Supun Gayantha Dissanayaka

Mr. Supun Gayantha Dissanayaka

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 August 2025 in Electronics
Verified
Journal Article
Open access
PDF file
WDWorrawat Duanyai
Worrawat Duanyai
WSWeon Keun Song
Weon Keun Song
MKMin-Ho Ka
Min-Ho Ka
DLDONG WOOK Lee
DONG WOOK Lee
SDSupun Gayantha Dissanayaka
Supun Gayantha Dissanayaka
A novel adaptive intelligent control system (AICS) with learning-while-controlling capability is developed for a highly nonlinear single-input single-output plant by redesigning the conventional model reference adaptive control (MRAC) framework, originally based on first-order Lyapunov stability, and employing customized neural networks. The AICS is designed with a simple structure, consisting of two main subsystems: a meta-learning-triggered mechanism-based physics-informed neural network (MLTM-PINN) for plant identification and a self-tuning neural network controller (STNNC). This structure, featuring the triggered mechanism, facilitates a balance between high controllability and control efficiency. The MLTM-PINN incorporates the following: (I) a single self-supervised physics-informed neural network (PINN) without the need for labelled data, enabling online learning in control; (II) a meta-learning-triggered mechanism to ensure consistent control performance; (III) transfer learning combined with meta-learning for finely tailored initialization and quick adaptation to input changes. To resolve the conflict between streamlining the AICS’s structure and enhancing its controllability, the STNNC functionally integrates the nonlinear controller and adaptation laws from the MRAC system. Three STNNC design scenarios are tested with transfer learning and/or hyperparameter optimization (HPO) using a Gaussian process tailored for Bayesian optimization (GP-BO): (scenario 1) applying transfer learning in the absence of the HPO; (scenario 2) optimizing a learning rate in combination with transfer learning; and (scenario 3) optimizing both a learning rate and the number of neurons in hidden layers without applying transfer learning. Unlike scenario 1, no quick adaptation effect in the MLTM-PINN is observed in the other scenarios, as these struggle with the issue of dynamic input evolution due to the HPO-based STNNC design. Scenario 2 demonstrates the best synergy in controllability (best control response) and efficiency (minimal activation frequency of meta-learning and fewer trials for the HPO) in control.

30 September 2024 in Engineering and Technology Horizons
Verified
Journal Article
Open access
PDF file
SDSupun Gayantha Dissanayaka
Supun Gayantha Dissanayaka
PSPitikhate Sooraksa
Pitikhate Sooraksa
SKSomyot Kaitwanidvilai
Somyot Kaitwanidvilai
JMJohn Morris
John Morris
A simple stereo system can be constructed from a single camera using a prism in the optical path to provide the required two views of a system. The simplicity of these systems has several advantages, particularly if the target is an underwater robot, where compact size and ability to seal the optical components are key factors. However, dispersion by the prism, in addition to the lens distortion, makes calibration challenging. By using a model-free approach, we were able to calibrate a prism-based stereo system effectively. We also aimed to use readily available 45° prisms, which present significant dispersion in the system, but retain simplicity and reduce cost, compared to custom low angle prisms. Modern LEDs provide high intensity, low bandwidth light sources and we used a set of three sources, roughly centered on the RGB channels of a readily available commercial camera. Our system used a circular target pattern covering the binocularly visible region in the scene and collected sets of images at known distances, using three separate light sources. From these images, we generated two look-up tables, one for each pixel in the image and a disparity derived by matching corresponding points, Cp(u,v,du), which has three dimensions, and another look-up table, which has a single dimension, Cz(z), so are not quite large, and not beyond the memory capability of even small modern camera systems, but provide fast, O(1), lookup times, suitable for real-time systems. Our calibration strategy enables a simple stereo system built from a single camera to measure depths in a scene: the single camera requires no electronic synchronization and is built from a single, inexpensive, and readily available optical component – a right-angle prism.

15 July 2025 in arXiv
Verified
Preprint
Open access
PDF file
SDSupun Gayantha Dissanayaka
Supun Gayantha Dissanayaka
Alexander FerreinAlexander Ferrein
Alexander Ferrein
THTill Hofmann
Till Hofmann
KNKosuke Nakajima
Kosuke Nakajima
MSMario Sanz-Lopez
Mario Sanz-Lopez
SYShohei Yasuda
Shohei Yasuda
The RoboCup Logistics League is a RoboCup competition in a smart factory scenario that has focused on task planning, job scheduling, and multi-agent coordination. The focus on production logistics allowed teams to develop highly competitive strategies, but also meant that some recent developments in the context of smart manufacturing are not reflected in the competition, weakening its relevance over the years. In this paper, we describe the vision for the RoboCup Smart Manufacturing League, a new competition designed as a larger smart manufacturing scenario, reflecting all the major aspects of a modern factory. It will consist of several tracks that are initially independent but gradually combined into one smart manufacturing scenario. The new tracks will cover industrial robotics challenges such as assembly, human-robot collaboration, and humanoid robotics, but also retain a focus on production logistics. We expect the reenvisioned competition to be more attractive to newcomers and well-tried teams, while also shifting the focus to current and future challenges of industrial robotics.

18 December 2025
Verified
Conference Paper
Closed access
PAPapinwich Asnapetch
Papinwich Asnapetch
SDSupun Gayantha Dissanayaka
Supun Gayantha Dissanayaka
DXDexin Xu
Dexin Xu
PKPoom Konghuayrob
Poom Konghuayrob
The state-of-the-art legged robot localization algorithms typically rely on both proprioceptive data and exteroceptive systems such as LiDAR, vision, and GNSS, fused using smoothing or filtering methods like Factor Graph Optimization or the Extended Kalman Filter (EKF). However, such methods are limited to environments with distinct features or robots containing specific hardware, specifically high quality GNSS receiver. This paper presents a Factor-graph-based localization framework that uses only proprioceptive sensing IMU and motor encoder kinematics - augmented with magnetometer yaw observations. The method is implemented on a Unitree Go2 quadrupedal robot in NVIDIA Isaac Sim and evaluated across five trajectories. The results show that the proposed approach achieves stable, drift-minimized localization, with a maximum deviation of less than 4% of the total trajectory length.