Autonomous Vehicle Predictive Planning for Fuel Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicle planning systems primarily focus on the lead vehicle, neglecting vehicles ahead, leading to excessive braking and throttling, which results in fuel inefficiency, actuator subsystem deterioration, and compromised driving comfort due to inadequate consideration of the entire traffic situation.
Innovation Solution
The system predicts the movement of lead and subsequent obstacles using a traffic model that accounts for the movement of all relevant vehicles, optimizing the ego vehicle's movement by minimizing an objective function that includes safety, comfort, and cost terms, thereby determining an optimized preferred following distance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the autonomous system focuses only on the lead vehicle for planning, then the planning complexity is reduced, but excessive braking and throttling occur leading to fuel inefficiency and actuator deterioration
Solution Approach 1:
The system performs preliminary prediction of the lead vehicle's movement using a traffic model that incorporates data from multiple vehicles ahead. This prediction is done in advance before the actual planning decision, allowing the system to anticipate traffic flow changes and plan smoother acceleration and braking profiles, thereby reducing energy loss from excessive actuator operations
Solution Approach 2:
The traffic model acts as an intermediary between raw sensor data and the planning function. It processes historical movement data from the lead vehicle and vehicles ahead to generate predicted movement trajectories, which then feed into the optimization function. This intermediary layer enables more accurate energy-efficient planning without significantly increasing overall system complexity
2Loss of energy
If the system considers the entire traffic situation including vehicles ahead of the lead obstacle, then fuel efficiency and driving comfort improve, but the computational complexity and data processing requirements increase
Solution Approach 1:
The system segments the traffic environment into distinct components: the ego vehicle, the lead vehicle, and vehicles ahead of the lead vehicle. Each segment is processed separately by the traffic model using historical movement data, and their predictions are integrated into the overall planning optimization. This segmentation allows comprehensive traffic consideration while managing computational complexity through modular processing
Solution Approach 2:
The system changes the parameter representation from raw sensor data to predicted movement trajectories generated by the traffic model. By transforming multiple vehicles' positional and velocity data into predicted future states, the system reduces the dimensionality of the optimization problem while retaining essential traffic flow information, thus balancing comprehensiveness with computational feasibility
3Measurement precision
If the system uses a learning-based algorithm to predict lead obstacle movement, then the prediction accuracy improves, but the computational time and processing requirements increase
Solution Approach 1:
The learning-based traffic model is trained offline in advance using historical traffic data. This preliminary training phase allows the model to learn complex traffic patterns and vehicle behaviors without real-time computational burden. During actual autonomous operation, the pre-trained model generates predictions quickly by applying learned patterns to current sensor data, thus achieving high accuracy without excessive real-time computational time
Solution Approach 2:
The system uses a learning-based model that processes more data than strictly necessary for basic prediction (excessive action during training), but this investment pays off during real-time operation where the model delivers accurate predictions efficiently. The offline training phase accepts higher computational costs to enable fast, accurate predictions during actual vehicle operation
Data Source
AI summary
Methods, systems, and non-transitory computer-readable media are configured to perform operations comprising acquiring data regarding historical movement of a lead obstacle and one or more obstacles in front of the lead obstacle; predicting movement of at least the lead obstacle based at least in part on the data; and applying the predicted movement of the lead obstacle to an optimizable function associated with movement of an ego vehicle.