Policy gradient based Reinforcement learning control design of an electro-pneumatic gearbox actuator

Bécsi, Tamás; Aradi, Szilárd; Szabo, Adam; Gáspár, Péter

doi:10.1016/j.ifacol.2018.11.577

Cited by 7 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Gear-change is an event-driven task, and the shift has a given time limit, and a well-defined goal, which makes this problem ideal for Reinforcement Learning (RL) based algorithms. Previous research showed promising results in the field [31], [32], leading to more extensive research presented in this article.…”

Section: Introductionmentioning

confidence: 81%

Reinforcement Learning Based Control Design for a Floating Piston Pneumatic Gearbox Actuator

et al. 2020

View full text Add to dashboard Cite

The project is also supported by the Hungarian Government and co-financed by the European Social Fund, EFOP-3.6.3-VEKOP-16-2017-00001: Talent management in autonomous vehicle control technologies ABSTRACT Electro-pneumatic actuators play an essential role in various areas of the industry, including heavy-duty vehicles. This article deals with the control problem of an Automatic Manual Transmission, where the actuator of the system is a double-acting floating-piston cylinder, with dedicated inner-position. During the control design of electro-pneumatic cylinders, one must implement a set-valued control on a nonlinear system, when, as in the present case, non-proportional valves provide the airflow. As both the system model itself and the qualitative control goals can be formulated as a Partially Observable Markov Decision Process, Machine learning frameworks are a conspicuous choice for handling such control problems. To this end, six different solutions are compared in the article, of which a new agent named PG-MCTS, using a modified version of the "Upper Confidence bound for Trees" algorithm, is also presented. The performance and strategic choice comparison of the six methods are carried out in a simulation environment. Validation tests performed on an actual transmission system and implemented on an automotive control unit to prove the applicability of the concept. In this case, a Policy Gradient agent, selected by implementation and computation capacity restrictions. The results show that the presented methods are suitable for the control of floating-piston cylinders and can be extended to other fluid mechanical actuators, or even different set-valued nonlinear control problems.

show abstract

Section: Introductionmentioning

confidence: 81%

Reinforcement Learning Based Control Design for a Floating Piston Pneumatic Gearbox Actuator

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Reinforcement learning have just become popular thanks to the extraordinary results in video games, 20 robotics, 21 and other control tasks. 22 In contrast to supervised learning, it has serious advantages: it does not require a huge number of labeled training data, which is expensive to create, and in case of RL the reachable result is not limited. 23 Although RL also has some drawbacks, like struggling with convergence in the training phase, or lack of robustness and reliability.…”

Section: Machine Learning Based Solutionmentioning

confidence: 99%

“…23 Although RL also has some drawbacks, like struggling with convergence in the training phase, or lack of robustness and reliability. 24…”

Section: Machine Learning Based Solutionmentioning

confidence: 99%

Area Scanning with Reinforcement Learning and MCTS in Smart City Applications

Haraszti¹,

Kővári²,

Kolat³

et al. 2020

RepTudKoz

View full text Add to dashboard Cite

This research project is focused on area scanning in the scope of smart city applications with unmanned aerial vehicles or UAVs. More powerful devices have been designed in terms of range, capacity and sensory capabilities in the recent years. This makes possible easier automation, thus suppressing the need for human resources. Some of the fields of applications include: traffic or pollution monitoring, land surveying, civil security control or natural disaster control and monitoring. With the increased number of UAV applications, the use and development of efficient algorithms is more and more essential. This paper investigates the possibility of using Monte-Carlo Tree Search (MCTS) and Reinforcement Learning (RL) in this area, which are already successful methods in other control tasks.

show abstract

Validation Environment for Deep Reinforcement Learning Based Gear Shift Controllers

Altenburg

Bause²,

Albers³

2022

Proceedings

View full text Add to dashboard Cite

Policy gradient based Reinforcement learning control design of an electro-pneumatic gearbox actuator

Cited by 7 publications

References 18 publications

Reinforcement Learning Based Control Design for a Floating Piston Pneumatic Gearbox Actuator

Reinforcement Learning Based Control Design for a Floating Piston Pneumatic Gearbox Actuator

Area Scanning with Reinforcement Learning and MCTS in Smart City Applications

Validation Environment for Deep Reinforcement Learning Based Gear Shift Controllers

Contact Info

Product

Resources

About