and lack the capability to independently determine safe driving corridors and trajectories, leading to potential collisions and inefficiencies.
Innovation Solution
An autonomous vehicle system that utilizes an Automated Driving Controller to determine drivable corridors and trajectories, with a Perception Controller generating stochastic and kinematic predictions of free space, and a Vehicle Controller sending steering, throttle, and braking commands to actuator controllers to avoid collisions and ensure ride comfort, while also integrating sensor data translation and operator availability monitoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automated systems assist driver functions (lane assist, pre-collision breaking, rear cross-track detection), then driver safety is improved, but the system still relies on driver intervention and cannot achieve full autonomy
Solution Approach 1:
The patent segments the autonomous driving system into distinct functional modules: perception controller for sensor data processing, automated driving controller for corridor and trajectory determination, and vehicle controller for actuator control. This modular segmentation enables full automation while managing system complexity through clear separation of concerns and independent module development.
2Reliability
If the system determines multiple planned driving corridors and trajectories, then collision avoidance is improved, but computational processing time increases
Solution Approach 1:
The perception controller performs preliminary stochastic prediction of free space and kinematic prediction of object locations in advance, generating multiple potential corridors and trajectories before the vehicle needs to execute them. This preliminary action enables comprehensive collision avoidance analysis while providing results in time for real-time decision-making.
Solution Approach 2:
The patent replaces traditional mechanical reaction-based collision avoidance with computational stochastic prediction and kinematic modeling. Instead of relying on driver reaction time or simple sensor-triggered braking, the system uses probabilistic algorithms to predict multiple future states and select optimal trajectories, achieving both high reliability and efficient processing.
3Productivity
If sensor data from multiple subsystems is processed independently, then data processing is simpler, but communication bandwidth is excessive and system efficiency is reduced
Solution Approach 1:
The patent merges sensor data processing by having the perception controller fuse measurements from multiple sensors into a single integrated estimate of the environment. This consolidation reduces communication bandwidth requirements while maintaining comprehensive situational awareness, as downstream controllers receive processed, integrated data rather than raw data from each sensor.
Solution Approach 2:
The perception controller acts as an intermediary between raw sensor subsystems and the automated driving controller. It translates and filters sensor data, providing a unified environmental model to downstream controllers. This intermediary role reduces communication overhead while enabling efficient use of processed information for trajectory planning and control.
Data Source
AI summary
An architecture for an autonomous vehicle uses a top-down approach to enable fully automated driving. The architecture is modular and compatible with hardware from different manufacturers. Each modular component can be tailored for individual cars, which have different vehicle control subsystems and different sensor subsystems.