Vehicle Velocity Profiles With Two-Level Route Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for vehicle speed control, particularly in autonomous vehicles, face challenges in integrating full-route optimization with local optimizations, leading to sub-optimal solutions due to reliance on heuristics and inequality-based constraints, which are computationally inefficient and result in strong variations.
Innovation Solution
A two-level optimization approach involving full-horizon and short-horizon optimizations, where parameter values from the full-horizon optimization are used to constrain short-horizon minimizations, ensuring smooth and coherent speed profiles without inequality constraints, using a model predictive control algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If inequality-based constraints are used to integrate full-route optimization with local optimizations, then the integration is achieved, but computational efficiency deteriorates and strong variations occur in local solutions
Solution Approach 1:
The patent transforms the constrained optimization problem into an unconstrained one by changing the parameter representation. Instead of using inequality constraints to enforce full-route optimization requirements, the invention uses parameter values (Lagrange multipliers) obtained from the full-horizon optimization to define a cost function for local optimizations. This parameter transformation eliminates computational inefficiency while maintaining integration coherence.
Solution Approach 2:
The patent replaces the mechanical constraint system (inequality-based constraints) with a functional substitution approach. The full-route optimization requirements are substituted into the local optimization cost function as parameterized objectives, eliminating the need for explicit constraint enforcement and improving computational efficiency while maintaining solution coherence.
2Device complexity
If heuristics are used to determine parameters in multi-objective schemes, then computational complexity is reduced, but solution optimality deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where parameter values from the full-horizon optimization are fed into the local optimization cost function. This feedback loop ensures that local optimizations remain aligned with global objectives without requiring heuristics, thereby maintaining solution optimality while managing computational complexity through the structured two-level approach.
Solution Approach 2:
The patent performs preliminary full-horizon optimization to obtain parameter values before executing local optimizations. This preliminary action provides optimal parameter settings that guide subsequent local optimizations, eliminating the need for heuristics and ensuring solution optimality is maintained throughout the multi-stage process.
3Reliability
If terminal energy and velocity constraints are imposed from high-level optimization to low-level optimization, then integration is achieved, but local solution stability deteriorates due to strong variations
Solution Approach 1:
The patent changes the approach from imposing hard terminal constraints to using parameter values (Lagrange multipliers) from high-level optimization to parameterize the low-level cost function. This parameter transformation smooths the integration process, maintaining solution stability while achieving integration coherence without causing strong variations in local solutions.
Data Source
AI summary
The invention is notably directed to a computer-implemented method for controlling a speed of a vehicle given energy-related quantities that pertain to moving the vehicle. Each of said energy-related quantities is being referred to as an energy expense in the following. The method involves two key operations (a full-horizon optimization and multiple, nested short-horizon optimizations), which are performed based on route parameters of a route segment and real-time signals capturing current states of the vehicle. First, a target function of a profile of an energy expense is optimized (S12) for a route segment, to determine a reference profile X*ref of this energy expense. The optimization is constrained with respect to objectives in respect of a total travel time and/or a total travel energy expense for the vehicle to travel said route segment. A profile of parameter values is obtained (S14-S28) according to a by-product of the optimization. Next, short-horizon speed profiles Vref of the vehicle are repeatedly determined (S32) to plan speeds Vref of the vehicle along local sections of the route segment. Such predictions are achieved by minimizing, for each of the local sections, a cost function as parameterized by one of the parameter values. The cost function captures a cost of deviating from an output of the optimized target function due to such speed profiles; said output and said one of the parameter values pertain, each, to said each of the local sections. Finally, control signals are obtained based on the planned speeds Vref. Such control signals can be used for controlling (S40) a speed of the vehicle along each of said local sections of the route segment. The invention is further directed to related systems, vehicles, and computer program products.