Jelena Vasiljević
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jelena Vasiljević has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Miloš Jovanović
Technical Test Center, Ministry of Defense, 11000 Belgrade, Serbia
3 shared publications - Aleksandar Pluškoski
School of Computing, University Union, 11000 Belgrade, Serbia
2 shared publications - Igor Ciganović
School of Computing, University Union, 11000 Belgrade, Serbia
2 shared publications - Dragan Lazić
Technical Test Center, Belgrade, Ministry of Defense Republic of Serbia, 11000 Belgrade, Serbia
1 shared publication - Igor Medenica
Faculty of Computer Science, University Union, 11000 Belgrade, Serbia
1 shared publication - Nikola Radulović
Faculty of Natural Sciences, University of Novi Sad, 21000 Novi Sad, Serbia
1 shared publication
Latest Publications
- Open Access
- 19 July 2026
- Published by MDPI AG in Electronics
- Vol. 15 (14) , 3170
Evaluation of Multi-Model Architecture Against Single-Model PPO in the MiniGrid Environment
Reward sparsity is a considerable problem in reinforcement learning when solving complex problems. In an environment where the reward is given only after successfully solving a problem in multiple steps, the signals for optimization can be too weak, preventing the agent from learning. However, this can be solved by breaking the problem into an array of simpler tasks that the agent can learn to perform. This paper presents a possible solution that uses a mixture-of-experts architecture. Separate models are used for planning, routing expert outputs, and solving singular sub-tasks. The solution was implemented in the MiniGrid environment, which allows for the design of arbitrarily complex problems. The results presented in this paper show that a multi-model agent can solve more complex problems comprising multiple steps than a single reinforcement learning agent.
- 19 July 2026
- Published by MDPI AG in Electronics
- Vol. 15 (14) , 3170
Open AccessEvaluation of Multi-Model Architecture Against Single-Model PPO in the MiniGrid Environment
Reward sparsity is a considerable problem in reinforcement learning when solving complex problems. In an environment where the reward is given only after successfully solving a problem in multiple steps, the signals for optimization can be too weak, preventing the agent from learning. However, this can be solved by breaking the problem into an array of simpler tasks that the agent can learn to perform. This paper presents a possible solution that uses a mixture-of-experts architecture. Separate models are used for planning, routing expert outputs, and solving singular sub-tasks. The solution was implemented in the MiniGrid environment, which allows for the design of arbitrarily complex problems. The results presented in this paper show that a multi-model agent can solve more complex problems comprising multiple steps than a single reinforcement learning agent.
- Open Access
- 1 May 2026
- Published by MDPI AG in Electronics
- Vol. 15 (9) , 1921
Benefits of Using LLM for Long-Term Planning with Distilled Subtask Model Compared to End-to-End Reinforcement Learning in the MiniGrid Simulator
Policy learning under delayed reward conditions remains a significant challenge for end-to-end reinforcement learning (RL) agents. The difficulty increases for problems that require long-term planning and the execution of multiple dependent subtasks. As a result, solutions based on a single monolithic policy often suffer from unstable training. One possible solution to this problem could be to delegate the long-term planning to a separate model. This paper presents an implementation comprising two models: a large language model (LLM) for long-term planning and an execution model that solves subtasks. The execution model was trained via distillation from multiple teacher models trained with RL on individual tasks. The results presented in this paper demonstrate the benefits of this approach. By delegating long-term planning to the LLM, the agent can solve more complex problems than end-to-end agents trained with the proximal policy optimization (PPO) algorithm.
- 1 May 2026
- Published by MDPI AG in Electronics
- Vol. 15 (9) , 1921
Open AccessBenefits of Using LLM for Long-Term Planning with Distilled Subtask Model Compared to End-to-End Reinforcement Learning in the MiniGrid Simulator
Policy learning under delayed reward conditions remains a significant challenge for end-to-end reinforcement learning (RL) agents. The difficulty increases for problems that require long-term planning and the execution of multiple dependent subtasks. As a result, solutions based on a single monolithic policy often suffer from unstable training. One possible solution to this problem could be to delegate the long-term planning to a separate model. This paper presents an implementation comprising two models: a large language model (LLM) for long-term planning and an execution model that solves subtasks. The execution model was trained via distillation from multiple teacher models trained with RL on individual tasks. The results presented in this paper demonstrate the benefits of this approach. By delegating long-term planning to the LLM, the agent can solve more complex problems than end-to-end agents trained with the proximal policy optimization (PPO) algorithm.
- Open Access
- 6 August 2025
- Published by MDPI AG in Electronics
- Vol. 14 (15) , 3127
Optimization of Delay Time in ZigBee Sensor Networks for Smart Home Systems Using a Smart-Adaptive Communication Distribution Algorithm
As smart homes and Internet of Things (IoT) ecosystems continue to expand, the need for energy-efficient and low-latency communication has become increasingly critical. One of the key challenges in these systems is minimizing delay time while ensuring reliable and efficient communication between devices. This study focuses on optimizing delay time in ZigBee sensor networks used in smart-home systems. A Smart–Adaptive Communication Distribution Algorithm is proposed, which dynamically adjusts the communication between network nodes based on real-time network conditions. Experimental measurements were conducted under various scenarios to evaluate the performance of the proposed algorithm, with a particular focus on reducing delay and enhancing overall network efficiency. The results demonstrate that the proposed algorithm significantly reduces delay times compared to traditional methods, making it a promising solution for delay-sensitive IoT applications. Furthermore, the findings highlight the importance of adaptive communication strategies in improving the performance of ZigBee-based sensor networks.
- 6 August 2025
- Published by MDPI AG in Electronics
- Vol. 14 (15) , 3127
Open AccessOptimization of Delay Time in ZigBee Sensor Networks for Smart Home Systems Using a Smart-Adaptive Communication Distribution Algorithm
As smart homes and Internet of Things (IoT) ecosystems continue to expand, the need for energy-efficient and low-latency communication has become increasingly critical. One of the key challenges in these systems is minimizing delay time while ensuring reliable and efficient communication between devices. This study focuses on optimizing delay time in ZigBee sensor networks used in smart-home systems. A Smart–Adaptive Communication Distribution Algorithm is proposed, which dynamically adjusts the communication between network nodes based on real-time network conditions. Experimental measurements were conducted under various scenarios to evaluate the performance of the proposed algorithm, with a particular focus on reducing delay and enhancing overall network efficiency. The results demonstrate that the proposed algorithm significantly reduces delay times compared to traditional methods, making it a promising solution for delay-sensitive IoT applications. Furthermore, the findings highlight the importance of adaptive communication strategies in improving the performance of ZigBee-based sensor networks.