Papers by Palina Bartashevich
2017 IEEE Symposium Series on Computational Intelligence (SSCI), 2017
This paper presents the Navigation Wind PSO (NW-PSO) as a search mechanism for aerial micro-robot... more This paper presents the Navigation Wind PSO (NW-PSO) as a search mechanism for aerial micro-robots with limited battery capacity acting in environments with unknown external dynamics (such as wind). The proposed method uses the concepts of multi-criteria decision making and let the individuals to decide on their movements almost following the flow with small course deviations in order to save as much energy as possible. One of the goals is to investigate how the arising premature energyloss on individual level affects the performance of collective search. The experiments show that NW-PSO can save a good amount of energy as well as perform search behavior with good approximation of the global optimal solution almost without awareness regarding disturbance factors and particle loss.
Proceedings of the Genetic and Evolutionary Computation Conference Companion, 2018
This paper investigates the possibility of evolving new particle swarm equations representing a c... more This paper investigates the possibility of evolving new particle swarm equations representing a collective search mechanism, acting in environments with unknown external dynamics, using Geometric Semantic Genetic Programming (GSGP). The proposed method uses a novel initialization technique-the Evolutionary Demes Despeciation Algorithm (EDDA)-which allows to generate solutions of smaller size than using the traditional ramped halfand-half algorithm. We show that EDDA, using a mixture of both GP and GSGP mutation operators, allows us to evolve new search mechanisms with good generalization ability. CCS CONCEPTS • Theory of computation → Design and analysis of algorithms; Bio-inspired optimization;
Lecture Notes in Computer Science, 2018
This paper presents a Vector Field Benchmark (VFB) generator to study and evaluate the performanc... more This paper presents a Vector Field Benchmark (VFB) generator to study and evaluate the performance of collective search algorithms under the influence of unknown external dynamic environments. The VFB generator is inspired by nature (simulating wind or flow) and constructs artificially dynamic environments based on time-dependent vector fields with moving singularities (vortices). Some experiments using the Particle Swarm Optimization (PSO) algorithm, along with two specially developed updating mechanisms for the global knowledge about the external environment, are conducted to investigate the performance of the proposed benchmarks.
Decision-making is considered as a vital mechanism that can allow linking multiple autonomous sys... more Decision-making is considered as a vital mechanism that can allow linking multiple autonomous systems (e.g., agents) together to design a more intelligent and capable artificial system than considering each of them in isolation. To model or to engineer such a successful collective autonomous distributed system is a very challenging task, as the resulting collective behaviour is a self-organised and emergent phenomenon arisen from the local interactions between the agents and their environment. It does not rely upon a particular leader and is based only on incomplete and noisy sensory information acquired by single individuals. In this way, the collective decision-making mechanism and the underlying collective information processing represent one of the means of designing autonomy at the global level. The current thesis focuses on the question of how individuals have to integrate the personal noisy assessments of their physical and social surroundings to make accurate collective deci...
In this paper, we study Particle Swarm Optimization (PSO) as a collective search mechanism for in... more In this paper, we study Particle Swarm Optimization (PSO) as a collective search mechanism for individuals (such as aerial micro-robots) which are supposed to search in environments with unknown external dynamics. In order to deal with the unknown disturbance, we present new PSO equations which are evolved using Genetic Programming (GP) with a semantically diverse starting population, seeded by the Evolutionary Demes Despeciation Algorithm (EDDA), that generalizes better than standard GP in the presence of unknown dynamics. The analysis of the evolved equations shows that with only small modifications in the velocity equation, PSO can achieve collective search behavior while being unaware of the dynamic external environment, mimicking the zigzag upwind flights of birds towards the food source.
The performance of self-organized collective decision-making systems highly depends on the intera... more The performance of self-organized collective decision-making systems highly depends on the interactions with the environment. The environmental bias factors can introduce indirect modifications in the behaviour of such systems, however, not all changes are for the worse. In this paper, we show how the isomorphic changes in the environment can improve the performance of the collective decision-making strategies, mostly used in the current state-of-the-art swarm robotics research. The idea is based on the usage of a special kind of an equivalence relation, namely isomorphism, which provides local changes in the environment while preserving the global information. The obtained results indicate that the isomorphic transformations, sharing a certain structure of the environment, can significantly accelerate the consensus time without compromising correctness of the final decision.
This paper studies the role of communication topologies in swarms performing the search under str... more This paper studies the role of communication topologies in swarms performing the search under strong negative influence coming from the unknown external environment affecting the individuals’ movements. The experiments are carried out on two modified versions of PSO, namely Power- and Zigzag-PSO, which act without any preliminary information about external forces modeled by vector fields. We propose four dynamic topologies inspired by the game-theoretic concepts and investigate their performance relative to the ordinarily static ones with regard to convergence and “energy expenses”, reflecting the amount of collective effort needed to eliminate the produced drift. The results reveal that the topology of social connections on its own is not an effective way to cope with the unknown disturbance during the search. However, within the predefined coping mechanism against the disturbance as in the case of Zigzag-PSO, the considered dynamic topologies show the advantage before static ones ...
Progress in Artificial Intelligence
2017 IEEE Congress on Evolutionary Computation (CEC)
This paper presents the Vector Field Map PSO (VFM-PSO) as a collective search algorithm for aeria... more This paper presents the Vector Field Map PSO (VFM-PSO) as a collective search algorithm for aerial micro-robots in environments with unknown external dynamics (such as wind). The proposed method is based on a multi-swarm approach and allows to cope with unknown disturbances arising by the vector fields in which the positions and the movements of the particles are highly affected. VFM-PSO requires gathering the information regarding the vector fields and one of our goals is to investigate the amount of the required information for a successful search mechanism. The experiments show that VFM-PSO can reduce the drift and improves the performance of the PSO algorithm despite incomplete information (awareness) about the structure of considered vector fields.
Swarm Intelligence
Collective perception allows sparsely distributed agents to form a global view on a common spatia... more Collective perception allows sparsely distributed agents to form a global view on a common spatially distributed problem without any direct access to global knowledge and only based on a combination of locally perceived information. However, the evidence gathered from the environment is often subject to spatial correlations and depends on the movements of the agents. The latter is not always easy to control and the main question is how to share and to combine the estimated information to achieve the most precise global estimate in the least possible time. The current article aims at answering this question with the help of evidence theory, also known as Dempster–Shafer theory, applied to the collective perception scenario as a collective decision-making problem. We study eight most common belief combination operators to address the arising conflict between different sources of evidence in a highly dynamic multi-agent setting, driven by modulation of positive feedback. In comparison ...
Progress in Artificial Intelligence
This paper presents nine different visual patterns for a Collective Perception scenario as new be... more This paper presents nine different visual patterns for a Collective Perception scenario as new benchmark problems, which can be used for the future development of more efficient collective decision-making strategies. The experiments using isomorphism and three of the well-studied collective decision-making mechanisms are conducted to validate the performance of the new scenarios. The results on a diverse set of problems show that the real task difficulty lies not only in the quantity ratio of the features in the environment but also in their distributions and the clustering levels. Given this, two new metrics for the difficulty of the task are additionally proposed and evaluated on the provided set of benchmarks.
Scalable Uncertainty Management, 2019
We study the effectiveness of consensus formation in multi-agent systems where there is both beli... more We study the effectiveness of consensus formation in multi-agent systems where there is both belief updating based on direct evidence and also belief combination between agents. In particular, we consider the scenario in which a population of agents collaborate on the best-of-n problem where the aim is to reach a consensus about which is the best (alternatively, true) state from amongst a set of states, each with a different quality value (or level of evidence). Agents' beliefs are represented within Dempster-Shafer theory by mass functions and we investigate the macro-level properties of four wellknown belief combination operators for this multiagent consensus formation problem: Dempster's rule, Yager's rule, Dubois & Prade's operator and the averaging operator. The convergence properties of the operators are considered and simulation experiments are conducted for different evidence rates and noise levels. Results show that a combination of updating on direct evidence and belief combination between agents results in better consensus to the best state than does evidence updating alone. We also find that in this framework the operators are robust to noise. Broadly, Yager's rule is shown to be the better operator under various parameter values, i.e. convergence to the best state, robustness to noise, and scalability.
Uploads
Papers by Palina Bartashevich