Autonomous Driving Topological Planning With Central Bound Pruning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing autonomous vehicle navigation systems face computational inefficiencies and memory inefficiencies in determining optimal trajectories due to the exponential complexity of topological tree expansion, leading to suboptimal or infeasible lane change and in-lane maneuvers, especially in dense traffic conditions.

Innovation Solution

A computer-implemented method using topological planning with bounds representations, involving a topological tree expansion that includes generating and storing bounds in a centralized storage, pruning nodes based on feasibility, strict redundancy, and fuzzy redundancy, and using sample index intervals to reduce memory usage and computational load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If the topological tree is expanded to consider all possible lane change variations (transition start/end locations, times, steering/speed profiles), then the completeness of trajectory evaluation is improved, but the computational cost becomes infinitely expensive and computationally infeasible

Engineering Contradiction:
Improvecompleteness of trajectory evaluationVSAvoidcomputational cost
Core Design Contradiction:
ReliabilityVSPower

Solution Approach 1:

The patent segments the continuous space of possible trajectories into discrete topological elements (nodes representing decision points, edges representing maneuver options). This segmentation allows the system to evaluate trajectories at critical decision points rather than continuously across all possible variations, reducing computational cost while maintaining evaluation completeness.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary action by pre-computing and storing bounds representations for each topological element before actual trajectory evaluation. These bounds (minimum and maximum values for parameters like longitudinal position, lateral position, speed, acceleration) are calculated in advance and stored in data structures, enabling efficient pruning during real-time planning without re-computing all possibilities.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 3:

The patent applies partial action by evaluating only the necessary subset of trajectory variations that fall within computed bounds, rather than exhaustively evaluating all possible lane change variations. The bounds pruning mechanism eliminates excessive computations by discarding trajectories that cannot satisfy constraints based on pre-computed bound comparisons.

Inventive Principle:
Principle #16Partial or excessive action

2Power

If a finite number of candidate trajectories are sampled through discretization or random sampling, then the computational cost is reduced, but the system may fail to find feasible or optimal trajectories that exist in the continuous space

Engineering Contradiction:
Improvecomputational costVSAvoidfeasibility of found trajectory
Core Design Contradiction:
PowerVSReliability

Solution Approach 1:

The patent changes parameters by representing trajectories not as discrete samples but as continuous functions bounded by parameter ranges (bounds). Each topological element has associated bounds for longitudinal position, lateral position, speed, acceleration, and other parameters. This parameter-based representation allows the system to efficiently query whether feasible trajectories exist within bound constraints without exhaustive sampling.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces bounds representations as an intermediary between the continuous trajectory space and the discrete evaluation process. These bounds act as a mediator that captures the essential constraints of continuous trajectories in a computationally tractable form, allowing the system to reason about continuous space using discrete bound comparisons rather than continuous optimization.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If the system stores vast amounts of information defining the roadway and calculates cost functions for each constraint in electronic memory, then the accuracy of trajectory evaluation is improved, but the memory usage and processing time increase significantly

Engineering Contradiction:
Improveaccuracy of trajectory evaluationVSAvoidmemory usage
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential constraint information needed for trajectory evaluation and stores it in optimized data structures. Rather than storing complete roadway definitions and all possible trajectory variations, the system extracts and stores only the bounds representations (minimum and maximum values for critical parameters) for each topological element, significantly reducing memory usage while preserving evaluation accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs preliminary action by pre-calculating and storing bounds representations for all topological elements before actual trajectory evaluation. This pre-computation phase extracts and stores only the necessary constraint information in an optimized format, eliminating the need to store and process vast amounts of raw roadway data during real-time operation.

Inventive Principle:
Principle #10Preliminary action

4Productivity

If the system prunes nodes based on feasibility, strict redundancy, and fuzzy redundancy, then the computational efficiency is improved, but the complexity of the pruning algorithms increases

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidalgorithm complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent segments the pruning process into three distinct phases: feasibility pruning (checking basic constraint satisfaction), strict redundancy pruning (removing identical or dominated trajectories), and fuzzy redundancy pruning (removing nearly identical trajectories based on tolerance thresholds). Each phase operates on segmented subsets of trajectories with specialized algorithms, making the overall complex pruning process more manageable and efficient.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by implementing pruning in hierarchical stages, applying only the necessary level of pruning for each situation. Feasibility pruning is applied first to eliminate obviously infeasible trajectories, then strict redundancy pruning for exact duplicates, and finally fuzzy redundancy pruning only when needed to reduce remaining candidates. This staged approach avoids the excessive complexity of implementing all pruning methods simultaneously.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP4291457B1Vehicle, method, and computer program product for topological planning in autonomous driving using bounds representations
Publication Date: 2025.11.19 VOLKSWAGEN GROUP OF AMERICA INVESTMENTS LLC
  • EP4291457B1 patent drawingFigure 1A
  • EP4291457B1 patent drawingFigure 1B
  • EP4291457B1 patent drawingFigure 2

AI summary

Provided are autonomous vehicles and methods of controlling autonomous vehicles through topological planning with bounds, including receiving map data and sensor data, expanding a topological tree by adding a plurality of nodes to represent a plurality of actions associated with the plurality of constraints, generating a bound based on a constraint in the geographic area, the bound associated with an action for navigating the autonomous vehicle relative to the at least one constraint, storing the bound in a central bound storage, linking a set of bounds of a tree node to the bound via a bound identifier, wherein the first bound is initially linked as an active bound, or alternatively, as an inactive bound after determining it is not the most restrictive bound at any sample index, and control the autonomous vehicle based on the topological tree, to navigate the plurality of constraints.