Work Vehicle Track Planning with Route-Guided Penalty Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing track calculation methods for autonomous vehicles at work sites face challenges due to high computational complexity and the risk of converging to local solutions, especially when dealing with multi-dimensional models and non-linear restriction conditions, leading to discontinuous tracks.

Innovation Solution

A two-stage approach is employed, where a route search unit reduces the model dimensions to minimize calculation, and a track search unit uses quadratic approximation with penalty functions to ensure continuity, avoiding local solutions and reducing computational load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a multi-dimensional mathematical model is used to search for a track considering position, posture, steering angle, speed, and acceleration, then the accuracy of the track calculation is improved, but the computational complexity increases and the calculation time becomes too long for on-line calculation

Engineering Contradiction:
Improvetrack calculation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the track search process into two distinct stages: route search and track search. The route search uses a simplified model to find the general path, while the track search refines this using a more detailed model. This segmentation allows the system to benefit from both simplified and complex models without incurring the full computational burden of using the complex model throughout.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The route search is performed as a preliminary action before the track search. By first finding a rough route using a simplified model, the system narrows down the search space for the subsequent track search, reducing the computational load of the more accurate but computationally intensive track calculation.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If the number of dimensions of the model is reduced to decrease the amount of calculation, then the computational load is reduced, but the obtained track may become discontinuous

Engineering Contradiction:
Improvecalculation speedVSAvoidtrack continuity
Core Design Contradiction:
ProductivityVSStability of the object's composition

Solution Approach 1:

The patent divides the track generation into two segments: route search that ensures overall continuity and track search that refines the path while maintaining continuity. This segmentation allows each stage to focus on specific aspects of track quality without compromising the other.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The track search uses the route from the route search as feedback to guide its optimization. By using the route as a reference and constraint, the track search can refine the path while ensuring it remains continuous and follows the general direction established by the route search.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If a detailed mathematical model with multiple dimensions is used for track search, then the track accuracy is improved, but the calculation may converge to a local solution rather than finding the optimal track

Engineering Contradiction:
Improvetrack accuracyVSAvoidconvergence to global solution
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The two-stage approach segments the optimization process, with the route search providing a global perspective and the track search providing local refinement. This segmentation helps avoid local optima by first establishing a reasonable global path before performing detailed local optimization.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The route search serves as a preliminary action that establishes a reasonable starting point for the track search. By first finding a feasible route, the system provides a good initial solution that guides the subsequent optimization, reducing the risk of converging to poor local solutions.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4290020B1System, method, and work vehicle
Publication Date: 2026.01.21 KOMATSU LTD
  • EP4290020B1 patent drawingFigure 1
  • EP4290020B1 patent drawingFigure 2
  • EP4290020B1 patent drawingFigure 3~4

AI summary

A route search unit searches for a series having elements of a position and posture of a vehicle, which is a route for moving from an initial position to a target position of the vehicle, based on a first restriction condition representing a position of an obstacle. A restriction condition generation unit generates a second restriction condition in which a penalty value increases according to a deviation distance from the route. A track search unit searches for a series having elements of a position, posture, speed, and steering angle of the vehicle, which is a track for moving from the initial position to the target position of the vehicle, based on the second restriction condition.