Towards Agile Robotics - Learning and Assigning Tasks

Tapio Heikkilä; Niko Känsäkoski; Martin J. Kollingbaum

doi:10.1109/CoDIT70676.2026.11630976

Towards Agile Robotics - Learning and Assigning Tasks

Tapio Heikkilä^*
, Niko Känsäkoski
, Martin J. Kollingbaum

^*Corresponding author for this work

BA5602 Intelligent robotics

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

Abstract

A solution for automated task planning and resource allocation in robot systems is proposed. Deep Reinforcement Learning is used to create task policies as generic task descriptions in the object context based on product CAD models, and modified Active Inference Framework is used to estimate the quality of task execution in the robot context based on robot accuracy and stiffness models. As an application, a robotic assembly task with finding a best robot to execute the task, is introduced.

Original language	English
Title of host publication	2026 12th International Conference on Control, Decision and Information Technologies (CoDIT)
Publisher	IEEE Institute of Electrical and Electronic Engineers
Pages	3104-3109
Number of pages	6
ISBN (Electronic)	979-8-3195-2077-7
DOIs	https://doi.org/10.1109/CoDIT70676.2026.11630976
Publication status	Published - 2026
MoE publication type	A4 Article in a conference publication
Event	12th International Conference on Control, Decision and Information Technologies, CoDIT 2026 - Bari, Italy Duration: 13 Jul 2026 → 16 Jul 2026

Conference

Conference	12th International Conference on Control, Decision and Information Technologies, CoDIT 2026
Country/Territory	Italy
City	Bari
Period	13/07/26 → 16/07/26

Funding

This research was part of the DOMINIC project, funded by the Research Council of Finland and VTT Technical Research Centre of Finland Ltd.

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

Access to Document

10.1109/CoDIT70676.2026.11630976

Cite this

@inproceedings{4e326326e37445818bf80156f8dca3f0,

title = "Towards Agile Robotics - Learning and Assigning Tasks",

abstract = "A solution for automated task planning and resource allocation in robot systems is proposed. Deep Reinforcement Learning is used to create task policies as generic task descriptions in the object context based on product CAD models, and modified Active Inference Framework is used to estimate the quality of task execution in the robot context based on robot accuracy and stiffness models. As an application, a robotic assembly task with finding a best robot to execute the task, is introduced.",

author = "Tapio Heikkil{\"a} and Niko K{\"a}ns{\"a}koski and Kollingbaum, \{Martin J.\}",

year = "2026",

doi = "10.1109/CoDIT70676.2026.11630976",

language = "English",

pages = "3104--3109",

booktitle = "2026 12th International Conference on Control, Decision and Information Technologies (CoDIT)",

publisher = "IEEE Institute of Electrical and Electronic Engineers",

address = "United States",

note = "12th International Conference on Control, Decision and Information Technologies, CoDIT 2026 ; Conference date: 13-07-2026 Through 16-07-2026",

}

Heikkilä, T, Känsäkoski, N & Kollingbaum, MJ 2026, Towards Agile Robotics - Learning and Assigning Tasks. in 2026 12th International Conference on Control, Decision and Information Technologies (CoDIT). IEEE Institute of Electrical and Electronic Engineers, pp. 3104-3109, 12th International Conference on Control, Decision and Information Technologies, CoDIT 2026, Bari, Italy, 13/07/26. https://doi.org/10.1109/CoDIT70676.2026.11630976

Towards Agile Robotics - Learning and Assigning Tasks. / Heikkilä, Tapio; Känsäkoski, Niko; Kollingbaum, Martin J.
2026 12th International Conference on Control, Decision and Information Technologies (CoDIT). IEEE Institute of Electrical and Electronic Engineers, 2026. p. 3104-3109.

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - Towards Agile Robotics - Learning and Assigning Tasks

AU - Heikkilä, Tapio

AU - Känsäkoski, Niko

AU - Kollingbaum, Martin J.

PY - 2026

Y1 - 2026

N2 - A solution for automated task planning and resource allocation in robot systems is proposed. Deep Reinforcement Learning is used to create task policies as generic task descriptions in the object context based on product CAD models, and modified Active Inference Framework is used to estimate the quality of task execution in the robot context based on robot accuracy and stiffness models. As an application, a robotic assembly task with finding a best robot to execute the task, is introduced.

AB - A solution for automated task planning and resource allocation in robot systems is proposed. Deep Reinforcement Learning is used to create task policies as generic task descriptions in the object context based on product CAD models, and modified Active Inference Framework is used to estimate the quality of task execution in the robot context based on robot accuracy and stiffness models. As an application, a robotic assembly task with finding a best robot to execute the task, is introduced.

UR - https://www.scopus.com/pages/publications/105047838568

U2 - 10.1109/CoDIT70676.2026.11630976

DO - 10.1109/CoDIT70676.2026.11630976

M3 - Conference article in proceedings

AN - SCOPUS:105047838568

SP - 3104

EP - 3109

BT - 2026 12th International Conference on Control, Decision and Information Technologies (CoDIT)

PB - IEEE Institute of Electrical and Electronic Engineers

T2 - 12th International Conference on Control, Decision and Information Technologies, CoDIT 2026

Y2 - 13 July 2026 through 16 July 2026

ER -

Towards Agile Robotics - Learning and Assigning Tasks

Abstract

Conference

Funding

UN SDGs

Access to Document

Other files and links

Fingerprint

DOMINIC: Developmental Multi-Robot Systems in Cognitive Manufacturing

Cite this

Towards Agile Robotics - Learning and Assigning Tasks

Abstract

Conference

Funding

UN SDGs

Access to Document

Other files and links

Fingerprint

Projects

DOMINIC: Developmental Multi-Robot Systems in Cognitive Manufacturing

Cite this