Autonomous Vehicle Control via Predictability-Based iMPC Activation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles operating in unpredictable environments face challenges in implementing Interactive Model Predictive Control (iMPC) due to the need for predictable conditions, which limits their ability to optimize fuel economy, drivability, and performance when encountering unpredictable objects or behaviors from other vehicles.
Innovation Solution
The system determines the predictability of the environment by assessing factors such as the presence of pedestrians, animals, and other vehicles' behaviors, and only implements iMPC when the environment is sufficiently predictable, allowing for optimal control of propulsion, steering, and braking by exchanging information between vehicles to achieve collective control goals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If iMPC is implemented in autonomous vehicles operating in unpredictable environments, then fuel economy, drivability, and performance can be optimized, but the control system fails due to violation of predictability assumptions
Solution Approach 1:
The patent applies dynamics by making the iMPC system configurable and adaptive based on environment predictability. The control system transitions between operational modes (fully autonomous iMPC, semi-autonomous, or manual override) depending on real-time assessment of environmental predictability. This allows the system to optimize performance when conditions permit while maintaining reliability by disengaging predictive control when unpredictability is detected, thus resolving the contradiction between optimization and reliability.
2Ease of operation
If iMPC is activated in all conditions, then optimal control function is provided, but safety is compromised when unpredictable objects are present
Solution Approach 1:
The patent implements feedback by continuously monitoring environmental conditions and object predictability to determine whether iMPC should remain active. The system assesses the predictability of objects in the vehicle's path and adjusts the control mode accordingly. When unpredictable objects are detected, the system provides feedback to disengage iMPC and switch to semi-autonomous or manual control, thus eliminating safety risks while preserving optimal control function when conditions are favorable.
3Reliability
If environment predictability is assessed and iMPC is conditionally implemented, then safety and efficiency are enhanced, but system complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the autonomous control system into distinct operational modes: fully autonomous iMPC mode, semi-autonomous mode, and manual override mode. Each mode has clearly defined characteristics and control strategies. The system segments the decision-making process into discrete states based on environment predictability assessment, making the complex system more manageable and easier to implement while maintaining safety enhancements through conditional operation.
Data Source
AI summary
A system includes a computing device programmed to receive a set of goals for a vehicle and identify a travel area for the vehicle for a time period. The computing device receives data indicating predictability of driving conditions of the travel area and determines that the predictability is sufficient to control the vehicle according to model predictive control. Controlling the vehicle includes determining instructions to control actuators related to the steering, propulsion and braking of the vehicle to minimize a cost function. The instructions are implemented for a first time slot. The time period is updated to remove the first time slot at the beginning and include an additional time slot at the end of the predetermined time period. The computing device determines an updated control solution, and implements the updated control solution for a second time slot.