Logistics Robot Routing Using ML-Based Route Issue Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Logistics operations face challenges in autonomously completing the last mile delivery due to the limitations of various robot types, such as legged, wheeled, and humanoid robots, which struggle with terrain, energy efficiency, and public acceptance, leading to unexpected issues that impact delivery times and costs.

Innovation Solution

A machine learning (ML) model is trained to associate map objects with route issues encountered by logistics robots, considering robot type, energy consumption, and dynamic conditions to optimize route planning and selection, using a logistics platform and database to identify potential obstacles and adjust routes accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If different robot types (legged, wheeled, humanoid) are deployed for logistics routes, then the robots can navigate different terrains and perform specific tasks, but each robot type has limitations (energy consumption, navigation capability, complexity, cost) that reduce overall system reliability

Engineering Contradiction:
Improverobot navigation capabilityVSAvoiddelivery completion reliability
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The system dynamically selects robot types and routes based on real-time environmental conditions, terrain characteristics, and robot performance data. The routing approach transitions from static to dynamic, allowing the system to adapt robot deployment to changing conditions and avoid situations that would cause failure, thereby improving reliability while maintaining versatility.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system implements feedback loops where robot performance data, route issues, and environmental conditions are continuously collected and used to improve future routing decisions. This feedback mechanism allows the system to learn from past experiences and optimize robot deployment, reducing failures and improving delivery reliability.

Inventive Principle:
Principle #23Feedback

2Productivity

If traditional routing approaches are used for logistics robots, then route planning is simple, but unexpected issues occur that impact delivery times and costs

Engineering Contradiction:
Improvedelivery speedVSAvoidroute completion reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The system performs preliminary analysis of route conditions and robot capabilities before deployment, identifying potential issues in advance. By pre-assessing terrain complexity, environmental conditions, and robot suitability, the system can prevent route failures before they occur, improving both delivery speed and reliability.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system continuously collects feedback on route issues and robot performance, using this information to improve future routing decisions. This iterative learning process enables the system to anticipate and avoid problems, reducing delivery delays and improving route completion reliability.

Inventive Principle:
Principle #23Feedback

3Reliability

If sophisticated routing approaches are implemented to improve delivery reliability, then route planning complexity increases, but unexpected issues still occur

Engineering Contradiction:
Improvedelivery completion reliabilityVSAvoidrouting system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system employs a universal routing framework that can handle multiple robot types, terrain conditions, and delivery scenarios through a single integrated approach. This multi-functional system reduces the need for separate complex routing algorithms for different situations, maintaining reliability while managing overall system complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses machine learning models that automatically learn from data and improve routing decisions without requiring manual configuration or complex rule-based systems. This self-learning capability enables the system to achieve high reliability through adaptive intelligence rather than through manually programmed complexity.

Inventive Principle:
Principle #25Self-service

4Productivity

If robot deployment is increased to meet growing delivery demand, then delivery capacity improves, but energy consumption and operational costs increase

Engineering Contradiction:
Improvedelivery capacityVSAvoidrobot energy consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The system optimizes energy consumption by dynamically changing operational parameters such as robot selection, route timing, and traversal speed based on environmental conditions and delivery priorities. This parameter optimization allows the system to maintain high delivery capacity while minimizing energy usage per delivery.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system dynamically adjusts robot deployment strategies based on real-time conditions, selecting the most energy-efficient robot types and routes for each delivery task. This dynamic optimization enables the system to scale delivery capacity without proportionally increasing energy consumption.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4686998A1Method, apparatus and computer program product for logistics robot deployment
Publication Date: 2026.02.04 HERE GLOBAL BV
  • EP4686998A1 patent drawingFigure 1
  • EP4686998A1 patent drawingFigure 2
  • EP4686998A1 patent drawingFigure 3a~3b

AI summary

Method, apparatuses and computer program products for training machine learning models for logistics robots and routing the robots are disclosed. A method of training an ML model involves providing a route to a logistics robot, obtaining route issue indications from the robot when it fails to traverse the route as expected, and associating map objects with the issue locations. The trained model can then be used to determine the likelihood of route issues for a specific logistics robot type based on the presence of certain map objects along the route. The disclosure further involves calculating route penalty value(s) based on the likelihood and updating the route accordingly, including selection of a logistics robot type for the route.